Asian Development Review: Volume 30, Number 2

Asian Development Review
Volume 30
2013
Number 2
Demographic Dividends Revisited Jeffrey G. Williamson Princelings and Paupers? State Employment and the Distribution of Human Capital Investments Among Households in Viet Nam Ian Coxhead and Diep Phan Foreign Firms and Indigenous Technology Development in the Peoples Republic of China Fredrik Sjholm and Nannan Lundin Dynamics of Household Assets and Income Shocks in the Long-run Process of Economic Development: The Case of Rural Pakistan Takashi Kurosaki Inequality of Human Opportunities in Developing Asia Hyun H. Son Political Connection and Firm Value James S. Ang, David K. Ding, and Tiong Yang Thong
EDITOR Masahiro Kawai EDITORIAL BOARD Chair Changyong Rhee Co-Managing Editors Maria Socorro G. Bautista Mario Lamberte KYM ANDERSON, University of Adelaide PREMACHANDRA ATHUKORALA, Australian National University IWAN AZIS, Cornell University PRANAB K. BARDHAN, University of California, Berkeley SHIN-ICHI FUKUDA, The University of Tokyo JONG-WHA LEE, Korea University MARCUS NOLAND, Peterson Institute for International Economics KEIJIRO OTSUKA, National Graduate Institute for Policy Studies EUSTON QUAH, Nanyang Technological University HYUN SONG SHIN, Princeton University KAR-YIU WONG, University of Washington WING THYE WOO, University of California, Davis CHARLES WYPLOSZ, Graduate Institute of International and Development Studies
HONORARY BOARD Chair Takehiko Nakao MONTEK SINGH AHLUWALIA, Government of India MASAHIKO AOKI, Stanford University PETER DRYSDALE, Australian National University JUSTIN LIN, Peking University MARI ELKA PANGESTU, Republic of Indonesia HAN SEUNG-SOO, Member, UN Secretary-Generals Advisory Board on Water and Sanitation LAWRENCE SUMMERS, Harvard University, John F. Kennedy School of Government
The Asian Development Review is a professional journal for disseminating the results of economic and development research relevant to Asia. The journal seeks high-quality papers done in an empirically rigorous way. Articles are intended for readership among economists and social scientists in government, private sector, academia, and international organizations. The views expressed in this publication are those of the authors and do not necessarily reflect the views and policies of the Asian Development Bank (ADB), the Asian Development Bank Institute (ADBI), ADBs Board of Governors, or the governments they represent. ADB and ADBI do not guarantee the accuracy of the data included in this publication and accept no responsibility for any consequence of their use. By making any designation of or reference to a particular territory or geographic area, or by using the term country in this document, ADB and ADBI do not intend to make any judgments as to the legal or other status of any territory or area. Please direct all editorial correspondence to the Co-Managing Editors, Asian Development Review, Economics and Research Department, Asian Development Bank, 6 ADB Avenue, Mandaluyong City, 1550 Metro Manila, Philippines. E-mail: asiandevreview@adb.org Note: In this publication, $ refers to United States dollars. For more information, please visit the website of the publication at www.adb.org/data/publications/1125
Asian Development Review

Volume 30
September 2013
2013
Number 2
Volume 30
2013
Number 2
1
Demographic Dividends Revisited Jeffrey G. Williamson Princelings and Paupers? State Employment and the Distribution of Human Capital Investments Among Households in Viet Nam Ian Coxhead and Diep Phan Foreign Firms and Indigenous Technology Development in the Peoples Republic of China Fredrik Sjholm and Nannan Lundin Dynamics of Household Assets and Income Shocks in the Long-run Process of Economic Development: The Case of Rural Pakistan Takashi Kurosaki Inequality of Human Opportunities in Developing Asia Hyun H. Son Political Connection and Firm Value James S. Ang, David K. Ding, and Tiong Yang Thong 2013 List of Referees Call for Papers Instructions for Authors
26
49
76
110
131
167 168 169
Demographic Dividends Revisited

JEFFREY G. WILLIAMSON
This paper revisits demographic dividend issues after almost 2 decades of debate. In 1998, David Bloom and I used a convergence model to estimate the impact of demographic-transition-driven age structure effects and calculated what the literature has come to call the demographic dividend. These early estimates seem to be similar to those coming from more recent overlapping generation models, when properly estimated. Research has shown that the demographic dividend is not simply a labor participation rate effect, but also a growth effect. Life-cycle savings, investment deepening, foreign capital ows, and schooling have all been greatly affected by the demographic transition. The paper discusses just how much of these positive growth effects are based on accelerating human capital accumulation induced by demand-side qualityquantity trade-offs versus a co-movement between demographic transitions and public schooling supply-side expansions. Since emigration has been driven in part by demography, it has wasted some of the demographic dividend by brain drain. In addition, within-country ruralurban migrations have also been driven in part by demographic transitions with different spatial timing. Finally, the paper shows how lifetimenot just annualincome inequality has been inuenced by demographic transitions. Keywords: demographic transitions, demographic dividends, growth, inequality, Asia JEL codes: J10, O11, O15, O40, O53
I. Looking Backwards: The Demographic Dividend Convergence Model
Back in the 1950s, 1960s, and 1970s, pessimists believed that rapid population growth in the Third World was immiserizing because it tended to overwhelm the contributions of technical change and capital accumulation (Coale and Hoover 1958, Ehrlich 1968). Optimists believed that rapid population growth helped an economy capture economies of scale from market size and promoted both technological and institutional innovation (Kuznets 1967, Boserup 1981, Simon 1981). Research culminating in the late 1980s defeated both viewspopulation growth was shown to have no signicant impact on economic growth, positive or negative (Kelley
Jeffrey Williamson is the Laird Bell Professor of Economics, emeritus, of Harvard University and an Honorary Fellow of the Economics Department at the University of Wisconsin. This paper draws heavily on my previous publications cited in the text, some of which have been with collaborators who have my thanks: David Bloom, Timothy Hatton, and Matthew Higgins. In addition, the comments of Noel de Dios, John Nye, Xin Meng, Feng Wang, Andy Mason, participants at the Asian Development Review conference held in Manila on 2526 March 2013, and the journals referees are gratefully appreciated.
Asian Development Review, vol. 30, no. 2, pp. 125
C 2013 Asian Development Bank and Asian Development Bank Institute
2 ASIAN DEVELOPMENT REVIEW
1988). These studies were typically based on cross-country regressions of per capita income growth on population growth, controlling for a variety of other inuences. As Allen Kelley and Robert Schmidt (1995, p. 543) put it: Possibly the most inuential statistical nding that has shaped the population debates in recent decades is the failure, in more than a dozen studies using cross-country data, to unearth a statistically signicant association between the growth rates of population and of per capita output. This nding was surprising, but it was unclear then whether it arose because population truly had no effect on economic growth or because the test had been somehow misspecied. Work immediately following decomposed population growth into its fertility and mortality components and examined their independent effects on economic growth (Coale 1986, Bloom and Freeman 1986, Barlow 1994, Brander and Dowrick 1994, Kelley and Schmidt 1995). These studies found that measures of fertility, specically past birth rates, were negatively and signicantly associated with economic growth, whereas the effect of mortality was insignicant. These contributions were the direct precursors to the demographic dividend literature that followed in the wake of the Asian Development Banks Emerging Asia conference (ADB 1997): This line of research justied the decomposition on the grounds that changes in fertility and mortality could imply very different changes in the age distribution. Population growth attributable to a fall in infant mortality and a rise in fertility both had an immediate negative effect on economic growth since this meant more mouths to feed. A fall in mortality everywhere across the age distribution could raise the adult labor force, giving an offsetting positive impact. However, we now understand that the negative demographic effect has a delayed positive impact on economic growth since the economically active population booms 2 decades later, long after the aggregate population growth effect may have disappeared. This positive effect on economic growth abates as the fertility rate declines, but with a long lag. Figure 1 plots the stylized version of this demographic transition, showing the transition from high fertility and high mortality when the country is poor to low fertility and low mortality when the country is rich. The two critical aspects of the transition are rst, that the initial mortality decline is driven primarily by a fall in infant mortality, and second, that fertility rates are very slow to decline in response. The transition takes decades to complete. The population growth rate is implicit in the rst panel of Figure 1 as the difference between fertility and mortality. The second panel makes the population dynamics explicit: the demographic transition is accompanied by a cycle in population growth and an even more dramatic cycle
DEMOGRAPHIC DIVIDENDS REVISITED 3

Figure 1. The Stylized Demographic Transition
Demographic Transition
Birth rate Death rate
Population growth rate
Birth rate Death rate Time
Population Growth and the Age Structure

Share working
Birth rate minus death rate
Growth
Percent in workforce
Time
Source: Bloom and Williamson (1998), Figure 1.
in the age structure. Figure 1 treats the demographic system as if it was closed, and thus it ignores external migration. This assumption will be relaxed later in the paper. The East Asian demographic evidence certainly supports Figure 1 (Bloom and Williamson 1997 and 1998; Feeney and Mason 2001; Oshima and Mason 2001; Lee 2003; Mason 2007a and 2007b; Bloom and Canning 2008; Mason, Lee, and Lee 2010), but the question back in 1997 was just how big the demographic transition impact on economic growth was. Indeed, how much of the East Asian miracle could it explain? David Bloom and I (1997 and 1998) contributed to this stage of the population debate in four ways.1 First, like Allen Kelley and Robert Schmidt (1995), we used the empirical convergence model (Barro 1991 and 1997) to isolate the effects of
1
These results were later conrmed in more detail by Bloom, Canning, and Malaney (2000).
demography. Second, we explored the possibility of reverse causality by using a twostage specication where instruments for population growth were used to account for possible endogeneity. Third, we introduced demography into the growth equations in a theoretically more appealing way than simply by the ad hoc addition of birth and death rates, specically by adding the growth rates of the total population and the economically active population. By doing so, population growth was allowed to affect economic growth by its overall rate and by its age structure. The distinction mattered. Fourth, we used these econometric results to assess the extent to which population dynamics could help account for a signicant portion of East Asias economic miracle. What did we nd? Between 1965 and 1990, the working age population in East Asia grew 2.4% per annum, dramatically faster than the 1.6% rate for the entire population, yielding a 0.8 percentage point differential. The working age population also grew faster than the entire population in Southeast Asia, but the difference was almost half that of East Asia, while in South Asia it was a quarter of the East Asian gure. Combining the coefcients from the estimated growth equations and the growth rates of the working age and total population, Table 1 reports that population dynamics explained between 1.4 and 1.9 percentage points of per capita GDP growth in East Asia (6.11% per annum)the biggest regional effect worldwideor as much as a third of the growth miracle (1.9/6.11 = 0.31).2 If instead the miracle were dened as the difference between current per capita GDP growth (a transitional rate where population dynamics matter) and some steady state of say 2% (when population is also in steady state and has no impact), then population dynamics explained almost half of the miracle (1.9/[6.112] = 0.46). In Southeast Asia, where the fertility decline took place a little later and the infant mortality decline was a little less dramatic, population dynamics still accounted for 0.9 to 1.8 percentage points of economic growth, or again, as much as half of their less impressive miracle (1.8/3.8 = 0.47). The East Asian economies that beneted most from these demographic events were Hong Kong, China; the Republic of Korea; Malaysia; Singapore; Taipei,China; and Thailand. It is no coincidence that these Asian tigers attracted most of Paul Krugmans attention when he asserted that the East Asian miracle was driven mainly by high rates of accumulation and labor force growth (Krugman 1994).3 Based on the coefcients of the estimated convergence model and the UN 2025 demographic projections, Bloom and I (1998) concluded that the future would look quite different (Table 2). In East Asia, per capita gross domestic product (GDP) growth attributable to demographic inuences was projected to be negative between 1990 and 2025, declining from a positive gain of 1.4 to 1.9 percentage points between
2 Eight years later, Kelley and Schmidt (2005) also used the convergence model to estimate a gure of 28% for all Asia in 19601995, quite close to our 31% for East Asia in 19651990. 3 Krugman relied on the results of Alwyn Young (1994a and 1994b) and Jong-Il Kim and Lawrence Lau (1994), but these results were subsequently challenged with much higher total factor productivity growth estimates.
Table 1. Contribution of Demographic Change to Economic Growth, 19651990

Estimated Contribution, 19651990 (4 model specications) (1) 1.04 1.71 1.25 0.66 0.14 0.43 1.03 0.94 0.74 (2) 1.64 1.87 1.81 1.34 1.10 0.52 1.54 1.34 1.14 (3) 0.86 1.60 1.07 0.48 0.07 0.39 0.87 0.81 0.62 (4) 0.73 1.37 0.91 0.41 0.06 0.33 0.74 0.69 0.53
Regions 2.32 1.58 2.36 2.27 2.64 0.53 2.06 1.72 1.57 2.76 2.39 2.90 2.51 2.62 0.73 2.50 2.13 1.89 1.56 0.25 1.66 1.95 2.92 0.15 1.71 1.11 1.00
Average Growth: Real GDP Per Capita (%) Average Growth: Economically Active Population (%) Average Growth: Dependent Population (%)
Average Growth: Population (%)
Asia East Asia Southeast Asia South Asia Africa Europe South America North America Oceania
3.33 6.11 3.80 1.71 0.97 2.83 0.85 1.61 1.97
Source: Bloom and Williamson (1998), Table 6.
Table 2. Contribution of Demographic Change to Future Economic Growth, 19902025

Estimated Contribution, 19902025 (4 model specications) (1) 0.61 0.40 0.83 1.02 0.98 0.32 0.82 0.21 0.22 0.99 0.14 1.10 1.38 1.63 0.16 1.15 0.645 0.24 (2) (3) 0.50 0.44 0.73 0.90 0.73 0.34 0.71 0.11 0.31 (4) 0.43 0.38 0.62 0.77 0.68 0.29 0.60 0.10 0.26
Regions 1.61 0.20 1.66 2.11 2.78 0.004 1.87 1.33 0.93 0.99 0.87 0.63 0.90 1.88 0.48 0.94 1.21 1.37
Projected Growth: Projected Growth: Projected Growth: Economically Active Dependent Population (%) Population (%) Population (%)
Asia East Asia Southeast Asia South Asia Africa Europe South America North America Oceania
1.36 0.43 1.29 1.65 2.40 0.17 1.50 1.28 1.08
Source: Bloom and Williamson (1998), Table 7.

Figure 2. Economic Growth and the Demographic Transition, East Asia
Growth rate of real GDP per capita
Economic miracle
Other transitional forces Demographic gift
Economic miracle
Youth demographic burden
Sutainable growth
c. 1945
c. 1960
c. 2010
c. 2025 Time
Source: Bloom and Williamson (1998), Figure 6.
1965 and 1990 to a loss of 0.1 to 0.4 percentage points up to 2025, a projected retardation of 1.5 to 2.3 percentage points due solely to demographic forces. South Asia was projected to see a 0.8 to 1.4 percentage point growth rate gain as it left the burden stage of the demographic transition entirely and entered the gift or dividend stage. Southeast Asia was predicted to register a smaller demographic dividend (0.61.1 percentage points): The biggest gainer was projected to be the Philippines while the biggest losers were projected to be Malaysia and Thailand. The macro evidence seemed to support the hypothesis that demographic events helped account for the East Asian economic miracle. Figure 2 offers a stylized version of the demographic dividend hypothesis where the sustainable growth rate is taken to be about 2% per annum. The actual growth rate in GDP per capita rst falls, due to a large and rising child share; then rises, as that large child cohort reaches young adulthood; then reaches a peak, reinforced by a lagged decline in fertility; after which aging and retirement lower the working age share and slow growth rates. The reader should note that the contribution of the demographic transition (labeled the demographic gift in Figure 2) to East Asian economic growth past, present, and future depends on how the miracle is dened. If it is dened as a share of per capita GDP growth between 1960 and 2010 in Figure 2, then it accounts for about a third of the miracle. If it is dened as the surplus over the sustainable rate, then it accounts for almost half. But if it is dened as the increase in growth rates from 19451960 to 19602010, then it accounts for almost three-quarters. These
are very big numbers. Could their size be attributable to the assumptions of the convergence model?
II. Decomposing the Convergence Model: Participation Rate and Productivity Effects
The demographic dividend literature took a big empirical step forward with a paper by Allen Kelley and Robert Schmidt (2005) which decomposed the growth effects uncovered by the na ve demographic dividend model into its transitory labor participation rate effects and its longer term productivity effects. The rst part is pure demographygiven labor productivity growth, any rise in activity or labor participation rate (LPR) will raise per capita income growth. Furthermore, the impact is transitory, although it may last decades. What about the second part? Kelley and Schmidt (2005) listed some possible channels of impact of demography on productivity growthscale economies, density, life-cycle savings and investment responses, and human capital accumulation, two of which will be discussed at length below. But they only used the list to motivate a reduced-form estimation of the two effects. Their main nding was that demography had no long run impact on productivity growth. It turns out that they spoke too soon.
III. How Big the Dividend? Computable OLG and Convergence Models
Analysts were suspicious of the size of the demographic dividend estimated with the reduced-form convergence model (taking accumulation as exogenous) since the underlying savings, capital accumulation, and schooling variables are all endogenous (Feyrer 2007, S anchez-Romero 2012). Thus, a more recent literature has emerged using computable overlapping generations (OLG) models which focus explicitly on the savings and accumulation response, and the demographic impact is typically estimated to be much smaller. For example, that literature nds only a small demographic impact on Japans economic growth in the late 20th century (Braun, Ikeda, and Joines 2009). While the OLG estimated impacts are bigger for Taipei,China (Lee, Mason, and Miller 2000, 2001, and 2003) and the Peoples Republic of China (PRC) (Curtis, Lugauer, and Mark 2011) than for Japan, all of the more recent OLG models seemed to yield smaller demographic dividends than did the original convergence models. But now it appears that they do so mainly by assumption. A recent and impressive paper by Miguel S anchez-Romero (2012) shows that there is actually little difference between the computable OLG and the early convergence model results when applied to Taipei,Chinas late 20th century history. S anchez-Romero nds that demography accounts for 22% of Taipei,Chinas per
capita output growth in 19652005. Why the difference between S anchez-Romero and Lee et al.? Both are using computable OLG models, but one adds an important reality missing from the other. Typically, the computable OLG models get their result by performing counterfactual experiments xing birth and death rates at those prevailing at the start of the period examined. In contrast, and more correctly, S anchez-Romero xes birth and death rates at the levels prevailing a generation before. The former underestimates the demographic impact; the latter reports that for Taipei,China in 19651990, 25% of per capita income growth can be explained by demography, much closer to the 28% estimate of Kelley and Schmidt (2005) for Asia and the 31% estimate of Bloom and Williamson (1998) for East Asia. Thus, it appears that the difference in the estimated growth impact of the demographic transition on Taipei,Chinas performance is much the same whether the convergence model or the OLG model is employed. What remains is to determine whether the result for Taipei,China generalizes to East and Southeast Asia.
IV. Channels of Dividend Impact: Savings and Physical Capital Accumulation A. Savings, Investment, and Accumulation
Since the key to the KelleySchmidt productivity impact of the demographic dividend must lie largely with accumulation responses, the life-cycle savings model and related literature has pursued this connection for Asia ever since Ansley Coale and Edgar Hoover (1958) wrote about dependency burdens a half century ago. An anecdotal fact illustrates the point. In the early 1970s, Korean authorities were concerned by their countrys heavy dependence on Japanese investment nancing and commissioned World Bank papers to explore why the Republic of Korea saved so little (Williamson 1979). By the late 1980s, the Republic of Korea had doubled its savings rate, and its current account balance share of GDP had swung from 8% (net capital inow) to +3.2% (net capital outow). Over the same period, the Korean dependency rate fell by more than 12 percentage points, a huge decline. Was the correlation spurious, or was demography driving some of the accumulation boom and its nancing? The subsequent literature was thus motivated by the following questions: How much of the impressive rise in East Asian savings rates across the late 20th century could be explained by the equally impressive decline in dependency burdens? How much of the fall in external capital dependency in East Asia since the 1970s could be explained by the same demographic forces? Over the past 2 decades, the literature has assigned a large role to the demographic transition in explaining accumulation-driven productivity gains underlying the East Asian miracle. Saving and investment rates can both be driven by demography, each tracing out an inverted-U. The explanation for the saving rate trend would be the famous life-cycle model with high saving rates in the middle of the
demographic transition where mature adults dominate.4 The explanation for the investment rate trend would be the impact of big working adult shares generating large investment shares, and thus accumulation rates, in the middle of the demographic transition. The following predictions logically follow: (i) If the saving rate inverted-U is more dramatic than that of investment, then net capital import shares, or foreign capital dependency, would be big early and small in the middle of the transition. (ii) If, in contrast, the investment inverted-U is more dramatic than that for savings, then net capital import shares would be small early and big in the middle of the transition. Theory cannot discriminate between the two predictions, but empirical work can. The earliest work to pursue the assessment for East Asia (Higgins and Williamson 1997, Higgins 1998, Williamson and Higgins 2001) found the following: rising fertility and falling infant mortality had a profound impact on East Asian saving rates, investment rates, and foreign capital dependency over the half century following 1950. Much of the impressive rise in East Asian saving rates after the late 1960s could be explained by the equally impressive decline in youth dependency burdens. Furthermore, the evidence supported the rst prediction abovecountries that kicked the foreign capital dependency habit rst and fastest were the countries where the youth dependency burden fell rst and most dramatically. Much of the contrasting foreign capital dependency between South and East Asia could be explained by the size of the youth dependency burden and its persistence. All of this implied that demography could explain a large share of the East Asian miracle through accumulation and KelleySchmidt productivity effects. These early studies made many assumptions along the way: that more abundant world savings supplies did not alter capitals incentive to seek new Asian opportunities, although it certainly did; that world capital markets stayed equally open throughout the half century, although they certainly did not; that the life-cycle model was an appropriate explanation for savings behavior; and that the demographic transition was exogenous to accumulation performance. Subsequent work has explored the importance of these assumptions at length, but never in a really persuasive way. Using an OLG model with xed household size and exogenous interest rates, we are told that demography explains almost none of Japans national savings rate
4 Life cycle trends in income and saving have been well documented in the literature, much of it for Asia (Mason, Lee, and Lee 2008, p. 12; Mason and Kinugasa 2008, p. 390). Recent work on the PRC since the 1990s suggests the same (Curtis, Lugauer, and Mark 2011; Wei and Zhang 2011; Banerjee, Meng, and Quian 2010).
(Hayashi and Prescott 2002; Chen, Imrohoro glu, Imrohoro glu 2006 and 2007), but using an OLG model with endogenous interest rates and variable household size, we are told that demography has a positive effect (Braun, Ikeda, and Joines 2009). Using an OLG model with xed interest rates and youth dependency, we are told that demography matters for Taipei,Chinas saving rate experience since 1960 (Lee, Mason, and Miller 2000, 2001, and 2003). More recently, and again using the OLG model with xed interest rates, we are told that most of the PRCs high saving rates are driven by demography (Curtis, Lugauer, and Mark 2011). These assumptions matter.
B. What About World Capital Markets?
Of all these assumptions invoked in the literature surveyed in the previous section, the nancial capital open economy assumption, or what the analysts call the interest rate assumption, is probably the most important, and the most poorly understood. If the interest rate is taken as exogenous, then we are assuming an economy open to world capital markets, where nancial capital is allowed to ow freely across borders, the world interest rate prevails locally, and domestic saving offers no constraint on domestic investment. We are also asked to assume that the world borrowing rate facing East and Southeast Asia was constant over the late 20th century although it certainly was not. Instead, their emerging market borrowing rates converged on the Organisation for Economic Co-operation and Development (OECD) rates (Obstfeld and Taylor 2004; Mauro, Sussman, and Yafeh 2006). If, instead, the interest rate is taken to be endogenous, then we are asked to assume an economy completely closed off from world capital markets, where capital does not ow across borders at all, domestic savings constrains domestic investment, and domestic saving and investment jointly determine the local interest rate. Some of these papers make the open economy assumption, some make the closed economy assumption, but noneas far as I knowexplore the wide reality in between and how it changed over time. Indeed, it is rare that we are told how much the assumption matters. Taipei,China is an exception. Between 1965 and 1990, S anchez-Romero (2012) reports that when the economy is assumed to have been open to capital ows, demography accounts for 25% of per capita output growth, and when it is assumed to have been closed, demography accounts for only 17.2%.5 In open economies, domestic savings is not a constraint on accumulation, and thus booming working adult shares have a bigger impact on investment, accumulation, and growth. Where is the literature that shows us that Asian economies open to world capital markets had bigger demographic effects than those closed? Perhaps
5 However, S anchez-Romero does not explore an open economy assumption where Third World borrowing rates fall, a characterization that comes closer to world capital market integration from the 1970s onwards.
more to the point, Asian capital markets were mostly closed before the 1970s, while they have been mostly open since. Thus, where is the literature that shows us that Asian economies had bigger demographic effects before the 1970s than afterwards? The empirical problem underlying the last research question is, of course, that stages of world capital market integration and stages of demographic transitions have been correlated in Asia over the past half century.
C. A Second Demographic Dividend?
The early convergence and OLG models assumed that older and retired workers dissaved, and that the demographic dividend evaporated as the demographic transition moved into its last stages. Some now think that the dividend persists. Why? The Asian data show that individuals consume much more than they produce well into old age (Mason, Lee, and Lee 2008; Mason and Kinugasa 2008, p. 390). Of course, this could be explained by massive intergenerational transfers within families, but it might also be explained by higher rates of saving for old age by mature working adults. There are at least three plausible reasons why current mature working adults might have higher saving rates than the previous generation thus raising aggregate saving rates late in the demographic transition. First, greater expected longevity would encourage that result. While the na ve demographic transition models stress the initial decline in child mortality, improvements in health environments also raise life expectancy at age 40 or 50, and those effects become increasingly important as the transition evolves. Second, mortality becomes less uncertain as disease is suppressed in poor countries. Lower uncertainty about life expectancy would also encourage more saving by mature working adults. Third, as family size declines, the spatial mobility of children rises, and retirement years increase, parents may be much less certain about the support they can expect from their children in their old age, offering another incentive for mature working adults to save even more. If these forces do indeed raise the saving rates of mature working adults, then they could create a second demographic dividend. All of this is certainly plausible, and it has been conrmed for Asia where greater longevity has raised saving rates across all adult groups (Bloom and Canning 2003). But the PRC, South Asia, and Southeast Asia are not far enough advanced with their demographic transition to offer the required evidence for assessing the magnitudes of a second demographic transition.6 All the Asian evidence comes from
6 Between 2000 and 2005, the population growth rate was 1.39% in ASEAN and 1.62% in India, while it was only 0.14% in Japan, 0.46% in the Republic of Korea, and 0.54% in Taipei,China. In 2000, the youth dependency rates were: 41.8% in ASEAN and 45.1% in India versus 20.5% in Japan; 28.9% in the Republic of Korea; 23.7% in Hong Kong, China; and 29.7% in Taipei,China. In 2000, the population shares 65 and older were: 4.9% in ASEAN and 4.6% in India versus 17.2% in Japan; 7.4% in the Republic of Korea; 11% in Hong Kong, China; and 8.1% in Taipei,China (Mason, Lee, and Lee 2010, Tables 1.1, 1.2, and 1.4).
the Republic of Korea; Japan; and Taipei,China. Ronald Lee and Andrew Mason nd very large second dividends for Taipei,China (Mason and Lee 2007; Mason, Lee, and Lee 2010) which they argue is consistent with the micro studies of Taipei,China (Deaton and Paxson 2000). For a sample of 25 European countries plus Australia, Japan, the Republic of Korea, Malaysia, New Zealand, Thailand, and the United States (US), the estimates of second dividends are large. Indeed, the two (saving rate) dividends are estimated to have been about equal (Mason and Kinugasa 2008, p. 398): declining child dependency led to a rise in saving rates by 6.9 percentage points (while) improvements in adult survival led to a rise in saving rates by 6.7 percentage points.7 Whether these second dividend magnitudes will hold up for the PRC, South Asia, and Southeast Asia cannot yet be asserted, but that future looks likely.
V. Channels of Dividend Impact: Human Capital Accumulation and Vintage Effects
We all agree that human capital accumulation is an important driver of growth, so how might it be connected to the demographic transition? There are two possibilities. First, and following Gary Becker (1960 and 1981) and H. Gregg Lewis (Becker and Lewis 1973), there may be a qualityquantity trade-off working at the family leveli.e., more children and lower investment per child versus fewer children and higher investment per child. This would clearly be a force endogenous to the demographic transition: A low-quality child cohort (produced under conditions of high fertility and big families) implies weak vintage effects when that cohort enters the labor force a decade or two later, while a high-quality child cohort (produced under conditions of low fertility and small families) implies a big vintage effect when the cohort enters the labor force a decade or two later. If those cohorts are themselves big, the vintage effect is even bigger. Second, there is the possibility of an independent co-movement between public spending on these investments and the size of the youth cohorts. If a publicschoolinghealth revolution coincides with the demographic transition, and if that revolution impacts on the quality of the new children, we have what might be called a quasi-endogenous effect. Without the public-schoolinghealth revolution, the vintage effects of the demographic transition would be weak and limited to the demand-side Becker effects, but with it, they would be strong and pushed by the supply side. In any case, both of these forces imply vintage effectsaverage human capital per worker rises as the well-schooled young workers replace the poorly-schooled older workers.
7
For an elegant and impressive extension of their views, see Lee and Mason (2011).
A.
Endogenous Schooling from the Demand Side: Quality versus Quantity
Some of the best work documenting the endogenous qualityquantity human capital effects coming from fertility declines in the second stage of the demographic transition, at least for Asia, have been produced by Andrew Mason, Ronald Lee, and their collaborators. The cross-section correlation between total public and private human capital investment per child and the total fertility rate is striking (Mason, Lee, and Lee 2010, Figure 1.8). In their research, Japan, the Republic of Korea, and Taipei,China are later in the demographic transition with high investment per child and low fertility rates. India, Indonesia, and the Philippines, meanwhile, are earlier in the demographic transition with low investment per child and higher fertility rates. Thailand is in the middle. Lee and Mason (2009) have shown how these qualityquantity correlations could have a powerful impact on human capital accumulation and growth. Although independent East Asian evidence certainly seems to support the Becker demand-side connection (Montgomery, Arends-Kuenning, and Mete 2000; Jun 2013), Lee and Mason never identify the source of the correlation they observe across Asia. Was it qualityquantity family effects that would be so clearly associated with demographic transitions, or was it something else? In addition, their measures of investment per child do not control for schooling qualitywhen parents demand more schooling for their children, it may generate crowding and rising teacher (and other schooling) costs and thus more spending, but maybe not more quality-adjusted schooling. This is exactly what T. Paul Schultz (1987) found for a large sample of countries for the period 19601981, and what I found for a smaller sample (but which included Hong Kong, China; Japan; the Republic of Korea; Malaysia; the Philippines; and Thailand). Controlling for other relevant variables, we found the relative price of teachers had a strong negative impact on schooling (Williamson 1993, p. 154). Why should we care? Because positive public supply-side forces should lower the relative schooling price while demand side forces should raise it. The behavior of the relative cost of schooling (and health) might help untangle the demand and supply side forces. Where is the literature that explores this important issue? In short, while other studies have conrmed the qualityquantity trade-off in micro data, Lee and Mason (2009) are dealing with macro data (and magnitudes across countries and over time) and we are not yet sure how much of the correlation they observe is driven by private family schooling demands, and how much by public schooling supplies. The differences will inuence interpretation and thus policy.
B. Supply Side: Demographic Transitions and Public Schooling Revolutions
Taipei,China was a schooling leader in Asia, and its experience illustrates the supply-side point. In it was established 6 years of compulsory education, and this
period was extended to 9 years in 1968. As a result, the proportion of illiterate . . . changed from 40% in 1940 to almost [zero] in 1970 (S anchez-Romero 2012, footnote 2; citing Huang 2001). Obviously, when Taipei,China passed from the youth dependency phase of the demographic transition to the young adult worker and the mature adult worker phases during 19502000 (Mason, Lee, and Lee 2010, Tables 1.2 and 1.3), the co-movement of the early public schooling revolution later raised average worker schooling due to these powerful vintage effects. The Asian public schooling revolution examples can be easily multiplied (Williamson 1993, Duo 2011).8 Indeed, the evidence shows that since 1950 East Asia has made a bigger commitment to schoolingand with a steeper rise in that commitmentthan anywhere else in the Third World. Why is that so, and why is it correlated with a more dramatic demographic transition? Is it public supply-side political economy at work, or is it private demand-side qualityquantity trade-offs? The recent neo-institutional contributions of Daron Acemoglu and James Robinson (e.g., 2006) and Stanley Engerman and Kenneth Sokoloff (e.g., 2012) would suggest low inequality and broad political participation might explain it. But we need far more research on this question.
VI. Emigration: Muting the Dividends Impact by Brain Drain A. Emigration Life Cycles
Countries typically pass through a migration transition, or what might be called an emigration life cycle, driven by income differences between the low-wage sending country and high-wage host countries, modernization, and economic growth at home, and of course, the sending countrys experience with its demographic transition. These emigration life cycles have been used by economic historians to describe European emigration from the 1840s to World War I, where each countrys emigration rate rose steeply from very low levels, after which the rise began to slow down as the emigration rates reached a peak and subsequently fell again to low levels. These historical emigration life cycles can be seen in aggregate country emigration rates, regional emigration rates within countries, and ruralurban emigration rates within countries (Hatton and Williamson 2005b, Chapter 4). Since these emigration life cycles are so pervasive in the historical data, it is hardly surprising that we also have seen them in the Third World since the 1950s (Hatton and Williamson 2005a and 2011, Williamson [Forthcoming]). While similar to 19th century European emigration patterns, the more recent Third World
8 The best paper covering the 19th and early 20th century (primary) schooling revolution in the now-advanced countries is by Richard Easterlin (1981). The reference to the post-World War II Asian schooling revolution is mainly secondary education.
country emigration life cycles cover shorter time periods. One oft-cited example is the Republic of Korea where emigration rose steeply to a peak in 1982 and subsequently declined just as quickly. Similar patterns have been observed for other Asian countries and it is tempting to associate these more compressed emigration life cycles with economic miracles and accelerated demographic transitions. While emigration life cycles have been examined for individual countries, they have also been explored at more aggregative continental levels, although the expectation is that the aggregate life cycles should be less dramatic. After all, some countries start and complete their emigration life cycles early, while others start and nish later, tending to partially smooth out the aggregate regional experience. Still, Asia reached peak emigration rates in 19801984, even though some economies reached peaks early, like the Republic of Korea and Taipei,China, while others much later, like Indonesia and the Philippines (Hatton and Williamson 2011, Table 1).9 This brief survey raises three questions. First, are emigration life cycles country-specic special cases or do they reect some common laws of motion? Second, if they do reect some common laws of motion, what are the shared economic and demographic fundamentals driving them? In particular, while we certainly see the correlationemigration life cycles following a decade or two after the rst stage of the demographic transitionexactly what role does the demographic transition play? Third, are the demographic forces powerful enough to account for at least some of the observed brain drain, and if so, are the drains big enough to mute some of the demographic dividends? Modern economic analysis of international migration uses the framework rst set out by Larry Sjaastad (1962) and rened by George Borjas (1987 and 1994), Barry Chiswick (2000), and others. Thus, the emigration decision is characterized as depending on the economic gain from migration net of its costs, the latter including waiting time (inuenced by short-run labor market conditions in host countries) and queues related to host country admission criteria (and illegal migration costs). Since young adults have the most to gain from long distance moves, they record by far the highest emigration rates. Thus, when young adult cohorts are big during the second stage of the demographic transition, aggregate emigration rates should be big as well, other things being equal. Since capital markets facing most poor households are imperfect or even absent, the demand to emigrate is constrained by poverty. But those constraints are released over time as growth miracles at home create more good jobs and thus the incomes necessary to nance more emigrants, and as remittances and in-kind help from previous (and increasing) emigrants living abroad rises.10
9 Asia is dened here very broadly to include East Asia, Southeast Asia, South Asia, the Middle East, and North Africa. 10 Uninformed observers often think that successful development at home will keep young adults from emigrating. On the contrary, the big income gaps between sending and host countries are still big, and now families have more resources to nance the next emigrant.
As growth miracles unfold, they generate two competing forces on the upswing of the emigration life cycle: poverty rates fall making it easier for families to nance the moves of their children and helping release the poverty constraint on emigration; but miracle growth implies catching up with the leaders, thus reducing the gains from the move. The rst dominates early in the emigration cycle, while the second dominates later in the cycle. Schooling matters as well. As the educational attainment of young cohorts rise, their ability to exploit labor market opportunities abroad rises as well. Emigration may also be constrained by host country immigration policy and perturbed by civil strife at home, but these events are more random. So goes the theory, but the econometric facts support the theory.
B. Emigration Life Cycle Fundamentals
Recent research has identied the main drivers of Third World emigration rates after the 1960s, the start of the great boom in world migration (Martin and Taylor 1996). The best evidence is for sending-country emigration to the US (Hatton and Williamson 2011). While changing immigration policy, civil strife at home, and other exogenous events have mattered, it is the underlying demographic and economic fundamentals that explain the common emigration life-cycle experience across sending countries. First, the US migrant stock effect made the most important contribution to the boom up to the 1990s, reecting both the importance of family reunication in US immigration policy and the previous impact of economic and demographic fundamentals on migration ows which then got embedded in the current migrant stock, thus raising current ows. Indeed, were it not for the migrant stock effect, Asian emigration rates would have fallen steeply after 19901994 rather than only slightly dropping. Second, the birth cohort effect played an important role in the downturn after 19901994 in Asia.11 Third, education catch-up also played an important role everywhere in the Third World, augmenting emigration rates, but it was especially powerful in Asia, where the schooling revolution was most dramatic. Fourth, while there was certainly per capita income growth catch-up in Asia, the growth miracles were not fast enough to reduce signicantly the income gap with the US, thus they contributed little to the emigration boom. Finally, while statistically signicant, the diminished poverty trap did not contribute as much to either the emigration boom in 19901994, or the decline thereafter.
C. Brain Drain and Wasted Demographic Dividends?
Perhaps it is obvious that emigration of well-schooled young adults diminishes the demographic dividend. True, some have shown that emigration of well-schooled
11
The same was true of Latin America but not Africa where the demographic transition is lagging behind.
young adults raises the expected rate of return to schooling for the next younger cohort left behind.12 Since this raises schooling rates, it offers a partial offset to the drain (Cervantes and Guellec 2002, Williamson 2007, Docquier and Rapoport 2012). Others have argued that remittances offset the brain drain losses (Fajnzylber and Lopez 2007, Yang 2008 and 2011, Guiliano and Ruiz-Arranz 2009), but the assessment ignores potential damage created by Dutch disease on manufacturing jobs or households use of those remittances to nance yet another childs emigration. Any assessment of the brain drain and demographic transition connection is made even more complicated by the powerful role of emigrant migration stocks abroad pulling even more young well-schooled adults abroad long after demography had its rst-order effect on emigration. As far as I know, the connection has not yet been assessed empirically.
VII. Beneath the Macro: RuralUrban Migration and Income Gaps
Oddly enough, it is hard to nd a single paper in the literature that breaks the demographic transition down into rural and urban component parts. We know that the richer, better educated, more progressive, and female-job friendly cities lead the poorer, less educated, more conservative, and less female-job friendly rural areas in the demographic transition. Postwar child mortality in Asia fell rst in the cities, and with a lag, fertility rates fell there rst as well. Where the rural lag (behind urban) has been big, we should see three things (ceteris paribus): (i) big migrations to the cities pushed by a young adult glut in the countryside; (ii) rising rural inequality driven by the same glut; and (iii) rising wage gaps also driven by the same glut. If the analyst just compares the size of youth cohort age shares in urban and rural areas, she will downplay these forces since ruralurban migration tends, at least partly, to equilibrate. Therefore, one has to look at the ruralurban fertility and child mortality differentials a couple of decades earlier to properly assess impact. I have said that these ruralurban differences in timing and magnitudes should have countrywide inequality implications (within regions, between regions, and countrywide). These will be explored in the next section, but here I will simply repeat the prediction that a glut of young adults who stay in rural areas will lower wages and incomes and raise ruralurban wage and income gaps. Anyone interested in ruralurban migrations in Asia and their distributional implications should pay attention to ruralurban differences with their demographic transition experiencesespecially for the PRC, Indonesia, the Philippines, and Thailand. As far as I know, nobody has yet done so.
12 That is, expected rates of return to schooling are raised by the possibility of emigration to high-wage countries (conrmed by the experience of older emigrating siblings).
VIII. The Demographic Transition and Inequality Connection
The cohort size hypothesis is simple enough: fat age cohorts tend to get lower rewards due to a supply glut and thin age cohorts get higher rewards due to scarcity. When the fat cohorts lie in the middle of the ageearnings curve where life-cycle income is highest, this labor market glut lowers income in the middle, thus tending to atten the ageearnings curve, and earnings inequality is moderated. When instead the fat cohorts are either young or old, these kinds of labor market gluts lower incomes at the two tails of the ageearnings curve, thus tending to augment earnings inequality. This demographic hypothesis has a long tradition in the US starting with the entry of the baby boomers into the labor market when their big numbers created poorer job prospects (Easterlin 1980), and the impact was surveyed not too long ago (Lam 1997, pp. 102324 and pp. 104452; Macunovich 1998). All such studies have shown that relative cohort size has had an adverse supply-side effect on the relative wages of the fat cohort in the US since the 1950s. What about the world more generally, and Asia in particular? If the cohort size hypothesis helps explain US (and European) postwar experience with earnings inequality, it might do even better elsewhere. After all, there is far greater variance in the age distribution of populations between countries than there has been over time in the US. More to the point, the post-World War II (WWII) demographic transition in Asia and the rest of the Third World has generated much more dramatic changes in relative cohort size than did the baby boom in the US and the OECD. In addition, the postwar OECD countries were already urbanized, so the postulated ruralurban inequality effects discussed in the previous section would have been far weaker there than should have been true of Asia and the rest of the Third World. As far as I know, nobody has yet explored this last hypothesis in the literature. One study (Williamson 2001, Higgins and Williamson 2002) used a world panel database of 92 countries to explain the DeiningerSquire (1996) income inequality Gini coefcients from the 1960s to the 1990s. The explanatory variables were: older labor force cohort size dened as the proportion of the adult population (taken to be persons 1569) who are ages 4059; a measure of openness; and GDP per capita entered nonlinearly to capture Kuznets Curve effects. The results were statistically signicant and robust. More importantly, the analysis assigned the biggest inuence to demographic-transition-induced cohort size effects. Indeed, compared with the East and Southeast Asian economies, inequality in Africa and Latin America in the 1990s was much higher, bigger by 7.2 points in Africa and by 10.8 points in Latin America.13 If Africa had the same demographic mix as East and Southeast Asia, inequality would have been lower by 3.6 points:
13 Asian economies here include the PRC; Hong Kong, China; Indonesia; Japan; the Republic of Korea; Malaysia; the Philippines; Taipei,China; and Thailand.
cohort size accounting for about half of the difference between the two regions. If Latin America had the same demographic mix as East and Southeast Asia, inequality would have been lower by 3.1 points: cohort size accounting for almost a third of the difference between the two regions. The study also showed that cohort size effects (rising mature labor force shares) should serve to lower inequality by more than 8 percentage points between the early 1990s and 2025, suggesting that these demographic changes could be a powerful force promoting reduced inequalitythe Gini coefcient for East and Southeast Asia was projected to fall from a relatively low 39.2 to a still lower 31.5 by 2025, after which it should stabilize. Harry Oshima and Andrew Mason also explored some of these issues for Asia, but their most novel contribution, at least in my opinion, was their stress on inequality of years of life (Oshima and Mason 2001, pp. 4045). Some time ago, Simon Kuznets (1976) stressed that it was inequality of lifetime earnings that we should be measuring. Since child and adult mortality are powerfully inuenced by poverty in very poor countries, low incomes and low life expectancy are correlated. Thus, lifetime incomes must have been much more unequal than annual incomes in most of Asia during the 1950s and 1960s. But if a fall in child mortality early in the demographic transition favors the poor (perhaps because it is driven mainly by public intervention), then it should offer another source of more egalitarian lifetime incomes. If a fall in adult mortality later in the demographic transition also favors the poor, it should be a source of more egalitarian lifetime incomes. Oshima and Mason (2001) report a very steep decline in the inequality of years lived by East Asians during the post-WWII era (something we also see across countries in the 20th century), contributing to much more egalitarian lifetime incomes. These trends are likely to continue in the near future, and they need more of our attention.
IX. Agenda
It seems to me that there are many unanswered questions involving the impact of the demographic transition on country economic performance. Estimates of the rst and second demographic dividends seem only to scratch the surface. We need to learn much more about the schooling, emigration, and inequality connections as well as ruralurban dynamics. The demographic transition matters!
References
Asian Development Bank. 1997. Emerging Asia: Change and Challenges. Manila: Asian Development Bank. Acemoglu, Daron, and James Robinson. 2006. De Facto Political Power and Institutional Persistence. American Economic Review 96(2): 32530. Banerjee, Abhijit, Xin Meng, and Nancy Quian. 2010. The Life Cycle Model and Household Savings: Micro Evidence from Urban China. Paper presented at the National Bureau of

Economic Research Working Group on the Chinese Economy, 12 October, Cambridge, Massachusetts. Barlow, Robin. 1994. Population Growth and Economic Growth: Some More Correlations. Population and Development Review 20(1): 15365. Barro, Robert. 1991. Economic Growth in a Cross Section of Countries. Quarterly Journal of Economics 106(2): 40743. . 1997. Determinants of Economic Growth: A Cross-country Empirical Study Cambridge, Mass.: MIT Press. Becker, Gary. 1960. An Economic Analysis of Fertility. In Gary S. Becker, ed. Demographic and Economic Change in Developed Countries. Princeton: Princeton University Press. . 1981. A Treatise on the Family. Cambridge: Harvard University Press. Becker, Gary, and H. Gregg Lewis. 1973. On the Interaction between the Quantity and Quality of Children. Journal of Political Economy 81(2): 106393. Bloom, David, and David Canning. 2003. Longevity and Life-Cycle Savings. Scandinavian Journal of Economics 105(3): 31938. . 2008. Global Demographic Change: Dimensions and Economic Signicance. Population and Development Review 34 Supp (8): 1751. Bloom, David, David Canning, and Pia Malaney. 2000. Population Dynamics and Economic Growth in Asia. Population and Development Review 26 Supp: 25790. Bloom, David, and Richard Freeman. 1986. The Effects of Rapid Population Growth on Labor Supply and Employment in Developing Countries. Population and Development Review 12(3): 381414. Bloom, David, and Jeffrey Williamson. 1997. Demographic Transitions, Human Resource Development, and Economic Miracles in Emerging Asia. In Emerging Asia: Change and Challenges. Manila: Asian Development Bank. . 1998. Demographic Transitions and Economic Miracles in Emerging Asia. World Bank Economic Review 12(3): 41955. Borjas, George. 1987. Self-Selection and the Earnings of Immigrants. American Economic Review 77(4): 53153. . 1994. The Economics of Immigration. Journal of Economic Literature 32(4): 1667717. Boserup, Ester. 1981. Population and Technological Change: A Study of Long-term Trends. Chicago: University of Chicago Press. Brander, James, and Steve Dowrick. 1994. The Role of Fertility and Population in Economic Growth: Empirical Results from Aggregate Cross-National Data. Journal of Population Economics 7(1): 125. Braun, R. Anton, Daisuke Ikeda, and Douglas Joines. 2009. The Saving Rate in Japan: Why it has Fallen and Why it will Remain Low. International Economic Review 50(1): 291 321. Cervantes, Mario, and Dominique Guellec. 2002. The Brain Drain: Old Myths, New Realities. OECD Observer No. 230. Paris: Organisation for Economic Co-operation and Development. Chen, Kaiji, Ayse Imrohoro glu, and Selahattin Imrohoro glu. 2006. The Japanese Saving Rate. American Economic Review 96(5): 185058. . 2007. The Japanese Saving Rate between 1960 and 2000: Productivity, Policy Changes, and Demographics. Economic Theory 32(1): 87104. Chiswick, Barry. 2000. Are Immigrants Favorably Self-Selected? An Economic Analysis. In Caroline Brettell and James Hollield, eds. Migration Theory: Talking Across Disciplines. New York: Routledge.
22 ASIAN DEVELOPMENT REVIEW Coale, Ansely. 1986. Population Trends and Economic Development. In Jane Mencken, ed. World Population and U S Policy: The Choices Ahead. New York: Norton. Coale, Ansely, and Edgar Hoover. 1958. Population Growth and Economic Development in LowIncome Countries. Princeton: Princeton University Press. Curtis, Chadwick, Steven Lugauer, and Nelson Mark. 2011. Demographic Patterns and Household Saving in China. NBER Working Paper No. 16828. Cambridge, MA: National Bureau of Economic Research. Deaton, Angus, and Christina Paxson. 2000. In Ronald Lee and Cyrus Chu, eds. Population and Economic Change in East Asia: A Supplement to Population and Development Review. New York: Population Council. Deininger, Klaus, and Lyn Squire. 1996. A New Data Set Measuring Income Inequality. World Bank Economic Review 10(3): 56591. Docquier, Frederic, and Hillel Rapoport. 2012. Globalization, Brain Drain, and Development. Journal of Economic Literature 50(3): 681730. Duo, Esther. 2011. Schooling and Labor Market Consequences of Schooling Construction in Indonesia. American Economic Review 91(4): 795813. Easterlin, Richard. 1980. Birth and Fortune: The Impact of Numbers on Personal Welfare. New York: Basic Books. . 1981. Why Isnt the Whole World Developed? Journal of Economic History 41(1): 119. Ehrlich, Paul. 1968. The Population Bomb. New York: Ballantine. Engerman, Stanley, and Kenneth Sokoloff. 2012. Economic Development in the Americas Since 1500: Endowments and Institutions. Cambridge: Cambridge University Press. Fajnzylber, Pablo, and Humberto Lopez, eds. 2007. Remittances and Development: Lessons from Latin America. Washington, DC: World Bank. Feeney, Grifth, and Andrew Mason. 2001. Population in East Asia. In Andrew Mason, ed. Population Change and Economic Development in East Asia: Challenges Met, Opportunities Seized. Stanford: Stanford University Press. Feyrer, James. 2007. Demographics and Productivity. Review of Economics and Statistics 89(1): 100009. Guiliano, Paola, and Marta Ruiz-Arranz. 2009. Remittances, Financial Development, and Growth. Journal of Development Economics 90(1): 14452. Hatton, Timothy, and Jeffrey Williamson. 2005a. What Fundamentals Drive World Migration? In George Borjas and Jeff Crips, eds. Poverty, International Migration and Asylum. Hampshire, UK: Palgrave-Macmillan for WIDER. . 2005b. Global Migration and the World Economy: Two Centuries of Policy and Performance. Cambridge, MA: MIT Press. . 2011. Are Third World Emigration Forces Abating? World Development 39(1): 2032. Hayashi, Fumio, and Edward Prescott. 2002. The 1990s in Japan: A Lost Decade. Review of Economic Dynamics 5(1): 20635. Higgins, Matthew. 1998. The Demographic Determinants of Savings, Investment and International Capital Flows. International Economic Review 39(2): 34369. Higgins, Matthew and Jeffrey Williamson. 1997. Age Structure Dynamics in Asia and Dependence on Foreign Capital. Population and Development Review 23(2): 26193. . 2002. Explaining Inequality the World Round: Cohort Size, Kuznets Curves, and Openness. Southeast Asian Studies 40(3): 268302. Huang, Fung-Mey. 2001. In Andrew Mason, ed. Population Change and Economic Development in East Asia: Challenges Met, Opportunities Seized. Stanford: Stanford University Press.

Jun, Bogang. 2013. The Trade-off between Fertility and Education: Evidence from the Korean Development Path. MPRA Working Paper No. 43971. Munich: Munich Personal RePEc Archive. Kelley, Allen. 1988. Economic Consequences of Population Change in the Third World. Journal of Economic Literature 27:1685728. Kelley, Allen, and Robert Schmidt. 1995. Aggregate Population and Economic Growth Correlations: The Role of the Components of Demographic Change. Demography 32(4): 543 55. . 2005. Evolution of Recent EconomicDemographic Modeling: A Synthesis. Journal of Population Economics 18(2): 275300. Kim, Jong-Il, and Lawrence Lau. 1994. The Sources of Economic Growth of the East Asian Newly Industrialized Countries. Journal of the Japanese and International Economies 8(3): 23571. Krugman, Paul. 1994. The Myth of Asias Miracle. Foreign Affairs 73(6): 6278. Kuznets, Simon. 1967. Population and Economic Growth. Proceedings of the American Philosophical Society 111(3): 17093. . 1976. Demographic Aspects of the Size Distribution of Income: An Exploratory Essay. Economic Development and Cultural Change 25(1): 194. Lam, David. 1997. Demographic Variables and Income Inequality. In Mark Rosenzweig and Oded Stark, eds. Handbook of Population and Family Economics. New York: Elsevier Science. Lee, Ronald. 2003. The Demographic Transition: Three Centuries of Fundamental Change. Journal of Economic Perspectives 17(4): 16790. Lee, Ronald, and Andrew Mason. 2009. Fertility, Human Capital, and Economic Growth Over the Demographic Transition. European Journal of Population 26(2): 15982. . 2011. Generational Economics in a Changing World. Population and Development Review 37 Supp (1): 11542. Lee, Ronald, Andrew Mason, and Timothy Miller. 2000. Life-cycle Saving and the Demographic Transition in Eastern and Southeastern Asia. Population and Development Review 26 Supp:194219. . 2001. Saving, Wealth, and the Demographic Transition in East Asia. In Andrew Mason, ed., Population Change and Economic Development in East Asia: Challenges Met, Opportunities Seized. Stanford: Stanford University Press. . 2003. Saving, Wealth, and Population. In Nancy Birdsall, Allen Kelley, and Steven Sinding, eds. Population Matters: Demographic Change, Economic Growth, and Poverty in the Developing World. Oxford: Oxford University Press. Macunovich, Diane. 1998. Relative Cohort Size and Inequality in the United States. American Economic Review 88(2): 259264. Martin, Philip, and J. Edward Taylor. 1996. The Anatomy of a Migration Hump. In J. Edward Taylor, ed. Development Strategy, Employment and Migration: Insights from Models. Paris: Organisation for Economic Co-operation and Development. Mason, Andrew. 2007a. Demographic Dividends: The Past, the Present, and the Future. In Andrew Mason and Mitoshi Yamaguchi, eds. Population Change, Labor Markets, and Sustainable Growth: Towards a New Economic Paradigm. Oxford: Elsevier. . 2007b. Demographic Transition and Demographic Dividends in Developed and Developing Nations. In Proceedings of the UN Expert Group Meeting on Social and Economic Implications of Changing Population Age Structures. New York: United Nations.
24 ASIAN DEVELOPMENT REVIEW Mason, Andrew, and Ronald Lee. 2007. Transfers, Capital, and Consumption over the Demographic Transition. In Robert Clark, Naohiro Ogawa, and Andrew Mason, eds. Population Aging, Intergenerational Transfers, and the Macroeconomy. Elgar Press. Mason, Andrew, Ronald Lee, and Sang-Hyop Lee. 2010. The Demographic Transition and Economic Growth in the Pacic Rim. In Takatoshi Ito and Andrew Rose, eds., The Economic Consequences of Demographic Change in East Asia. Chicago: University of Chicago Press. Mason, Andrew, Sang-Hyop Lee, and Ronald Lee. 2008. Will Demographic Change Undermine Asias Growth Prospects? Background paper for the Asian Development Banks study Emerging Asian Regionalism: Ten Years after the Crisis. Mason, Andrew, and Tomoko Kinugasa. 2008. East Asian Economic Development: Two Demographic Dividends. Journal of Asian Economics 19(5/6): 38999. Mauro, Paolo, Nathan Sussman, and Yishay Yafeh. 2006. Emerging Markets and Financial Globalization: Sovereign Bond Spreads in 18701913 and Today. Oxford: Oxford University Press. Montgomery, Mark, Mary Arends-Kuenning, and Cem Mete. 2000. The Quantity-Quality Transition in Asia. Population and Development Review 26(2): 22356. Obstfeld, Maurice, and Alan Taylor. 2004. Global Capital Markets: Integration, Crisis, and Growth. Cambridge: Cambridge University Press. Oshima, Harry, and Andrew Mason. 2001. Population and Inequality in East Asia. In Andrew Mason, ed. Population Change and Economic Development in East Asia: Challenges Met, Opportunities Seized. Stanford: Stanford University Press. S anchez-Romero, Miguel. 2012. The Role of Demography on Per Capita Output Growth and Saving Rates. Journal of Population Economics (online). Schultz, T. Paul. 1987. School Expenditures and Enrollments 19601980: The Effects of Income, Prices, and Population. In David Johnson and Ronald Lee, eds. Population Growth and Economic Development. Madison: University of Wisconsin Press. Simon, Julian. 1981. The Ultimate Resource. Princeton: Princeton University Press. Sjaastad, Larry. 1962. The Costs and Returns of Human Migration. Journal of Political Economy 70(5): 8093. United Nations. 1991. Global Estimates and Projections of Populations by Age and Sex. New York: United Nations. Wei, Shang-Jin, and Xiaobo Zhang. 2011. The Competitive Savings Motive: Evidence from Rising Sex Ratios and Saving Rates in China. Journal of Political Economy 119(3): 51164. Williamson, Jeffrey. 1979. Why Do Koreans Save So Little? Journal of Development Economics 6(3): 34362. . 1993. Human Capital Deepening, Inequality, and Demographic Events Along the Asia Pacic Rim. In Naohiro Ogawa, Gavin Jones, and Jeffrey Williamson, eds. Human Resources and Development Along the Asia-Pacic Rim. Oxford: Oxford University Press. . 2001. Demographic Change, Economic Growth and Inequality. In Nancy Birdsall, Allen Kelley, and Steven Sinding, eds. Population Matters: Demographic Change, Economic Growth, and Poverty in the Developing World. Oxford: Oxford University Press. . 2007. Inequality and Schooling Responses to Globalization Forces: Lessons from History. In Migration, Trade and Development. Dallas: Dallas Federal Reserve Bank. . Forthcoming. World Migration in Historical Perspective: Four Big Issues. In Barry Chiswick and Paul Miller, eds. Handbook for Economics of International Migration. Elsevier Press.

Williamson, Jeffrey, and Matthew Higgins. 2001. The Accumulation and Demography Connection in East Asia. In Andrew Mason, ed. Population Change and Economic Development in East Asia: Challenges Met, Opportunities Seized. Stanford: Stanford University Press. Yang, Dean. 2008. International Migration, Remittances, and Household Investment: Evidence from Philippine Migrants Exchange Rate Shocks. Economic Journal 118(528): 591630. . 2011. Migrant Remittances. Journal of Economic Perspectives 25(3): 82242. Young, Alwyn. 1994a. Lessons from the East Asian NICS: A Contrarian View. European Economic Review 38(3): 96473. . 1994b. The Tyranny of Numbers: Confronting the Statistical Realities of Asian Growth Experience. NBER Working Paper No. 4680. Cambridge, MA: National Bureau of Economic Research.
Princelings and Paupers? State Employment and the Distribution of Human Capital Investments Among Households in Viet Nam
IAN COXHEAD AND DIEP PHAN
Inequality in access to education is known to be a key driver of income inequality in developing countries. Viet Nam, a transitional economy, exhibits signicant segmentation in the market for skilled labor based on more remunerative employment in government and state rms. We ask whether this segmentation is also reected in human capital investments at the household level. We nd that households whose heads hold state jobs keep their children in school longer, spend more on education, and are more likely to enroll their children in tertiary institutions relative to households whose heads hold nonstate jobs. The estimates are robust to a wide range of household and individual controls. Over time, disparities in educational investments based on differential access to jobs that reward skills and/or credentials help widen existing income and earnings gaps between well-connected princelings and the rest of the labor market. Capital market policies that create segmentation in the market for skills also crowd out investment in private sector rms, further reducing incentives for human capital deepening. Keywords: human capital, state-owned, education, connections, inequality, Viet Nam JEL codes: J24, J45, O15, P23
I. Policy Distortions, Connections, and Inequality in Transition Economies
Inequality in Viet Nam is low by the standards of Asian economies but has risen during that countrys transition to market-led socialism. Part of the rise can be explained by reference to the weakening of many socialist-era policies that repressed returns to skills, ability, and entrepreneurial activity, in which case greater income disparities might be regarded as necessary and even benecial consequences of liberalization. But despite huge strides toward a market economy, the state retains signicant power over some product markets, notably those in which there is potential
Ian Coxhead is Professor, Department of Agricultural and Applied Economics at the University of WisconsinMadison. Diep Phan is Assistant Professor of Economics at Beloit College. We thank Bernard Thiam Hee Ng, seminar participants at the Asian Development Bank and the University of Wisconsin-Madison, and two anonymous referees for helpful comments on earlier drafts. Remaining errors are ours alone.
PRINCELINGS AND PAUPERS? 27
for monopolistic behavior, and factor markets, especially those for capital and land. It follows that part of the observed rise in inequality may be due to less positive trends, including rent capture by those with access to high-level government positions or to employment in rms that are in some way protected by policy or market structure. As in the Peoples Republic of China (PRC), another economy undergoing a broadly similar transition, there is widespread concern in Viet Nam over the alleged capture of state-owned or partially privatized rms and inuential public service positions by princelingsthat is, members of the former nomenklatura and their close relatives and associates.1 The same concerns extend to many other developing economies in which privileges extended to well-connected owners of capital and land have led to diminished earnings and opportunities for owners of labor (Berg and Ostry 2011). Paradoxically, in some economies, the structure of product and labor markets (and especially that for skilled labor) is such that the rewards to skills are signicantly higher in public sector employment than in the private sector, leading to queuing and competition for such jobs.2 In Viet Nam, income inequality across the state/nonstate divide is clearly visible in household survey data. The data also reveal that this inequality is caused by just a few households at the very top of the political hierarchy. Figure 1, which is based on representative national household data, shows that the per capita incomes of state households (those containing at least one member working for a state rm or in public administration) are appreciably higher than those for nonstate households, whether in urban or rural areas.3 The gap widened between 2004 and 2008, a period of very rapid growth in Viet Nam during which average real income per capita rose by almost 50%. Why should the labor market connections of households be important, and why should returns to skills be higher for state employees than those in the private sector? A default model would predict equal returns to labor of equal skills in equilibrium. Moreover, the experience of the transition from socialism in Eastern Europe and the former Soviet Union was overwhelmingly one in which movement of white-collar workers to the private sector was strongly positively selected, reecting more productive employment opportunities with higher earnings to match (Adamchik and Bedi 2000, Munich, Svejnar, and Terrell 2005). Viet Nam, like the PRC, shows the opposite trend (Phan and Coxhead 2013). Our answer, while speculative, ts with the stylized facts of an economy in an as yet incomplete transition from command to market economy. The skill intensity of civil service employment is high, but the demand for workers is limited by budget
1 Princelings in [the Peoples Republic of] China Use Family Ties to Gain Riches, New York Times, 17 May 2012. See also, In Viet Nam, Message of Equality is Challenged by Widening Wealth Gap, New York Times, 1 September 2012. 2 Young Feel Hungrier for Golden Rice Bowl Jobs: Record Numbers Eye Public Sector, Financial Times, 25 October 2012. 3 Simple t-tests also show statistically signicant difference in the mean per capita income between the two groups of households, for both rural and urban areas (at signicance levels 7% or lower).
28 ASIAN DEVELOPMENT REVIEW Figure 1. Distribution of Per Capita Income for State and Nonstate Households (D million, January 1998 prices)
Kernel Density Estimate .00006
.00004 Density state urban .00002 nonstate urban 0 0 20000 40000 pcinc kernel = epanechnikov, bandwidth = 2089.57 Kernel Density Estimate .0001 .00008 Density .00006 .00004 .00002 0 0 20000 40000 60000 pcinc 80000 100000 60000 80000 100000
state rural
nonstate rural
kernel = epanechnikov, bandwidth = 1501.76

Note: State households are those with at least one member working for the state (in either a state rm or for the government in public administration). Source: Viet Nam Household Living Standards Survey 2004, 2006, and 2008.
constraints on government agencies. State-owned enterprises (SOEs), on the other hand, benet from capital market interventions that lower their borrowing costs; they therefore adopt relatively capital-intensive techniques. Because capital and skills are complementary inputs, more highly educated workers are drawn to these rms and compete to be hired by them. But since SOEs primarily supply goods and
services to (often highly-regulated) domestic markets, their expansionand thus their demand for laboris bounded in ways that do not apply to privately-owned and trade-oriented rms. What both SOEs and civil administration have in common, however, is access to rents, which when distributed among their workers, generate potential for incomes that are higher than the earnings of equivalent workers in competitive industries.4 Nonstate rms, meanwhile, suffer from crowding out in capital markets and so adopt less capital-intensive techniques. In these rms, capital-skills complementarity means they will hire fewer skilled workers, and will offer to compensate them at a lower rate, commensurate with their lower value marginal product.5 In this system, segmentation in the skilled labor market arises indirectly, from capital market distortions, market structure, and budget constraints limiting hiring by state entities. This segmentation is visible as queuing by applicants and demands for up-front payments from prospective employers. The difference in earnings for equivalent workers at each type of rm will persist so long as the favored rms face market or regulatory conditions that generate rents and so encourage them to restrict hiring. Thus capital market policies will lead to job rationing and incentives for corrupt behavior across a broad spectrum of the white-collar labor market, not merely among the few that have direct access to the highest levels of political power. In earlier work (Phan and Coxhead 2013), we uncovered evidence in support of an important part of the above narrative. Viet Nam displays a high degree of industrial policy distortion, with a clear bias in favor of state-owned enterprises in the markets for banking sector credit, equity capital, and land (World Bank 2005, Sj oholm 2008, Hakkala and Kokko 2008, Leung 2009, Nixson and Walters 2010, IMF 2012) and in trade and pricing policies (Athukorala 2006). We discovered a highly signicant state sector premium in earnings and in returns to educationas of 2008, this premium had persisted despite two decades of economic reforms. We also found that family connections to state sector employers increase an individuals own probability of having a state sector job. Together, these results indicate that connections to the state sector increase individual earnings and returns on human capital investments, and that these effects operate at least in part through householdlevel connections. The concerns raised by these ndings are not limited to inequality. A countrys long-term economic growth depends on its ability to accumulate and efciently deploy human capital, and the acquisition of human capital is a key determinant of improvements in individual earning power. The preceding narrative suggests that
For an equivalent account using Chinese labor market data, see Xin Meng (2000). A recent study using labor force data covering ve broad industrial sectors conrms this prediction, nding that in manufacturing, post-secondary qualications earn a premium of only 40%50% over primary education, and concluding that there is currently not a strong demand for workers with either professional training or tertiary education in either low-value or medium-value industries. . . workers with post-secondary qualications are therefore likely to gravitate towards better-remunerated jobs in government and administration and the services (Baulch, Dat, and Thang 2012, p. 2223).
5 4
in the presence of certain policies and capital market distortions, household-level incentives for educational investments will depend, in part, on the likelihood that their children will be able to secure jobs in which skills or credentials are rewarded. If this probability is viewed as small, then returns to additional years of education are perceived as low, and capital-constrained parents will spend less on education and/or withdraw children from school earlier; those children will enter the labor force at a younger age and with less formal training. That perceived returns matter for educational investments is well established in empirical studies. In a study from the Dominican Republic, for example, Jensen (2010) nds that when children are given information on higher measured returns to education, they complete on average 0.200.35 more years of schooling over the next 4 years. Similarly, Jensen (2012) nds that young rural women in India are signicantly more likely to enter the labor market or obtain more schooling instead of getting married and having children if they have access to recruiting services, which increase their awareness of job opportunities. Lower perceived returns to schooling reduce educational investments and the schooling achievements of children. As a result their lifetime earnings proles will be atter and their capacity to invest in the education of their own children will be diminished. In this way, initial household-level disparities in opportunity may become persistent over more than one generation. Moreover, the potential productivity of less-educated workers will also be lower, so the economy as a whole will face a diminished growth rate and lower steady-state income per capita relative to the counterfactual of one in which perceived returns are higher. In this paper, we test the hypothesis that households with close connections to Viet Nams state sector invest more in their childrens human capital. Specically, we ask whether children from households headed by state employees are more likely to attend high school or university, and whether those households spend more on their childrens education. The rst of these questions explores the extensive margin of educational attainment in an economy where only a small minority of school-leavers continue on to higher education. The second question investigates the intensive margin of educational investments in a system in which household spending on discretionary educational items, such as tutors and private schools, is an important component of total educational expenditures. After controlling for characteristics of the potential student, the household head, and the household, we nd robust evidence that children from households whose heads work for the state are more likely to attend university, and that these households also spend signicantly more on their childrens education. Our work extends the existing literature on educational inequality by identifying and focusing on a previously neglected dimension. The state/nonstate dichotomy among households proves to be an important dividing line for investments in the most growth-oriented form of individual and household capital, education. Policies that restrict access to schooling or reduce incentives to remain in school create and
exacerbate inequality, a phenomenon now widely seen as reducing growth (Stiglitz 2012). By discouraging educational investments by some, they also reduce the efcacy of efforts to promote and sustain long-run economic growth. The rest of this paper is organized as follows. In Section II we review basic data and trends on educational attainment and related indicators in Viet Nam. In Section III we motivate a somewhat deeper analysis and conduct an econometric exploration of our main hypotheses. In Section IV we discuss our results, and in Section V we draw some tentative conclusions, speculate on the generalizability of our ndings, and propose some avenues for future research.
II. A New Dimension of Inequality: State and Nonstate Households A. Data
Our primary data source is the Viet Nam Household Living Standards Survey (VHLSS), with rounds in 1993, 1998, 2002, 2004, 2006, and 2008.6 These surveys gather data on household income and expenditure and are designed to measure living conditions and poverty and inequality (Grosh and Glewwe 2000). They are intended to be representative at the national and provincial levels. They also include modules on education and on employment and wages. Early rounds of VHLSS were smaller in size (4,800 households in 1993 and 6,000 in 1998). The survey year 2002 had the largest number of households (29,533). In the most recent three rounds, the number of households from which both expenditure and income data are collected has stabilized at around 9,000.7 In our regressions, we use data from the 2004, 2006, and 2008 rounds. Consistent questionnaires and denitions of variables in all these three survey years allow us to pool data into one estimating model. To measure years of education, most other studies using VHLSS data have used the surveys original schooling year variable, which ranges from 0 through 12 years. We adjust schooling years for highest educational level (junior college corresponds to 14 years of education, a college degree 16 years, masters degree 18 years, and PhD 21 years). As a result, our calculated years of schooling are higher than other studies. In the descriptive analysis in the next section we also use the Viet Nam Enterprise Survey for data on rms. This is an annual survey begun in 2000.8 It covers all enterprises with independent accounting systems that are established under and governed by the Law on State-owned Enterprises, the Law on Cooperatives, the Law
Data from the most recent VHLSS rounds, in 2010 and 2012, are not yet available to us. These are survey subsamples that have both income and expenditure modules. There are also much larger surveys with expenditure modules only, which we do not want to use because data are considered less reliable when income cannot be compared against expenditures. 8 This survey existed in the 1990s, but in a somewhat different form and quality.
7 6
32 ASIAN DEVELOPMENT REVIEW Table 1. Structure of Employment by Type and Ownership (%)
1993 Self employed Wage employed State wage employed Nonstate wage employed 73.4 26.6 4.3 22.2 1998 59.4 40.6 6.7 33.9 2002 60.9 39.1 10.2 28.9 2004 57.0 43.0 14.7 28.3 2006 55.7 44.3 15.4 28.9 2008 55.2 44.8 14.7 30.1
Source: Authors calculations from Viet Nam Household Living Standards Survey, various years.
Table 2. Average Years of Education of Workers in State and Nonstate Sectors

1993 State Nonstate 11.3 5.6 1998 9.9 5.8 2002 11.7 6.7 2004 11.1 6.9 2006 11.3 7.0 2008 11.8 7.3
Table 3. Workers with College Degree in State and Nonstate Sectors (%)
1993 State Nonstate 17.31 0.38 1998 22.22 0.78 2002 27.86 1.38 2004 25.81 0.98 2006 26.07 1.12 2008 29.92 1.77
on Enterprises, and the Law on Foreign Investment in Viet Nam. All formal sectors and industries, including agriculture, are covered. However, because the criterion for inclusion in the survey is establishment and governance under the law, the informal sector is largely ignored. The survey questionnaire includes various business and production activities: labor and employment, incomes of employees, number of establishments, assets and liabilities, investments, capital stock, production costs, turnover, products, prots, inventories, taxes, research and development investments, IT applications, and others.
B. Employment and Wages in State and Nonstate Firms
In the course of Viet Nams more than 20 years of transition to a market economy, the state sector has retained an extraordinarily prominent role. SOEs account for between one-third and two-fths of economic activity. According to VHLSS data, employment in SOEs has even risen as a share of total wage labor employment (which itself has risen substantially since the 1990s). Table 1 provides a summary of employment by ownership type. Table 2 conrms that state sector rms are far more intensive in their use of educated workers than are nonstate rms. Although the average education of nonstate workers has risen steadily, from 5.6 years in 1993 to 7.8 in 2008, this gure still remains far below years of education in state sector rms. Moreover, as Table 3 shows, the proportion of workers in nonstate rms with a college degree remains extremely low (1.8% in 2008) by comparison with state rms (30%).

Table 4. Capital Intensity (D million per worker)
2000 State rms Private rms Foreign rms 194 53 391 2002 234 49 378 2004 329 65 292 2006 407 90 499 2008 324 99 478
Source: Authors calculations from Viet Nam Enterprise Survey, various years.
Table 5. Hourly Wage of State vs. Nonstate Workers (D thousand per hour)
1993 State Nonstate State/Nonstate ratio 1.71 1.89 0.90 1998 3.23 2.73 1.18 2002 5.49 3.14 1.75 2004 5.26 3.35 1.57 2006 5.61 3.65 1.54 2008 8.02 5.01 1.60
One reason for the higher skill intensity of state sector employment is that some jobs in this sector are government administrative positions that require higher education. Another important reason is technological: capital and skills are complements in production (Griliches 1969). Table 4 reveals much higher capital per worker in state and foreign-invested rms compared to nonstate rms. In the case of SOEs this is undoubtedly the result of their privileged access to capital. The combination of technological factors and policy distortions results in higher demand for skilled workers by SOEs than by private rms. These differentials in capital and skill intensities between state and nonstate rms no doubt contribute to the discrepancy in average hourly wages, shown in Table 5. In 1993, workers in the state sector earned less on average than workers in the nonstate sector. But this was prior to the implementation of most market and labor reforms. Since then, they have earned about 40% more per hour than nonstate workers. Of course, these are simple averages and do not take account of differences in the composition of the labor force in each type of rm. But as mentioned earlier, our earlier econometric investigation conrms that workers in state rms indeed receive higher returns to education, even after controlling for many differences in worker and job characteristics and that, far from being erased by the economic transition, this disparity has increased over time (Phan and Coxhead 2013).
C. Educational Attainment and Investments by State and Nonstate Households
There is a large literature on the inuence of family environment on human capital investments (for a recent and highly relevant survey, see Yi et al. 2011).9
9 See especially their footnote 21, a discussion of the true meaning of parental education which has been shown to have a strong inuence over schooling outcomes for children. Parts of this literature explore just what parental education stands forwhether income, or cultural norms favoring education, or other things.
For Viet Nam, the VHLSS data reveal well-known sources of cross-household variation in schooling rates and educational expenditures such as between urban and rural subpopulations, ethnic minority and Kinh majority households, households in different regions of the country, and of course variation based on differences in per capita income (Vu 2012).10 These are the familiar break points in econometric analyses of educational attainment. But even after controlling for these sources of variation, differences in human capital investments persist. What explains this? One clue to the remaining differences comes from our prior work on returns to education. Our estimates of the treatment equation for employment in the state sector revealed that workers in households where another household member works for a state rm are far more likely themselves to have state sector jobs, even after controlling for other household factors and individual characteristics. This empirical insight points to a previously unexamined fracture line in the educational attainment and investment data: variation by state versus nonstate households. We now turn to the examination of these data. In this section we use three different measures of educational attainment and investments: average years of schooling, net and gross enrollment rates in different levels of schooling, and household educational expenditures.11 No matter which measure we use, state households (those with at least one member working for a state rm or for the government) tend to invest more in the education of their children. Figure 2 shows that for all age groups, the average years of schooling of children from state households are higher than those of children from nonstate households. The differences in means are small at lower levels of education, but widen progressively through college. Figures 3a and 3b reveal a similar story. Net and gross enrollment rates are higher for children from state households in both rural and urban areas. Again, the differences are greatest at the college levelthe level most relevant for the skilled labor market. Finally, Figure 4 conrms that state households tend to spend more on their childrens education than do nonstate households, in both rural and urban areas. In summary, the available data point to a large premium for educated workers employed in the state sector. If the presence of a parent or other elder with a state sector job means easier access for others in the same household, then schooling decisions appear to be driven (at least in part) by access to jobs in the favored sector. There are, however, many covariates to be controlled for when analyzing educational investments. Therefore we now turn to a more rigorous examination of the determinants of household educational investment.
And compare these differentials in similar countries, such as the PRC (Wang et al. 2011). Net enrollment rate is the ratio of enrolled children in the ofcial school age group over the total number of children in that age group. Gross enrollment rate is the ratio of enrolled children of all ages over the total number of children in the ofcial school age group.
11
10

Figure 2. Average Years of Schooling
12.0 10.0 8.0 6.0 4.0 2.0 0.0 College Age Group (1822) High School Age Group Primary and Middle (1517) School Age Group (614) Nonstate Households
State Households
Source: Viet Nam Household Living Standards Survey, 2006.
III. Determinants of Household Educational Investments A. Empirical Strategy
Our goal is to understand disparities in household educational investments. As is well known, education is often terminated at discrete points such as the completion of a given level of schooling. Therefore we focus on specic schooling decisions. These are rst, the decision to enroll a child in high school given completion of middle school, and second, the decision to enroll a potential student in university given that he or she has completed high school. Our empirical strategy is based on well-established human capital theory (Schultz 1960 and 1963, Becker 1964). This maintains that household demand for education is affected by expected returns (r) and costs (c, including direct and opportunity costs) of education: yi = f (ri , ci , Di , u i ) = a + ri + ci + Di + u i (1)
where yi is a latent variable capturing the decision to enroll in university or high school by potential student i; Di is a vector of controls including demographic variables such as gender, age, ethnicity, geographic variables, and other exogenous inuences; and ui is a random error term that is assumed to be independently and identically distributed with expected value E(ui ) = 0. We have assumed a linear functional form for the explanatory variables.
36 ASIAN DEVELOPMENT REVIEW Figure 3a. Net Enrollment Rate
1.00 0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 Primary School Middle School High School College and Junior College
0.28 0.18 0.15 0.09 0.61 0.61 0.90 0.90 0.86 0.88 0.88 0.79 0.80 0.74 0.79
0.45
Urban State Households Rural State Households

Urban Nonstate Households Rural Nonstate Households
Figure 3b. Gross Enrollment Rate
1.40
1.25
1.20
1.05
1.00 0.80
0.99
1.01
1.06 1.00 0.95
1.02 0.94 0.91
1.06
0.69
0.60 0.40 0.20 0.00 Primary School Middle School High School College and Junior College
0.35 0.20 0.21 0.10
Urban State Households Rural State Households

Urban Nonstate Households Rural Nonstate Households

Figure 4. Mean Household Educational Expenditure (D thousand)
3,500 3,000 2,500 2,000 1,500 1,000 500 0 Urban State Households
2,941
1,871
1,839
915
Rural Nonstate Households
The observed variable is yi , which is a dummy indicating whether the potential student i enrolls in university (or high school). Specically, yi = 1 if yi > 0, = 0 otherwise. Thus, the frequency of the observed outcome, or the probability of enrolling in university or high school is equal to Pi = Pr ob( yi = 1) = Pr ob( yi > 0) = Pr ob( + ri + ci + Di + u i > 0) = Pr ob(u i > ri ci Di ) = 1 G ( ri ci Di ) where G(.) is the cumulative distribution of the error term (Maddala 1983). If we use the logistic cumulative distribution function and take the log of the odds ratios, then we have a logit model: logit ( Pi ) = log [ Pi /(1 Pi )]) = + ri + ci + Di . (2) and
Our primary focus is on explaining differential incentives to invest in education, given differences in expected returns to education in private and state sector
jobs. That is, the decision to progress from middle to high school, or to enroll in tertiary education, takes into account the trajectory of returns to education in state and nonstate employment adjusted by the probabilities of employment in either sector. From Phan and Coxhead (2013) we know that the average rate of return to education in the state sector (rS ) is higher than in the nonstate sector (rN ). If i is the probability of state sector employment for potential student i, then his expected returns to education is ri = rS i + rN (1 i ). Under the assumption of uniformity within each employment sector, we can normalize rN = 1 and rewrite ri = 1 + [rS 1] i . Substituting this into the previous equation, our logit model becomes: logit ( Pi ) = ( + ) + (r S 1)i + ci + Di . (3)
Equation (3) is our estimating equation.12 The variable of interest is i , the probability of getting a state sector job. Lacking direct observations on this, we use two proxies: an SOE dummy which equals 1 if household head works for a state rm, and a civil dummy equal to 1 if the head of household works in public administration. These dummies aim to capture a households connection to the state sector, whether state-owned enterprises or government. The null hypothesis is that these dummies are not statistically different from zero; that is, the sector of employment of a potential students household head has no effect on probability of high school or university enrollment. Rejection of the null is evidence of state employment as a source of household differences in educational incentives. We discuss the implications of such a nding in Section IV . We include three other sets of explanatory variables: (i) household characteristics, specically log of household income,13 household composition by age, urban or regional dummies, ethnic minority dummy (1 if household is not of Kinh or Chinese ethnicities); (ii) household head characteristics, namely age, gender (1 if male), years of schooling, and skill dummy (1 if household head holds a skilled job); and (iii) child characteristics, comprising gender (1 if male) and age. Among the control variables, those of greatest economic interest are the households ability to nance education, as measured by household income, and the household heads years of schooling. Obviously, wealthier households are more able to pay for education and to keep children out of the labor force until a later age. A
The model is estimated using maximum likelihood method using Stata. We also ran a series of regressions using household expenditure, which yield similar results. Details are available from the authors.
13 12
large empirical literature conrms that more educated parents tend to invest more in the education of their own children.14 Controlling for parents education also helps reduce (although not completely eliminate) omitted variable biases, because this variable serves as proxy for many monetary and nonmonetary factors that affect schooling decisions. For example, we have no information on childrens intelligence or ability: smarter children do better in school and are more likely to attend high school and college, other things equal. Another omitted variable is preferences: some children may love schooling for the sake of knowledge, not for the monetary returns. If these omitted variables are correlated with employment sector of household head, their omission causes upward biases in estimates of the SOE and civil dummies. But it is reasonable to assume that parents who are smarter and who love education tend to transfer their intelligence and love of schooling to their children. Such parents also tend to attain more education. So parents education partly proxies and controls for the two omitted variables, childrens ability and love of schooling. In estimating determinants of school enrollment probability, we look only at the children and grandchildren of household heads. We do this intentionally to avoid potential endogeneity due to reverse causality. Since we examine the impact of the older generations employment on educational investments for the younger generation and it is not plausible that the younger generations educational achievement increases the likelihood of state employment by the older generation, there is no reverse causality problem. In addition to estimating the probability of university or high school enrollment, we also estimate household educational expenditures, another measure of household educational investment. Since many households have zero educational expenditure, our sample suffers from censoring. As a result, we use a Tobit model for this estimation.
B. Results
Estimation results are presented in Tables 6 through 9 (a table of summary statistics of key variables is in the Appendix). In Tables 6 and 7, we use logistic regressions to estimate school enrollment probabilities (see equation 3). We report both estimated coefcients and odds ratios. Standard errors are robust and adjusted for clustering at the commune level. In both Tables 6 and 7, most variables have coefcients of expected signs and statistical signicance. For brevity, we restrict our discussion to the main variables of interest, the civil and SOE dummies. In Table 6, these dummies are not statistically signicant. The household heads employment does not seem to affect childrens probability of enrolling in high school. But in Table 7, both dummies are positive and statistically signicant. The estimated odds ratios imply that the university enrollment probability of a potential student whose
14 A recent study of rural households in southern Viet Nam conrms that as agricultural incomes increase, the schooling years of children (especially daughters) increase. See Edmonds (2005).
40 ASIAN DEVELOPMENT REVIEW Table 6. Determinants of High School Enrollment Probability

Odds Ratio Household Head Characteristics Head works for SOE (SOE dummy) Head works for govt (civil dummy) Head has a skilled job (skill dummy) Head years of schooling Gender of household head (male = 1) Age of household head Household Characteristics Log of household income Ethnic minority dummy = 1 Controls Family demographic structure Childs age and sex Region Year Constant N Pseudo R-squared 1.264 0.998 1.392 1.080 0.918 1.011 1.224 1.481 Coeff. 0.234 0.002 0.331 0.077 0.086 0.011 0.202 0.393 Yes Yes Yes Yes 3.348 12,333 0.299 S.E. 0.161 0.117 0.113 0.009 0.065 0.003 0.036 0.099 p-value 0.146 0.984 0.003 0.000 0.185 0.000 0.000 0.000
0.448
0.000
= statistically signicant at 5% level or less, = statistically signicant at 5%10% level. Notes: 1. Estimates are from logit regressions with robust standard errors adjusted for clustering at commune level. 2. Dependent variable: dummy equals 1 if potential student was currently enrolling in high school during survey year; sample includes children and grandchildren (who already obtained middle school diploma) of household heads. 3. Controls: family demographic structure (number of children 5 years old, children 610 years old, children 1114 years old, children 1822 years old, family members 2365 years old); region (8 administrative regions); year (dummies for 2006 and 2008). Source: Authors computations using data from Viet Nam Household Living Standards Survey 2004, 2006, and 2008.
household head works in an SOE is 1.3 times greater than that of someone whose household head works elsewhere. For someone whose household head works in public administration, the probability is 1.4 times greater. That connection to the state sector is not a source of inuence in schooling decisions at lower levels of education is consistent with a theoretical model in Phan and Coxhead (2013). In that paper, we argue that the rising wage premium of state sector jobs is a result of government intervention in both capital markets and the market for skilled labor. As the government allocates capital to the state sector, this raises the demand for skilled labor in that sector (because of skillcapital complementarity). But since access to state sector jobs is restricted, the increase in demand for skilled labor translates to higher earnings for skilled workers in this sector. These distortions do not apply to unskilled labor markets, which require lower education. Therefore we do not expect a relationship between state connection and investment in lower levels of schooling. In Tables 8 and 9, we estimate the determinants of households educational expenditures and the share of education in total household expenditures. Once again, employment of the household head is inuential. In Table 8, the SOE and civil dummies have positive and statistically signicant impacts on (log of) household

Table 7. Determinants of University or College Enrollment Probability
Odds Ratio Household Head Characteristics Head works for SOE (SOE dummy) Head works for govt (civil dummy) Head has a skilled job (skill dummy) Head years of schooling Gender of household head (male = 1) Age of household head Household Characteristics Log of household income Ethnic minority dummy = 1 Controls Family demographic structure Childs age and sex Region Year Constant N Pseudo R-squared 1.332 1.438 1.210 1.115 0.922 1.018 1.325 0.546 Coeff. 0.287 0.363 0.191 0.109 0.081 0.018 0.281 0.604 Yes Yes Yes Yes 4.585 8,531 0.106 S.E. 0.148 0.145 0.117 0.012 0.087 0.005 0.044 0.177 p-value 0.053 0.012 0.102 0.000 0.352 0.000 0.000 0.001
0.558
0.000
= statistically signicant at 5% level or less, = statistically signicant at 5%10% level. Notes: 1. Estimates are from logit regressions with robust standard errors adjusted for clustering at commune level. 2. Dependent variable: dummy equals 1 if potential student was currently enrolling in high school during survey year; sample includes children and grandchildren (who already obtained middle school diploma) of household heads. 3. Controls: family demographic structure (number of children 5 years old, children 610 years old, children 1114 years old, children 1822 years old, family members 2365 years old); region (8 administrative regions); year (dummies for 2006 and 2008). Source: Authors computations using data from Viet Nam Household Living Standards Survey 2004, 2006, and 2008.
education expenditure, after controlling for household income and other characteristics. In Table 9, the civil dummy has a positive and statistically signicant impact on the share of household education expenditure in total expenditure. To summarize these results, even after controlling for the usual determinants of educational investments we decisively reject the null hypothesis of no effect of state employment of parents on investments in the education of children. The implication is that due to underlying distortions that give rise to state sector wage premia and labor market segmentation, factors other than ability are helping determine who among children receives more and better educational opportunities. This has consequences for both equity and growth in the long run.
IV. Discussion A. Inequality
When there is rationing in the market for skills, discrepancies in the household distribution of assets, including claims on state sector jobs, can be transmitted across
42 ASIAN DEVELOPMENT REVIEW Table 8: Determinants of Household Educational Expenditure: Total Spending
Dep. var: hhold educational expenditure (D) Household Head Characteristics Head works for SOE (SOE dummy) Head works for govt (civil dummy) Head has a skilled job (skill dummy) Head years of schooling Gender of household head (male = 1) Age of household head Household Characteristics Log of household income Ethnic minority dummy = 1 Controls Family demographic structure Childs age and sex Region Year Constant N Pseudo R-squared Coeff. 0.495 0.503 0.313 0.190 0.414 0.045 0.812 0.937 Yes Yes Yes Yes 6.726 27,563 0.115 S.E. 0.128 0.101 0.099 0.009 0.068 0.003 0.036 0.107 p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
0.401
0.000
= statistically signicant at 5% level or less, = statistically signicant at 5%10% level. Notes: 1. Estimates are from Tobit regressions with robust standard errors adjusted for clustering at commune level. 2. Dependent variable: log of educational expenditure. 3. Controls: family demographic structure (number of children 5 years old, children 610 years old, children 1114 years old, children 1822 years old, family members 2365 years old); region (8 administrative regions); year (dummies for 2006 and 2008). Source: Authors computations using data from Viet Nam Household Living Standards Survey 2004, 2006, and 2008.
generations. It follows (although we do not address this directly) that unequal access to education and an unequal distribution of incentives to enroll in tertiary education could both exacerbate initial inequality and increase the rate at which the fortunes of favored and less favored households diverge during the course of economic growth and transition to a market economy. Our main result elaborates a point that has been made before in other settings. Becker and Tomes (1993), for example, explore the potential of capital market imperfections to lower parents investments in their childrens acquisition of human capital. The data from Viet Nam suggest two potentially interesting twists on this account. First, it is not only widely-studied constraints such as access to credit that distinguish the fortunes of one family from those of another. Labor market segmentation may also give rise to a structural disparity that is hard for households to bridge. Second, and related to this, in our data it appears that a key asset that parents can bequeath to their children is access to state jobs, and this in turn inuences the expected rate of return to the latters education. But access, like human capital itself, does not function well as collateral. Two implications follow. First, even if a household could self-nance a higher level of education for its children, that investment might still not pay off in the labor market. Second, public policies aimed

Table 9. Determinants of Household Educational Expenditure: Expenditure Share
Dep. var: share of education in total hhold expenditure Household head characteristics Head works for SOE (SOE dummy) Head works for govt (civil dummy) Head has a skilled job (skill dummy) Head years of schooling Gender of household head (male = 1) Age of household head Household Characteristics Log of household income Ethnic minority dummy = 1 Controls Family demographic structure Childs age and sex Region Year Constant N Pseudo R-squared Coeff. 0.004 0.007 0.010 0.004 0.011 0.001 0.008 0.023 Yes Yes Yes Yes 0.091 27,563 0.396 S.E. 0.003 0.003 0.003 0.000 0.002 0.000 0.001 0.002 p-value 0.228 0.009 0.000 0.000 0.000 0.000 0.000 0.000
0.009
0.000
= statistically signicant at 5% level or less, = statistically signicant at 5%10% level. Notes: 1. Estimates are from Tobit regressions with robust standard errors adjusted for clustering at commune level. 2. Dependent variable: share of education in total household expenditure. 3. Controls: family demographic structure (number of children 5 years old, children 610 years old, children 1114 years old, children 1822 years old, family members 2365 years old); region (8 administrative regions); year (dummies for 2006 and 2008). Source: Authors computations using data from Viet Nam Household Living Standards Survey 2004, 2006, and 2008.
at broadening access to education might not raise enrollments if incentives are reduced by distortions that lower the probability of getting a suitably rewarding job.
B. Growth
In low-income countries, and especially in those undergoing transition, aggregate growth can be sustained for a time through mobilization of underutilized resources and efciency gains due to economic policy reforms. But in the longer run, continued growth relies on increases in total factor productivity and investments in reproducible factors of production, most notably human capital. Developing economies that underinvest in human capital risk landing in a middle-income trap in which their continued progress up the product quality ladder or participation in global production networks is constrained by human capital shortages. This leaves them reliant on other sources of income, typically natural resource extraction or low-skill manufactures, for which global competition among suppliers is intense. Our results suggest that at present, a large majority of workers in Viet Nam have a low probability of nding employment in an industry or occupation where returns to higher-quality education or tertiary credentials will justify the cost of
their acquisition. Even if other constraints (such as access to credit) do not bind, some fraction of high school and college age children are rationally opting out of education to join the unskilled or semi-skilled labor force. It may be possible for these individuals to acquire additional skills later in life, for example through vocational training or on-the-job training, but at present, neither of these channels appears promising in Viet Nam. Vocational training is not widespread nor is it respected by employers in Viet Nam (Chirot and Wilkinson 2010). Due to the crowding out of banking sector capital for small and medium enterprises, capital intensity outside of the state sectors and a few other well-connected industries is very low. Accordingly, returns to education and skills are also low in these industries (Baulch, Dat, and Thang 2012). Viet Nam therefore confronts a medium-run future in which a large part of its adult labor force is inadequately prepared to support a move up the product quality ladder, from assembly to innovation. Finally, as discussed at the start of this paper, inequality and long-run economic growth are increasingly widely regarded as positively linked, with access to education and incentives to acquire human capital as key conduits. This was succinctly stated in a recent report on rising inequality around the world: If income gaps get wide enough, they can lead to less equality of opportunity, especially in education. Social mobility in America, contrary to conventional wisdom, is lower than in most European countries. The gap in test scores between rich and poor American children is roughly 30%40% wider than it was 25 years ago. And by some measures class mobility is even stickier in [the Peoples Republic of] China than in America.15 The reports primary proposed remedy is to dismantle monopolies and vested interests, be they state-owned enterprises in the PRC, Indias unequal oligarchy or worse, or big banks on Wall Street.16 Although by the usual measures inequality in Viet Nam is much lower than in most economies, so too is its per capita income. Therefore, the prospect of a lower growth rate due to inequality in human capital investments must be treated very seriously.
C. Generalizability
Our data are drawn from just one developing economy, but there are strong indications that the analysis is relevant to a larger group. In many countries with less restrictive regimes than Viet Nam, SOEs enjoy qualitatively similar positions in terms of their access to domestic capital, dominance in domestic markets, and
15 16
Crony Tigers, Divided Dragons, The Economist, 13 October 2012. True Progressivism, The Economist, 13 October 2012. The quote is attributed to Raghuram Rajan.
connections to political decision makers. State involvement in developing economies is not disappearing as quickly as was perhaps expected in the post-Cold-War, Washington Consensus era. In fact there are some recent signs of a resurgence of state engagement.17 Thus the role of personal connections, so highly signicant in the case of one rather heavily distorted developing economy, may well be present, albeit less starkly so, in many others. Our results are necessarily conditional on a set of parameters that is subject to change along with the economic policies of Viet Nam. In particular, the statenonstate gap in labor productivity (and thus in returns to skills and earnings) could be reduced or eliminated by structural reforms that reduce or eradicate the special economic status of state sector enterprises. Alternatively, greater access by rms, and especially by small and medium enterprises, to foreign capital would be inuential since it would offset the productivity-reducing effects of capital crowding-out by SOEs. Whether either of these changes is sufcient to reverse the trend toward greater disparity that we have documented remains an open question. Further analysis with the most recent rounds of VHLSS (2010 and 2012), once they become available, will help answer this question.
V. Conclusion
In this paper we have explored a hitherto neglected dimension of inequalitythat arising in a transitional economy between households with access to rationed, rent-yielding state sector jobs and those without. Data from several rounds of Viet Nams household living standards surveys conrm a substantial difference in per capita incomes across this divide. Our research then reveals that the statenonstate dichotomy also marks a signicant discrepancy in households educational investments in their children. When state employers are the largest demanders of skills, and when they compensate those skills far more generously on average than do nonstate employers, there is competition among workers to win state jobs. These jobs are effectively rationed by public sector budget constraints and limited domestic markets for the goods and services produced by state-owned corporations. As a result, connections based on family ties become important factors raising the probability of winning a state sector job. Households for which these connections are strong tend to invest more heavily than others in education and credentials for their offspring. In other households, the data suggest that opportunity cost drives
17 The Great Slowdown, The Economist, 21 July 2012. The article comments on this as follows: Sadly, many emerging-world governments have interpreted the crisis in rich-world nance as a reason to preserve a more muscular role for the state. [The Peoples Republic of] China has reserved some sectors for state-owned enterprises. In Brazil the big state-controlled oil company, Petrobras, and the state-controlled banks have become virtual appendages of government policy. Having so much leverage over the economy is indeed helpful during a crisis, but in the long run it will stie competition, starve the private sector of capital, deter foreign investment and know-how, and breed corruption.
children whose parents cannot help them win lucrative state sector jobs out of education earlier and at lower levels, and that those households invest less overall in education of their children. These divisions depend on policy, not on secular development trends, and so could well persist across generations, creating a twotrack labor market of princelings and paupers. International evidence suggests further that unequal incentives to invest in human capital will in the long run lead to slower rates of aggregate economic growth. If so, then Viet Nam could face an earlier onset of the middle income trap than would be the case under the counterfactual of a more level playing eld for labor and education. Prior experience of other regional economies like Indonesia, Malaysia, and Thailand pinpoints underinvestment in human capital formation as a key contributing factor to growth slowdowns and vulnerability to economic crises (Booth 2003, Warr 2005, Coxhead and Li 2008). These are important lessons for policymakers in Viet Nam. The insight from our study is that skills shortages will not necessarily be resolved simply by increasing the supply of schools and teachers. There is a potentially binding demand-side constraint on human capital deepening, and this can be eliminated only by deeper structural reforms.
References
Adamchik, Vera, and Arjun Bedi. 2000. Wage Differentials between the Public and Private Sectors: Evidence from an Economy in Transition. Labour Economics 7(2): 20324. Athukorala, Prema-chandra. 2006. Trade Policy Reforms and the Structure of Protection in Viet Nam. World Economy 29(2): 161187. Baulch, Bob, Vu Hoang Dat, and Nguyen Thang. 2012. Do Vietnamese Schools Provide the Right Education for an Industrializing Country? Young Lives Project Working Paper No. 81. Becker, Gary S. 1964. Human Capital: A Theoretical Analysis With Special Reference to Education. National Bureau for Economic Research, Columbia University Press, New York and London. Becker, Gary, and Nigel Tomes. 1993. Human Capital and the Rise and Fall of Families. In Gary Becker, ed. Human Capital. 3rd edition. Chicago: University of Chicago Press. Berg, Andrew, and Jonathan Ostry. 2011. Inequality and Unsustainable Growth: Two Sides of the Same Coin? IMF Staff Discussion Note No. SDN 11/08. Washington, DC: International Monetary Fund. Booth, Anne. 2003. Education and Economic Development in Southeast Asia: Myths and Realities. In Kwama Sundarman Jomo, ed. Southeast Asian Paper Tigers? From Miracle to Debacle and Beyond. London: RoutledgeCurzon. Chirot, L., and B. Wilkinson. 2010. The Intangibles of Excellence: Governance and the Quest to Build a Vietnamese Apex Research University. http://www.ash.harvard.edu/extension/ ash/docs/Apex.pdf, accessed 15 May 2103. Coxhead, Ian, and Muqun Li. 2008. Prospects for Skills-based Exports in Resource-rich Developing Economies: Indonesia in Comparative Perspective. Bulletin of Indonesian Economic Studies 44(2): 199228. Edmonds, Eric. 2005. Does Child Labor Decline with Improving Economic Status? Journal of Human Resources 40(1): 7799.

Griliches, Zvi. 1969. CapitalSkill Complementarity. Review of Economics and Statistics 51(4): 46568. Grosh, Margeret E., and Paul Glewwe. 2000. Designing Household Survey Questionnaires for Developing Countries: Lessons From 15 Years of the Living Standards Measurement Study. Washington, DC: World Bank. Hakkala, Katariina, and Ari Kokko. 2008. The State and the Private Sector. In Ari Kokko, ed. Viet Nam: 20 Years of Doi Moi. Ha Noi: Gioi Publishers. International Monetary Fund. 2012. 2012 Article IV Consultation. IMF Country Report No. 12/165. Washington, DC: International Monetary Fund. Jensen, Robert. 2010. The (Perceived) Returns to Education and the Demand for Schooling. Quarterly Journal of Economics 125(2): 515548. . 2012. Do Labor Market Opportunities Affect Young Womens Work and Family Decisions? Experimental Evidence from India. Quarterly Journal of Economics 127(2): 753792. Leung, Suiwah. 2009. Banking and Financial Sector Reforms in Viet Nam. ASEAN Economic Bulletin 26(1): 4457. Maddala, G. S. 1983. Limited Dependent and Qualitative Variables in Econometrics. New York: Cambridge University Press. Munich, Daniel, Jan Svejnar, and Katherine Terrell. 2005. Returns to Human Capital under the Communist Wage Grid and During the Transition to a Market Economy. Review of Economics and Statistics 87(1): 100123. Nixson, Frederick, and Bernard Walters. 2010. Vietnamese Enterprises: Toward Global Competitiveness. Hanoi: United Nations Development Programme. Phan, Diep, and Ian Coxhead. 2013. Long-run Costs of Piecemeal Reform: Wage Inequality and Returns to Education in Vietnam. Journal of Comparative Economics, published online 16 April 2013. doi: http://dx.doi.org/10.1016/j.jce.2013.04.001 Pincus, Jonathan, Vu Thanh Tu Anh, Phan Duy Nghia, Benjamin Wilkinson, and Nguyen Xuan Thanh. 2012. Structural Reform for Growth, Equity and National Sovereignty. Ho Chi Minh: Fulbright Economics Teaching Program Discussion Papers. Available: http://www.fetp.edu.vn/en/policy-papers/discussion-papers/ Schultz, Theodore W. 1960. Capital Formation by Education. Journal of Political Economy 68(6): 571583. . 1963. The Economic Value of Education. New York: John Wiley. Sj oholm, Fredrik. 2008. State-owned Enterprises and Equitisation in Viet Nam. In Ari Kokko, ed., Viet Nam: 20 Years of Doi Moi. Ha Noi: Gioi Publishers. Stiglitz, Joseph. 2012. The Price of Inequality: How Todays Divided Society Endangers Our Future. New York: W.W. Norton. United Nations Development Programme. 2006. The State as Investor: Equitization, Privatization, and the Transformation of SOEs in Viet Nam. UNDP Viet Nam Policy Dialogue Paper No. 2006/3. Ha Noi: UNDP. Vu, Hoang Linh. 2012. An Overview of Access to and Inequality in the Education System of Viet Nam. Asia-Pacic Development Journal 19(1): 3762. Warr, Peter. 2005. Boom, Bust and Beyond. In Peter Warr, ed. Thailand Beyond the Crisis. London and New York: RoutledgeCurzon. Wang, Xiaobing, Chengfang Li, Linxiu Zhang, Renfu Luo, Thomas Glauben, Yaojiang Shi, Scott Rozelle, and Brian Sharbono. 2011. College Education and the Poor in China: Documenting the Hurdles to Educational Attainment and College Matriculation. REAP Working Paper no. 213. Available: reapchina.org/reap.stanford.edu
48 ASIAN DEVELOPMENT REVIEW World Bank. 2005. Viet Nam Business: Viet Nam Development Report 2006. World Bank Report No. 34474-VN. Washington, DC. Xin Meng. 2000. Labour Market Reform in the Peoples Republic of China. Cambridge: Cambridge University Press. Yi, Hongmei, Linxiu Zhang, Renfu Luo, Yaojiang Shi, Di Mo, Xinxin Chen, Carl Brinton, and Scott Rozelle. 2011. Dropping Out: Why are Students Leaving Junior High in the Peoples Republic of Chinas Poor Rural Areas? REAP Working Paper No. 227. Available: reapchina.org/reap.stanford.edu Appendix: Summary Statistics of Main Variables
2004 No. households HH head works for civil admin (%) HH head works for SOE (%) HH Head education (years) (s.d.) Ethnic minority HH (%) HH per capita income (million VND) (s.d.) HH per capita expenditure (million VND) (s.d.) HH expenditure on education (million VND) (s.d.) HH with zero educ. expenditure (%) No. individuals Completed middle school (%) Of which, attended high school (%) Completed high school (%) Of which, attended tertiary instn (%)
Notes: Data are from VHLSS 2004, 2006, and 2008.
2006 9,192 6.17 3.23 7.13 (3.99) 15.06 15,710 (40,801) 6,130 (5,019) 1,397 (2,521) 32.04 39,071 18.69 26.03 10.63 19.25
2008 9,189 6.26 2.83 7.35 (4.16) 15.00 24,596 (67,443) 8,233 (6,451) 1,987 (4,028) 34.01 38,253 19.60 21.49 12.02 20.13
9,188 6.92 3.17 7.01 (3.99) 14.30 11,887 (24,726) 4,927 (4,410) 1,121 (2,167) 28.50 40,439 22.21 19.60 11.14 13.78
Foreign Firms and Indigenous Technology Development in the Peoples Republic of China
FREDRIK SJOHOLM AND NANNAN LUNDIN
The Peoples Republic of China (PRC) is currently promoting indigenous technology development through support of Chinese rms and, arguably, by restricting operations of foreign multinational rms. This policy seems to overlook the impact of foreign rms on technology development in local rms. For instance, technology might leak out to local rms though spillovers. Moreover, competition from foreign rms might force local rms to engage in technology development. We examine the impact of foreign direct investment (FDI) on technology development in the PRC. We start by surveying a large and growing literature on FDI and spillovers in the country. Most previous studies nd evidence of positive spillovers. We then continue to examine the effect of FDI on competition in the Chinese manufacturing sector and the effect of competition on rms research and development (R&D). Our analysis is conducted on a large dataset including all large- and medium-sized Chinese rms over the period 19982004. Our results show that FDI increases competition but there are no strong indications of competition affecting investments in R&D. Keywords: Peoples Republic of China, FDI, spillovers, technology, R&D JEL codes: F23, L11, O31
I. Introduction
Rapid development of the Peoples Republic of China (PRC) during the last few decades is closely related to its success in attracting foreign direct investment (FDI). The PRC is one of the worlds largest recipients of FDI which has substantially contributed to production and export. Moreover, FDI to the PRC, according to some reports, has undergone a structural change away from simple manufacturing towards more technology-intensive activities. For instance, the PRC has become the third most important offshore research and development (R&D) location for multinational enterprises (MNEs) according to a survey by UNCTAD (2005).
Fredrik Sj oholm is Professor at the Department of Economics, Lund University and the Research Institute of Industrial Economics. Nannan Lundin is Managing Director at Tekfors Ltd. The authors are grateful for comments and suggestions made by participants at the ADB Conference on Development Issues in Asia held in November 2012 and from two anonymous referees. Xiaojing Guan, He Ping, and Jinchang Qian from the National Bureau of Statistics of China have been helpful in providing the data. Sj oholm gratefully acknowledges nancial support from the Ragnar S oderberg Foundation.
Despite the increase in FDI and R&D, there is a concern in the PRC that inows of FDI do not contribute to technology development to the same extent that they contribute to production and export. It is noted by policymakers and academics alike that while foreign rms account for a large share of exports and production, their share of R&D is small. The recent emphasis on indigenous innovation and indigenous capacity building in Chinese science and technology policy partly reects an uncertain and even skeptical attitude towards FDI. However, the discussion on FDI and R&D in the PRC neglects how indigenous technology development is affected by FDI. Such an effect could arise if, for instance, domestic rms learn from foreign-owned rms. FDI might also affect the competitive pressure in the market which, in turn, could affect the amount of technology development in domestically-owned rms. The expected impact of FDI on the R&D of domestically-owned (indigenous) rms, however, is not clear. First and as previously mentioned, there could be a demonstration effect or technology externalities from FDI that might affect productivity in domestic rms. This effect is typically attributed to a spillover effect from foreign to domestic rms. A large number of studies have identied such spillovers in various countries. The positive effect on productivity could be caused by technological externalities but could also capture other aspects such as more efcient use of existing technologies through a competition effect. There could also be an effect of FDI on R&D in domestic rms. An impact on R&D is perhaps more closely related to technology development than the above mentioned spillover effect on productivity. The most obvious channel would be through the impact of FDI on the market structure. The direction of this effect is more uncertain since it depends both on how FDI affects market structure and how market structure affects R&D. Starting with FDI and market structure, foreign rms might increase the degree of competition in the local market but it could also happen that successful foreign rms force local rms to exit the market with a resulting increased industry concentration. Moreover, increased competition could both increase R&D, by rms struggling to compete, or decrease R&D because of diminishing monopoly rents. This paper examines how FDI might affect technology development in local Chinese rms. We will examine both general productivity effects through spillovers from FDI and the effect of FDI on market structure and how market structure, in turn, affects investments in R&D. Our section on spillovers surveys the large and growing existing literature on the PRC and concludes that there is ample evidence of positive spillovers. The exact nature of these spillovers and how they are affected by various factors are less clear. We then continue with our second issue where we use rm-level data containing detailed information on operational and R&D activities of large- and medium-sized Chinese manufacturing rms for the period 19982004. Our results suggest that FDI tends to increase competition, as measured by price cost margins, but there is no visible effect of competition on R&D intensity in Chinese rms.
FOREIGN FIRMS AND INDIGENOUS TECHNOLOGY DEVELOPMENT 51
The paper is organized as follows. Section II discusses FDI and indigenous technology development in the Chinese context. In Section III, we set up a conceptual framework for our empirical analysis and briey review previous studies in the eld of interest. We show the econometric models in Section IV , give a detailed description of the dataset and show some descriptive statistics in Section V , and present the results in Section VI. We then conclude the paper in Section VII.
II. The Chinese Context
The policy focus in the PRC has changed in some important respects over the last years. One key change following the peoples party congress in 2006 has been the increased focus on technology development (Chinese Ministry of Science and Technology 2006, Sj oholm and Lundin 2010a). The ambition was to make the PRC an innovation-driven economy, the background being increased uneasiness with the dependence on manufacturing of labor-intensive goods. The perception has been that an upgrading of production towards higher-value-added goods would be desirable. It is argued that continued assembly of imported inputs would not enable the country to continue its rapid development. In light of this, Chinese ofcials worry that some technology indicators other than aggregate R&D, which is increasing, seem to suggest a low level of technology sophistication. For example, it has been shown that most new Chinese rms are heavily concentrated in relatively simple manufacturing activities (Sj oholm and Lundin 2010a). But it is not clear that the Chinese concern is necessary. For instance, Brandt, Van Biesebroeck, and Zhang (2012) show that total factor productivity (TFP) growth has been very high in the PRC between 1998 and 2007, which suggests that there have been considerable technology improvements taking place. This focus on technology development has been repeated by various policymakers at frequent occasions. For instance, in August 2009, Science and Technology Minister Wan Gang argued that The most effective way to withstand the impact of the global economic meltdown is to accelerate technological innovation, the new economic growth engine.1 More recently, President Hu Jintao argued for the importance of technology in a more broad sense, We must focus on promoting innovation in science and technology if we want to push forward reform and opening up policy, the modernization of socialism, and achieve the overall target for building a moderately prosperous society in an all-round way, improve the peoples living standard, as well as achieve the great rejuvenation of the Chinese nation.2
Government Pledges Strong Support for Innovation-based SMEs, China Daily, 1 September 2009. Top Leaders Urge Innovation in Science, Technology, China.org.com, accessed 8 July 2012. Available at www.china.org.cn/china/2012-07/08/content_25847720.htm
2 1
The Chinese innovation strategy is also related to discussions on the middleincome trap. This refers to the seemingly difcult task for countries that wish to move from a middle-income position to become a high-income economy. Few countries have managed such a transition in recent decades as rising wage costs tend to erode competitiveness with slowing growth as a result (Grifth 2011). Excluding the small city states, only three Asian economies have managed the transition from middle-income to high-income positions: Japan in the 1960s and Taipei,China and the Republic of Korea in the 1990s. The middle-income trap is frequently discussed also in a Chinese context. The World Bank (2012, xvi) discusses in a report how continued high economic growth in the PRC can be secured and highlights the need to accelerate the pace of innovation and create an open innovation system. It is argued that a higher degree of technological sophistication could balance rising labor costs and enable the PRC to move to higher-value-added production. An important aspect of the PRCs technology policy is the emphasis on indigenous technology development. In other words, with general technology development deemed insufcient, indigenous Chinese rms in particular should develop the new products and production processes that would secure continued economic development. This is arguably an important change of strategy in a country as dependent on foreign multinationals as the PRC. The aspect of indigenous technology development was already stressed in the 2006 5-year plan and then again in the 12th 5-year plan in 2011. [The Peoples Republic of] China should upgrade its capabilities in indigenous research and innovation in science, technology and administration, train more innovative talents and improve education for workers. In a word, we will strive to speed up the construction of an innovation country (National Peoples Congress 2011, p. 3). Current policy promotes indigenous technology development through access to nance and subsidies and by preferential treatment to indigenous rms in public procurement. Another aspect of the policy has to do with increased restrictions on foreign multinationals in the PRC. Koyama and Golub (2006) nd that in a sample of 42 countries, the PRC and India tend to have the most restrictions on FDI. The Chinese investment regime has since become even more restrictive. For instance, a new investment regime in 2006 has made it more difcult to acquire Chinese companies (Sauvant 2009, p. 12). The Chinese government has also closed some industries to entry of foreign multinational rms. These sectors are typically the ones considered important from a national security perspective. In recent years, the economic security criteria has additionally been used to keep foreign rms out of the PRC.
Further restrictions were introduced in 2008. Foreign acquisitions can now be stopped by the authorities if considered a threat to local state-owned companiesan option that was used to stop Coca Colas $2.3 billion purchase of Huiyuan Juice in 2009. Other frequently used restrictions include requiring foreign rms to have local joint-venture partners, tap local suppliers, and locate R&D in the country. The new Chinese policies are not unproblematic. Public strategies for technology development are common in developing countries and it seems fair to say that the results are often disappointing. Subsidies and protection of indigenous high-tech rms have a tendency to result in a lack of international competitiveness. Moreover, labor-intensive production in foreign multinational rms are presumably a suitable development strategy for a country where a large share of the population still lives in poverty, income inequality is large and growing, and modern sector employment greatly needed (Karlsson et al. 2009, Sj oholm and Lundin 2010b).
III. FDI and R&D: Conceptual Framework and Previous Studies
As discussed above, Chinese authorities seem to be concerned by the level of indigenous technology development. This has resulted in a more hesitant approach towards FDI. However, there are reasons to believe that indigenous technology development might be affected by FDI. A suitable conceptual framework to discuss this angle can be based on two strands of literaturethe effect of FDI on the host countrys market structure and the effect of market structure on rms investment in R&D. The presence of foreign MNEs may exert a signicant inuence on the host countrys market structure. However, different theoretical models and previous empirical evidence show the relationship between FDI and market structure to be highly complex.3 FDI can both increase and decrease the degree of competition depending on the specic context. On the one hand, FDI may increase the number of rms in an industry and thereby decrease the concentration and increase the competition in the market, particularly in industries with high start-up costs and high barriers to entry (e.g., Barba Navaretti and Venables 2004, p. 174). This is true for greeneld investments but not for mergers and acquisitions, and the former type therefore has a more competitive effect on the local economy (Haller 2005). Moreover, the entrance of foreign MNEs might have a positive effect on production in existing domestic rms through spillovers and even increase the number of rms if employees in MNEs leave to set up their own businesses (Caves 1996). This would also tend to increase competition.
3
See, for example, UNCTAD (1997) and OECD (2002) for more detailed reviews.
On the other hand, FDI may raise the level of concentration in the hostcountry market (Aitken and Harrison 1999). Foreign MNEs possess competitive rm-specic assets and might therefore be able to capture a leading market position. The number of rms in an industry might then fall after the entry of foreign MNEs, if only the most efcient rms can survive and less efcient (domestic) rms are forced to exit. As a consequence, the industry will become more concentrated. It is important to note that in this case, high concentration is associated with initial high intensity of competition. Once rms are forced out of the market, competition will tend to decline. With regard to the impact of market structure on innovation, this issue has been addressed in a large body of theoretical work, which often yields conicting results (e.g., Aghion and Howitt 1992, Aghion et al. 2005). Innovations are, in the classic Schumpeterian view of creative destruction, made by rms that earn no rents if they fail to innovate and that obtain monopolistic power if they succeed. The market would be characterized by Arrows replacement effect, i.e., new rms replacing monopolists that fail to innovate (Arrow 1962). However, when competition intensies and in turn trims down monopoly rents, the incentive to innovate decreases. This theory therefore predicts a negative relationship between market competition and innovation. In contrast to the replacement effect, the selection effect of market competition predicts a positive relation between competition and innovation. Competition may stimulate innovation when rms with innovation advantages further strengthen their innovation capabilities in order to escape competition with neck-to-neck rivals (e.g., Vickers 1997, Boone 2000, Aghion and Schankerman 1999). In more recent theoretical work, the relationship between competition and innovation is described as non-monotone, which can happen when there are different types of innovators in terms of leaders and followers. Both the level of the technology gap and the degree of rivalry are important aspects that are taken into account in so-called step-by-step innovation models (e.g., Aghion et al. 2001, Boone 2001). The step-by-step innovation models are of particular relevance for analyzing unlevelled industries, i.e., industries where different rms have different levels of innovation capacity. In such industries, there are technically laggard rms which have to catch up with the leading-edge technology before they can compete with their more technologically advanced rivals. The ability to catch up partly depends on the level of competition. When competition is low, the leading rms will invest relatively little in R&D, which means that laggard rms have a higher potential for catching up and thereby a higher incentive to innovate. In the case of high market competition, the leading rms have a higher incentive to innovate to remain in their strong position. This makes it more difcult for followers to catch up and will, in turn, tend to decrease their incentives to innovate, and they will instead try to nd industrial niches with less competition from the leading rms.
As to the question of the impact competition, economic theory gives us little guidance in making predictions on how this would affect R&D. There are reasons to expect the effect to be positive as well as negative, and we have to address the issue by empirical analysis to obtain information on how the relationship works in a Chinese context.
A. Previous Studies on FDI and Competition
There are relatively few empirical studies on the competitive effects of FDI. One exception is Co (2001) who found both positive and negative effects of FDI on competition in the United States depending on market conditions and the extent of spillovers. Chung (2001) also nds mixed effects and suggests that the degree of competitive pressure imposed by FDI depends on both the entry mode and the various investment traits. The most interesting paper for our purposes is by Fu and Wu (2012) who study the effect of FDI on prots in Chinese rms. They nd an inverted U-shape effect of FDI on prots, where prots tend to increase with smaller inows of FDI but then decrease with larger ows.
B. Previous Studies on FDI and Domestic Innovation
Studies on the impact of FDI on domestic innovation and R&D are also rare. Veugelers and Vanden Houte (1990) study the Belgian manufacturing sector and nd that domestic rms tend to have lower innovation intensities the higher the share of FDI in the industry. It is not examined if this result is caused by an effect of FDI on the market structure. Two studies on the PRC by Jefferson et al. (2006) and Girma, Gong, and G org (2006) are related to our work. Jefferson et al. (2006) use a similar measure of investment in R&D, the rm-level R&D-to-sales ratio, and nd a negative relationship between rm size and R&D intensity and a positive but fragile effect of high industry concentration. Girma, Gong, and G org (2006) nd state-owned enterprises (SOEs) with foreign capital participation to have relatively high degree of innovation activity. Innovations in SOEs without foreign capital participation can be both positively and negatively affected by FDI depending on the absorptive capacity of these rms.
C. A Survey of Literature on FDI and Spillovers in the PRC
As discussed above, there are few studies on FDI, competition, and R&D in indigenous rms. This stands in large contrast to the extensive literature on spillovers from FDI. This literature examines how productivity in domestic rms can be affected by FDI and captures a broader set of productivity-increasing mechanisms than just the rms own R&D.
56 ASIAN DEVELOPMENT REVIEW Table 1. Papers on Spillovers from FDI in the PRC
Author(s) Wei and Liu (2006) Years Dependent Measure of Foreign Variable Presence Shares of output, employment, capital Share of output Share of sales Share of output Share of output Share of output Share of output Result Negative at a national level, positive at a province level Short-term negative effects, long-term positive effects Positive Positive Positive No effect Mixed
19982001 Output
Liu (2008) Abraham et al. (2010) Du, Harrison, and Jefferson (2011) Du, Harrison, and Jefferson (2012) Girma and Gong (2008) Xu and Sheng (2012)
19951999
TFP
20022004 TFP 19982007 Output 19982007 Output 19992002 20002003 TFP TFP
FDI = foreign direct investment, PRC = Peoples Republic of China, TFP = total factor productivity.
The presence of foreign multinational rms is believed to benet the host country through access to advanced technologies, some of which might be absorbed by domestic rms. Local rms might, for instance, imitate foreign rms techniques, either by reverse engineering or by hiring away some of their employees. These types of externalities or spillovers are present within industries. There might also be backward spillovers to suppliers and forward spillovers to customers. The former occurs if, for instance, foreign rms help suppliers to improve upon their products, the latter if domestic rms benet from better inputs. The issue of spillovers from FDI has been extensively debated and examined since the inuential article by Caves (1974). The increased availability of rm-level databases in recent years has led to new waves of studies. A few generalizations from the large literature can be made. First, some of the previously large spillover effects found in earlier papers using industry-level data were probably exaggerated. More recent papers on rm- and plant-level data tend to nd smaller effects (G org and Strobl 2001). Second, the results on within-industry spillovers are mixed, and there are more uniform ndings of positive spillovers to local suppliers and customers. Finally, the extent of spillovers differs between countries and industries and depends on the absorptive capacity, type of FDI, and degree of competition in the local economy (Lipsey and Sj oholm 2005). These generalizations seem to hold also for the PRC. Table 1 lists reviewed and published studies on spillovers from FDI in the PRC using rm-level data.4 Most studies nd some sort of spillover in the PRC, more often to local rms in linkage
4 There are also a large number of spillover studies on the PRC using industry-level data (e.g., Buckley, Clegg, and Wang 2002) and regional data (e.g., Madariaga and Poncet 2007). Most of these studies tend to nd positive spillovers from FDI.
industries rather than within industries. For instance, Du, Harrison, and Jefferson (2012) examine spillovers in a long panel of Chinese rms and nd little evidence of spillovers within the industries of investment but strong evidence for spillovers to both supplying industries and customer industries. Similar results are found in Liu (2008) who nds evidence of both intra-industry and interindustry spillovers but concludes that backward linkages seems to be the strongest ones. The paper by Xu and Sheng (2012) differs from the studies above and nds positive spillovers from regional presence of multinational rms within the industry but negative effects of regional presence in backward and forward industries. It seems plausible that geographic proximity could enhance intra-industry spillovers but it is more difcult to understand why the presence of foreign multinational rms should hurt local rms in the same region in neighboring industries. Accordingly, Wei and Liu (2006) nd evidence of positive intra-industry and interindustry spillovers at the regional level but negative ones at a national level. Hence, their results suggest that local rms benet from foreign rms in the same region but are hurt by foreign rms in other regions, irrespective of which industries the foreign rms are located in. The contrasting effects might be caused by a different balance between competition effects and spillover effects from FDI within and between regions. As found in studies on other countries, it seems that spillovers to local Chinese rms are not automatic but affected by various factors. For instance, absorptive capacity seems important. Girma and Gong (2008) examine spillovers to state-owned Chinese enterprises. The authors nd little evidence of spillovers and attribute this to a low level of absorptive capacity. Other studies focus on the role of geographic proximity in spillovers, an issue already touched upon above, and on how spillovers from different types of multinational rms differ. The latter issue is often analyzed in the context of spillovers from FDI from Hong Kong, China; Taipei,China; and Macau, China (often referred to as SouthSouth FDI) versus FDI from other countries (often referred to as NorthNorth FDI). Most studies nd spillovers to be higher from North FDI than from South FDI. For instance, Du, Harrison, and Jefferson (2012) argue that the lack of spillovers from FDI from Hong Kong, China; Taipei,China; and Macau, China in their study suggests that much of the ows may really be round-tripping capital rather than FDI. An additional nding is that FDI in rms beneting from tax incentives to investing rms generates greater productivity spillovers than unsubsidized rms (Du, Harrison, and Jefferson 2012, p. 28). Another paper on the PRC based on 4 years of census data and which uses the distinction between FDI from Hong Kong, China and Taipei,China and FDI from all other locations nds evidence of smaller spillovers from the SouthSouth FDI (Xu and Sheng 2012).
Abraham, Konings, and Slootmaekers (2010) also compare spillovers in the PRC from NorthSouth FDI with those from SouthSouth FDI. They dene SouthSouth FDI as FDI from Hong Kong, China; Macau, China; and Taipei,China (or HMT) plus FDI from tax havens, which they include following the suggestion of Naughton (2007, p. 164) who states that FDI from tax havens are generally diverted investment from HMT or [the Peoples Republic of] China itself for tax evasion. Spillovers from HMT and non-HMT FDI are both positive and statistically significant, but those from HMT are larger though not signicantly so. However, FDI from HMT is negatively related to the productivity of domestic exporters and rms located in special economic zones (SEZs). Girma and Gong (2008) nd distance to be a factor in spillovers, in this case, spillovers from FDI to SOEs. One result they describe as robust across all specications is that there is no evidence of productivity spillovers. . .outside the region FDI takes place (Girma and Gong 2008, p. 735736). SOEs also appear to lose from the presence of ethnic Chinese FDI (SouthSouth FDI) in downstream sectors in their regions. Wei and Liu (2006, p. 553), using similar industry data and denitions of sources of FDI in the PRC, conclude that FDI from Organisation for Economic Co-operation and Development (OECD) countries has played a much greater positive role in inter-industry productivity spillovers to indigenous Chinese rms than FDI from Hong Kong, China; Macau, China; and Taipei,China, adding that FDI from these two different sources has played a similar role in terms of magnitude in intra-industry productivity spillovers within regions. They suggest that the contributions of FDI from the two sources to Chinese rms productivity may be of a different nature (Wei and Liu 2006, p. 553). Specically, technologies transferred or diffused by HMT rms (SouthSouth FDI) may be more compatible with (the mainlands) current resource endowments. Authors also state that Foreign-invested rms from OECD countries have higher technological capabilities, and their productivity spillovers may concentrate on the enhancement of technological knowledge and competence in indigenous Chinese rms, and this is very important for (the PRCs) move to a higher development stage. A few conclusions can be made from the literature surveyed above. First, most studies nd evidence of some type of spillover from FDI in relation to the PRC. Second, the studies differ quite substantially on exactly what type of spillovers they nd and on rm and industry characteristics that affect the degree of spillover. There are results pointing to both interindustry and intra-industry spillovers in the PRC, but with more evidence of the former. Moreover, there seems to be a clear geographic component with FDI beneting local rms primarily within the same region. Absorptive capacity seems to be important and may explain why SOEs do not appear to benet from spillovers. Finally, the results from previous studies seem to indicate less spillovers from FDI coming from Hong Kong, China; Taipei,China; and Macau, China than from FDI coming from other countries.
IV. Econometric Models A. The Effect of FDI on Market Structure
We use a two-step econometric approach. First, we investigate the impact of FDI on the market structure. We follow a standard approach and use the price cost margin (PCM) as a measure of competition. This is dened as:5 PCM = Value added payroll Value added (1)
A high value on PCM means a large markup and presumably a low level of competition. Our measure of FDI penetration is calculated as the share of sales by foreign rms in total sales in the domestic market at the 4-digit industry level. Exports by foreign rms are excluded from the sales gures, since these do not impose competitive pressure on the Chinese market. The baseline econometric model is specied as: PCM jt = + FDIi ,t n + Firm jt + Hi ,t n + DOWNERw + 1 DTt + 2 DINDi + 3 D Re gion r + jt where PCM jt is the price cost margin of rm j at time t; FDIi ,t n indicates the presence of FDI in industry i at time t at the 4-digit industry level, where n is the number of lags; Firm jt represents a vector of rm-level control variables such as capital intensity, market share, export intensity, and relative TFP; Hi ,t n is the Herndahl index in industry i as a proxy for industrial concentration at the 4-digit industry level; and DOWNERw represents a vector of ownership dummy variables. DT, DIND, and DRegion denote year dummy variables, industry dummy variables at the 4-digit level, and region dummy variables at the 2-digit level (31 geographic units). In the above specication, FDI and the Herndahl index are two industrylevel variables that measure the effect of market structure on the PCM. The key hypothesis is that high concentration raises market power and, hence, increases PCM. Moreover, FDI may have a negative or positive effect on competition in the
5 The approach makes the restrictive assumption of constant returns to scale. Konings, Van Cayseele, and Warzynski (2005) apply an alternative methodology developed by Roeger (1995). The need to impose the constraint that markup (alternative expression of PCM) is the same for all rms within the same industry makes this methodology less suitable when rms are as heterogeneous as in the PRC.
(2)
same industry. We do not examine if FDI affect downstream and upstream local rms. In examining the effect of competition on the PCM, it is also important to control for efciency effects. A high PCM does not necessarily only reect low competition in the market, but may also be associated with higher efciency in the rm. To control for such efciency effects, we include a rms TFP relative to average TFP in the industry. Furthermore, a rms domestic market share and export intensity might also pick up the efciency aspect. Market share may also capture rm-specic market power. Following previous empirical studies by Scherer and Ross (1990) and Roberts and Tybout (1997), we specify a nonlinear relationship between domestic market share and PCM and add a quadratic term of market share. The effect of export share on PCM depends on the relative price elasticity of demand for the rms product in the home market and abroad.
B. The Effect of Competition and FDI on R&D Intensity
In the second step, we examine the effect of competition on R&D intensity at the rm level and estimate the following model: RDINT jt = + RDINT j ,t n + PCM j ,t n + FDIi ,t n + Firm jt + DOWNERw + 1 DTt + 2 DINDi + 3 D Re gion r + jt where RDINT jt is the ratio of R&D expenditures to sales of rm j at time t; and Firm jt denotes a vector of rm-level control variables such as the share of science and technology personnel in total employment, export share, and rm size. An important methodological issue when estimating this type of model is the treatment of persistence in rms R&D investment behavior. We include lagged R&D intensity to deal with this aspect, which means that we estimate a dynamic model. However, one econometric problem in estimating a dynamic model is that OLS estimates are likely to suffer from a dynamic panel bias. Therefore, we follow standard approaches and use system generalized method of moments (GMM) estimates developed by Arellano and Bover (1995) and Blundell and Bond (1998) which imply that RDINT and any other potentially endogenous variables are instrumented. The system rst uses differenced and level versions of the estimating equation, where lagged values in the former and lagged differences in the latter can serve as valid instruments. The differentiated transformed instruments are assumed to be uncorrelated with unobserved xed effects, implying that rst differentiated variables can act as instruments for variables in levels, i.e., instrumenting levels with differences. (3)
We use the Sargan/Hansen test to evaluate the instruments and the Arellano and Bond (1991) test for autocorrelation in the idiosyncratic disturbance term eit .
V. Data and Descriptive Statistics A. Data
Our data on large- and medium-sized manufacturing enterprises (LMEs) are compiled annually by the National Bureau of Statistics of China (NBS) and cover the period 19982004. All rms classied as LMEs are included making it a census rather than a survey.6 The classication of LMEs follows the NBSs classication from 2000. In this classication, employment, turnover, and xed capital are applied as a combined indicator of rm size (see Table A2 in the appendix).7 The sample of rms is unbalancedthere are almost 16,000 rms included in 1998 and more than 24,000 in 2004. These rms constitute roughly a third of total industrial sales and about a fourth of total industrial employment. One source of potential bias is that rms that shrink in size may exit our sample. In other words, R&D in weak rms (shrinking in size) might be differently affected by foreign entry. This calls for some caution in interpreting the results. The NBS collects information on a large number of rm characteristics such as sales, employment, labor cost, materials, xed assets, export, and ownership. The information on LMEs R&D activities includes R&D expenditures and the number of employees involved in science and technology. R&D gures are sometimes claimed to be exaggerated in the PRC, partly as a consequence of tax breaks for R&D, but also because of policies mandating annual R&D targets for rms. Unfortunately, we are not in a position to evaluate how serious this possible exaggeration could be and if it biases our results. What we can say from other studies is that R&D seems to be highly correlated with other R&D indicators such as sales of new products and patents (e.g., Sj oholm and Lundin 2010a), suggesting that it is a reasonable indicator of innovative activities at the rm level. The industry classication is similar to the classication ISIC, Rev. 3. The included sectors at the 2-digit industry level can be found in Table 3. In our econometric analysis, we construct industry-level variables such as industry concentration and FDI penetration at the 4-digit industry level in order to have industrial control variables that are as disaggregated as possible. Finally, the effect of FDI on competition and R&D might differ for rms with different kinds of ownership. Therefore, we divide our dataset into domestic
See Jefferson et al. (2003) for a detailed description of the same data. To compare market structure over time, we have reclassied the rms in 19981999 according to the new classication. This explains why the number of rms can differ from the ofcially published sources where the classication of rm size is not the same for the period prior to and after 2000.
7 6
62 ASIAN DEVELOPMENT REVIEW Table 2. Number of Foreign Firms and Foreign Share of Chinese Manufacturing, 19982004 (share of total manufacturing)
Year 1998 1999 2000 2001 2002 2003 2004 Number of Foreign Firms 3,489 3,764 4,221 4,585 5,327 6,512 8,745 Firms 0.22 0.23 0.25 0.27 0.29 0.31 0.36 Value-added 0.26 0.28 0.30 0.31 0.33 0.36 0.40 R&D Expenditure 0.21 0.23 0.20 0.23 0.23 0.25 0.29 Export 0.58 0.61 0.63 0.66 0.68 0.71 0.76 Employment 0.14 0.16 0.18 0.20 0.23 0.27 0.34
R&D = research and development. Source: Authors calculations on Chinese rm data.
and foreign subsamples according to the classication given by Table A3 in the appendix.
B. Descriptive Statistics
The share of foreign rms in our data on Chinese manufacturing between 1998 and 2004 is seen in Table 2.8 Foreign rms include wholly foreign-owned rms and joint ventures between foreign and domestic rms.9 Foreign ownership has increased substantially in relative as well as absolute terms. For instance, the number of foreign-owned rms increased by 150% over the periodfrom 3,489 in 1998 to 8,745 in 2004. This increase was higher than in domestically owned rms, with the foreign share of LMEs rising from about 22% to 36%. The other variables show a similar pattern of rapid increases. Foreign-owned rms account for about one-third of employment, 40% of value-added, and a staggering 76% of total exports. Turning our attention to the focus of this paper, R&D, it is interesting to note that foreign share is relatively small. Foreign share of R&D expenditures was 21% in 1998, which was larger than the foreign share of employment and about the same as the foreign share of LMEs. However, this increased to only about 29% in 2004, or lower than the foreign share of any other economic indicator in Table 2. One conclusion that can be drawn from the gures is that even if the PRC has become increasingly attractive as a location of foreign rms R&D, as suggested for instance by UNCTAD (2005), this trend has grown more slowly than the increase in foreign rms production, employment, and export.
8 The share of foreign rms in total Chinese manufacturing is likely to be lower since domestic rms dominate among the non-included small rms. 9 Hence, rms with any foreign ownership share are classied as foreign. Changing the threshold of foreign ownership that is required for a rm to be classied as foreign does not have any major impact on the econometric results.

Table 3. Foreign Firms, Price Cost Margins, and R&D Intensities across 2-digit Level Industries, 2004
Number of Firms FDI Domestic 13 Processing food from 252 agriculture 14 Production, processing 214 of food 15 Beverage 191 17 Textiles 776 18 Wearing apparels 483 19 Leather, footwear 376 20 Wood, timber, bamboo 79 products 21 Manufacture of 184 furniture 22 Pulp and paper 180 23 Publishing, print 110 24 Musical instruments, 257 sport goods 25 Rened petroleum 27 products 26 Basic chemicals 221 27 Pharmaceuticals, 158 medicinal chemistry 28 Manufacture of 58 chemical ber 29 Rubber products 173 30 Plastics products 454 31 Non-metallic mineral 322 products 32 Ferrous metals 114 33 Non-ferrous metals 98 34 Metal product 331 35 Machinery, general 383 36 Machinery, special 215 purpose 37 Transport equipment 437 39 Electrical machinery 795 and apparatus 40 Computer, 1445 communication 41 Ofce machinery, 221 measuring instrument 42 Manufacture n. e. c. 186
Source: Authors calculations.
Share of Foreign Firms (%) VA 38 43 39 28 45 67 33 82 43 46 69 9 21 23 31 46 55 23 11 16 50 34 24 46 43 86 82 45 R&D 8 29 43 32 15 50 11 82 38 37 54 7 15 22 18 23 28 26 4 25 21 24 13 32 23 49 42 07
PCM (%)
R&D Intensity (%)
FDI Domestic FDI Domestic 11.3 15.6 19.9 10.6 9.6 9.4 9.0 11.1 12.3 17.7 9.0 16.4 12.0 14.8 8.1 10.0 11.6 14.3 15.7 11.4 15.4 14.1 15.6 13.1 11.0 11.9 12.1 11.6 15.4 13.2 17.0 8.1 11.0 11.4 14.5 10.5 11.5 16.2 8.0 15.9 12.3 10.9 8.5 11.9 14.5 13.1 12.4 11.3 12.1 8.2 7.3 8.0 10.8 13.5 10.3 4.6 0.09 0.15 0.25 0.32 0.09 0.11 0.21 0.15 0.23 0.21 0.19 0.08 0.67 1.44 0.21 0.20 0.27 0.50 0.41 0.82 0.15 0.67 0.60 0.70 0.47 0.60 0.84 0.09 0.22 0.40 0.37 0.36 0.17 0.23 0.37 0.28 0.26 0.21 0.66 0.16 0.77 1.34 0.63 0.69 0.59 0.35 0.24 0.56 0.54 1.29 1.43 1.30 1.17 2.93 2.85 1.47
654 361 351 1673 353 158 123 80 432 185 80 340 1443 593 165 187 268 1482 813 460 475 1265 742 1231 965 488 169 154
We continue by looking at the foreign share of different sectors. Table 3 shows that the number of foreign rms is highest in electronic products (sectors 3941) and textiles, clothes, and shoes (1719). In relative terms, FDI is of great importance in many industries and of particularly high importance in footwear,
furniture, sporting goods, computers, and ofce machines, where more than twothirds of the value added come from foreign-owned rms. Once more, the foreign share of R&D expenditures tends to be lower than the shares of other economic indicators. For instance, the foreign share of R&D expenditures is 50% or above in only three sectors (leather and footwear, furniture, musical instruments and sporting goods) and the foreign share of R&D is lower than the foreign share of value added in all but four industries (beverage, textiles, non-metallic mineral products, and non-ferrous metals). The PCM is higher in foreign rms than in domestic rms in about twothirds of the industries. There seem to be important sector-specic effects in PCMs as foreign and domestic rms show a similar pattern across sectors. For instance, PCMs are particularly high in food, beverage, petroleum, metals, and machinery and low in clothing, footwear, wood products, and some chemical industries. For domestic rms, the PCM is also relatively high in publishing and plastics and low in various machinery sectors. More importantly, there is no obvious relation between the share of foreign rms in a sector and the PCMs in domestic rmsPCMs are relatively high in some sectors with low foreign presence such as ferrous metals and petroleum products, but also high in some sectors with high foreign presence such as plastics and computers. Finally, R&D intensities are relatively high in pharmaceuticals, machinery, transport equipment, electronics, computers, and ofce machinery, but there is a large difference between foreign and domestic rms. R&D intensity is higher in foreign rms than in domestic ones in only three sectors: non-metallic mineral products, ferrous metals, and non-ferrous metals. It is difcult to detect a direct relationship between PCM and R&D intensity. Domestic rms conduct relatively large shares of R&D in some sectors with high PCMs such as computers and in some sectors with low PCMs such as machinery. Accordingly, there is no obvious relation seen in Table 2 between FDI share and R&D intensity in domestic rms.
VI. Results A. The Effect of FDI on Price Cost Margins
We start by estimating the effect of FDI on the PCM using the complete sample of rms with results shown in Table 4. A number of different estimators are used: OLS with and without industry and regional dummies and a xed-effect estimator. It should be noted that the relatively short time period together with lagged dependent variables makes the xed-effect estimations relatively weak but including them gives us a sense of the robustness of the results.10 Moreover, there are obvious
10 We also tried to estimate random effect models but they did not pass the Hausman specication test and were therefore not included.

Table 4. Determinants of Price Cost Margins at the Firm Level, 19982004 (full sample, domestic, and foreign rms)
Variables OLS (1) OLS (2) 0.001 (0.009) 0.031 (0.009) 2.585 (1.486) FE (3) 0.002 (0.008) 0.023 (0.009) 2.239 (1.352) OLS (4) 0.000 (0.022) 0.037 (0.018) 3.068 (3.099) 0.003 (0.070) 0.055 (0.041) 0.192 (0.064) 0.002 (0.0008) 2.947 (0.539) 0.197 (0.013) 0.022 (0.005) OLS (5) 0.002 (0.011) 0.038 (0.010) 1.727 (2.035) 0.009 (0.037) 0.035 (0.022) 0.276 (0.050) 0.003 (0.0008) 3.243 (0.234) 0.208 (0.011) 0.017 (0.005) 1.051 (0.359) 0.880 (0.330) 1.689 (0.418) Yes Yes Yes 0.33 23,123 FE (6) 0.001 (0.010) 0.032 (0.010) 1.622 (1.782) 0.009 (0.032) 0.048 (0.027) 0.110 (0.038) 0.001 (0.0005) 3.189 (0.195) 0.220 (0.004) 0.002 (0.007) FDI penetration 0.002 (with 1 lag) (0.016) FDI penetration 0.027 (with 2 lags) (0.015) Herndahl index 4.365 (with 1 lag) (1.562) FDI penetration Herndahl index (with 1 lag) FDI penetration Herndahl index (with 2 lags) Market share 0.192 (0.067) Market share 0.002 market share (0.0008) Capitallabor ratio 2.946 (0.544) Relative TFP 0.196 (0.013) Export intensity 0.022 (0.005) Ownership dummy (Private) Ownership dummy (JVHTM) Ownership dummy (JV foreign and Foreign) Year dummies Yes Industry dummies No Regional dummies No R2 0.20 23,123
0.275 (0.047) 0.003 (0.0008) 3.244 (0.234) 0.208 (0.015) 0.017 (0.005) 1.050 (0.359) 0.880 (0.331) 1.691 (0.418) Yes Yes Yes 0.33 23,123
0.111 (0.038) 0.001 (0.0005) 3.187 (0.195) 0.220 (0.004) 0.002 (0.007)
Yes Within: 0.23 Between: 0.19 Overall 0.20 23,123
Yes No No 0.20 23,123
Yes Within: 0.23 Between: 0.19 Overall 0.20 23,123
Observations
= signicant at 1% level, = signicant at 5% level, = signicant at 10% level, HTM = Hong Kong, China; Taipei,China; Macau, China, JV = joint venture. Note: Standard errors in parentheses are adjusted both for heteroskedasticity and potential dependency among rms in the same industry at the 4-digit level. Source: Authors computations.
instances of omitted variables and simultaneity biases. We try to control for these aspects but the results should be interpreted with caution and be seen as correlations rather than causations. Our rst estimation shows that FDI has a negative impact on the PCM, but there is a time lag before the competition from FDI exhibits an effect (Table 4). Lag 1
of FDI is statistically insignicant but lag 2 is signicant. Excluding lag 1 of FDI did not change the results for lag 2, while including lag 2 of the Herndahl index rather than lag 1 did not change its signicance (not shown). Moreover, the PCM is high in concentrated markets. The results seem stable across different estimations (columns 2 and 3). Turning to the other variables, results show that capital-intensive rms with high levels of TFP have high PCMs and that, surprisingly, rms with large market shares and export intensities have low PCMs.11 Signicant and negative coefcients of ownership dummy variables in columns (2) and (4) comprise another interesting outcome. This suggests that, compared to the reference group of SOEs and collective rms, both domestic private rms and foreign-owned rms have low PCMs. This can be due to both market-related and institutional effects. As an example, SOEs often obtain subsidies in terms of, for instance, access to capital at below market interest rates, which might explain the relatively high PCMs. The above estimations on the effect of FDI on PCMs are rather conservative considering that we control for market shares and industry concentration. These variables are likely to be endogenous to FDI and we therefore run the risk of underestimating the impact of FDI on competition. Excluding the market share and Herndahl index increased the coefcient on FDI marginally (not shown). Accordingly, experimenting with different subsets of control variables had no major impact on the results. To further examine the robustness of the results, and whether the effect of FDI on the PCM depends on the degree of market concentration, we insert an interaction term of FDI and the Herndahl index. As shown in columns 46, the negative effect of FDI on PCMs remains robust and a signicant interaction effect can only be observed in the xed-effect estimation where the competitive effect imposed by FDI seems to be weaker in industries with high concentration. Our main interest is to examine how FDI affects the competition for domestically owned rms. Therefore, we repeat the estimations above but exclude foreign-owned rms and joint ventures. The results are shown in Table 5.12 The estimations yield fairly similar results, but with two differences if compared to the full sample estimations in Table 4. First, the negative effect of export intensities on the PCM disappears. Second, domestic rms in concentrated industries do not have comparably high PCMs. The results are again robust to the inclusion of interaction variables as seen in columns 46 and to changes in the lag structure (not shown).
11 We also used labor productivity as an alternative productivity measure. Firms with high labor productivity were found to have high PCM, while the results for the other included variables remained unchanged (not shown). 12 The estimations include SOEs, collective rms, and private rms. Shareholding rms and other rms are not clearly dened (i.e., can include foreign ownership) and have therefore been excluded. As an additional robustness check, they were included in the domestic subsample but this had little impact on the results (not shown).

Table 5. Determinants of Price Cost Margins at the Firm Level, 19982004 (domestic rms)
Variables OLS (1) OLS (2) 0.017 (0.010) 0.033 (0.014) 1.138 (2.846) FE (3) 0.019 (0.015) 0.013 (0.016) 3.288 (2.487) OLS (4) 0.004 (0.042) 0.073 (0.036) 2.752 (4.431) 0.023 (0.167) 0.309 (0.200) OLS (5) 0.019 (0.020) 0.042 (0.022) 0.248 (3.431) 0.021 (0.086) 0.078 (0.139) FE (6) 0.024 (0.020) 0.038 (0.021) 2.481 (2.693) 0.020 (0.080) 0.176 (0.094)
FDI penetration 0.010 (with 1 lag) (0.030) FDI penetration 0.037 (with 2 lags) (0.024) Herndahl index 6.386 (with 1 lag) (2.760) FDI penetration Herndahl index (with 1 lag) FDI penetration Herndahl index (with 2 lags) Market share 0.202 (0.091) Market share 0.002 Market share (0.0010) Capitallabor ratio 3.805 (1.186) Relative TFP 0.179 (0.021) Export intensity 0.014 (0.010) Ownership dummy (Private) Year dummies Yes Industry dummies No Regional dummies No 0.16 R2 Observations
12,891
0.248 0.157 0.220 (0.058) (0.055) (0.099) 0.003 0.002 0.003 (0.0008) (0.0007) (0.001) 3.189 3.281 3.803 (0.439) (0.287) (1.170) 0.182 0.196 0.179 (0.014) (0.005) (0.021) 0.000 0.015 0.014 (0.011) (0.015) (0.010) 0.421 (0.372) Yes Yes Yes Yes No Yes No 0.35 Within: 0.17 0.16 Between: 0.14 Overall 0.15 12,891 12,891 12,891
0.250 0.158 (0.059) (0.055) 0.003 0.002 (0.0008) (0.0007) 3.190 3.274 (0.439) (0.288) 0.182 0.196 (0.014) (0.005) 0.000 0.015 (0.010) (0.015) 0.425 (0.373) Yes Yes Yes Yes 0.35 Within: 0.17 Between: 0.14 Overall 0.16 12,891 12,891
= signicant at 1% level, = signicant at 5% level, = signicant at 10% level. Note: Standard errors in parentheses are adjusted both for heteroskedasticity and potential dependency among rms in the same industry at the 4-digit level. Source: Authors computations.
The proper denition of a market in a large country like the PRC can be discussed. Our denition assumes the competitive effect of FDI to be the same throughout the country. This might be questionable hence we also constructed FDI penetration variables as the foreign share of a regionindustry (region being dened as east, west, or central). The overall results remained unchanged, but the FDI coefcient was slightly larger suggesting the competitive effect of FDI to be strongest within the same region. We also experimented with alternative estimation methods. For instance, we tried to specify a dynamic model with lagged PCMs as the independent variable.
68 ASIAN DEVELOPMENT REVIEW Table 6. Determinants of R&D Intensity at the Firm Level, 19982004 (GMM estimations)
Variables R&D intensity (with 1 lag) PCM (with 1 lag) PCM PCM (with 1 lag) FDI penetration (with 1 lag) Skill share Export intensity Firm size Ownership dummy (Private) Ownership dummy (JV-HTM) Ownership dummy (JV foreign & Foreign) Year dummies Industry dummies Regional dummies R2 AR (2) Hansen test Observations
Note:
All Firms 0.273 (0.034) 0.003 (0.004) 0.0000 (0.0001) 0.001 (0.001) 0.069 (0.011) 0.0000 (0.001) 0.156 (0.054) 0.149 (0.038) 0.103 (0.061) 0.053 (0.062) Yes No No 0.212 0.475 39,687
Domestic Firms 0.289 (0.039) 0.003 (0.005) 0.000 (0.000) 0.003 (0.003) 0.069 (0.013) 0.002 (0.002) 0.177 (0.067) 0.124 (0.051)
FDI Firms 0.204 (0.038) 0.003 (0.004) 0.000 (0.000) 0.001 (0.001) 0.070 (0.008) 0.001 (0.001) 0.118 (0.061) 0.057 (0.029) Yes No No 0.144 0.835 16,745
High-tech Firms 0.387 (0.064) 0.027 (0.012) 0.0004 (0.0002) 0.001 (0.004) 0.090 (0.019) 0.002 (0.003) 0.174 (0.135) 0.358 (0.215) 0.121 (0.263) 0.023 (0.277) Yes No No 0.261 0.719 4,986
Non high-tech Firms 0.225 (0.033) 0.004 (0.003) 0.000 (0.000) 0.002 (0.001) 0.062 (0.012) 0.001 (0.001) 0.114 (0.054) 0.124 (0.033) 0.152 (0.058) 0.125 (0.057) Yes No No 0.282 0.668 34,701
Yes No No 0.971 0.428 22,942
= signicant at 1% level, = signicant at 5% level, = signicant at 10% level, HTM = Hong Kong, China; Taipei,China; Macau, China, JV = joint venture. Source: Authors computations.
However, although the estimations conrmed previous results on PCMs and R&D, the models failed to pass the Sargan/Hansen specication tests and are therefore not shown. To sum up the results so far, it has been shown that FDI imposes signicant competitive pressure on Chinese rms. The result is robust to different estimators and various industry- and rm-level controls as well as to different subsamples. We now continue to examine if this increased competition has an effect on investments in R&D.
B. The Effect of Price Cost Margins on R&D Intensities
Results from the GMM estimations are shown in Table 6. Most results are stable across different models and samples.
The results show no strong signs of an effect of competition on R&D. The PCM coefcient is statistically insignicant in the different samples: all rms, only domestic rms, only foreign rms, and non-high-tech rms. The one possible exception to a non-signicant effect of competition can be seen in high-tech industries where high competition (low PCM) has a negative effect on R&D.13 We also included an FDI variable, which is expected to capture the effect of FDI on R&D after controlling for the indirect effect on competition. Such an effect could be through demonstration effects or technology spillovers. The results consistently show that no such effect seems to exist in Chinese manufacturing. The coefcients of lagged R&D intensity are positive and highly signicant, indicating persistence in R&D and justifying inclusion of the lag.14 We also observe a signicant and positive effect of skills share but a signicant and negative effect of rm size on R&D intensity.15 These results are robust across the different specications. Finally, private domestic and foreign rms seem to have lower R&D intensities than SOEs after controlling for various rm characteristics. We tried to estimate specications with the same control variables as in the PCM estimations. Excluding skills share and replacing rm size with market share had no impact on the results (not shown). Moreover, estimations including TFP unfortunately did not pass the specications tests. As an alternative robustness check, we estimated a xed-effect model but the results did not change (not shown). Moreover, a relatively large proportion of the rms do not engage in R&D at all. It might be that such rms are located in small segments of industries where they are not signicantly affected by foreign rms or the industry level of competition. We examined this issue by including only those rms that have positive R&D spending in at least one year of their existence. Once more, the results remained unchanged (not shown). Hence, we conclude by noting that the effect of rms PCMs on R&D intensities seems to be insignicant with the result appearing robust across samples and estimation methods.
VII. Concluding Remarks
FDI can be an important channel for developing countries to gain access to new technology. The impact of FDI on the technology development of domestically owned rms is less examined, but it is frequently argued that technology externalities or a demonstration effect could have a positive impact. Another little examined effect
13 We follow OECD (2006) and dene high-tech as including pharmaceuticals; aircraft and spacecraft; radio, television, and communication equipment; ofce, accounting, and computing machinery; and medical, precision, and optical instruments. 14 We also estimated the model by including a 2-year lag of R&D intensity. The coefcients on the second lag turned out to be insignicant, and those on the rst lag remained signicant (not shown). 15 Jefferson et al. (2006) nd a similar effect of rm size on R&D intensity.
of FDI on technology development in domestically owned rms works through the impact on competition. FDI might affect the degree of competition in the market which in turn may affect efforts to upgrade technology in domestic rms. We have surveyed the literature on FDI and spillovers in the PRC. It seems that the spillovers are relatively large for suppliers and customers, for rms within the same region as the foreign rms, and from FDI coming from OECD countries. The studies on spillovers focus on the effect on productivity (labor or TFP) in local rms. It can be argued that this is only an indirect measure on technology and not the primary concern of the Chinese government. We therefore center our analysis on the impact of FDI on R&D in Chinese rms. Our hypothesis was that FDI could positively or negatively affect R&D through an impact on the degree of competition. We nd a strong and robust negative effect of FDI on PCMs of rms which suggests that FDI does increase the level of competition in Chinese manufacturing. This result stands in some contrast to studies on countries other than the PRC, which typically nd mixed effects of FDI on competition (e.g., Co 2001, Chung 2001, Fu and Wu 2012). We continue the analysis by examining determinants of R&D with a special focus on the role of competition. The general conclusion is that we nd a high degree of persistence in R&D and little evidence of any effect of competition, negative or positive, on the level of R&D. The result is slightly different from Jefferson et al. (2006) who nd a positive (albeit fragile) effect of industry concentration on R&D. Moreover, there is no indication of a direct effect of FDI on R&D in domestic rms, which contrasts with the nding on FDI in Belgium by Veugelers and Vanden Houte (1990) who estimate a negative effect. Finally, rms with high R&D intensities tend to have a relatively skilled labor force and to be relatively small in size. Moreover, SOEs tend to be more R&D intensive than domestic and foreign private rms. It is clear from our analysis that many issues need more research. For instance, the available data does not permit us to examine how small Chinese rms are affected by FDI, or how rms in upstream and downstream industries are affected. Accordingly, the effect of FDI on alternative measures of technology development such as patents or changes in the product mix awaits further research. It is also possible that FDI might have an effect on the impact of R&D in local rms even if the amount of R&D is not affected.16 Bearing these caveats in mind, we believe our results contribute to the ongoing policy debate in the PRC. Most importantly, there is substantial evidence of a positive effect of FDI on productivity in local rms. Some of these productivity gains might be caused by technology spillovers but there could also be other explanations. We do not nd any positive impact of FDI on R&D in domestically owned rms. Hence, it seems that although FDI has contributed to productivity, it has not been an important force for promoting R&D investment in domestic rms.
16
We are grateful to an anonymous referee who highlighted these areas as suitable for future research.
References
Abraham, Filip, Jozef Konings, and Veerle Slootmaekers. 2010. FDI Spillovers in the Chinese Manufacturing Sector: Evidence of Firm Heterogeneity. Economics of Transition 18(1):143182. Aghion, Philippe, Nicholas Bloom, Richard Blundell, Rachel Grifth, and Peter Howitt. 2002. Competition and Innovation: An Inverted-U Relationship. NBER Working Paper No. 9269. Cambridge, Mass.: National Bureau of Economic Research. . 2005. Competition and Innovation: An Inverted-U Relationship. Quarterly Journal of Economics 120(2):701728. Aghion, Philippe, Christopher Harris, Peter Howitt, and John Vickers. 2001. Competition, Imitation and Growth with Step-by-step Innovation. Review of Economic Studies 68(3):467492. Aghion, Philippe, and Peter Howitt. 1992. A Model of Growth through Creative Destruction. Econometrica 60(2):323351. Aghion, Philippe, and Mark Schankerman. 1999. Competition, Entry and the Social Returns to Infrastructure in Transition Economies. Economics of Transition 7(1):79104. Aitken, Brian, and Ann Harrison. 1999. Do Domestic Firms Benet from Foreign Investment? Evidence from Venezuela. American Economic Review 89(3):605618. Arellano, Manuel, and Stephen Bond. 1991. Some Tests of Specication for Panel Data: Monte Carlo Evidence and an Application to Employment Equations. Review of Economic Studies 58(2): 27797. Arellano, Manuel, and Olympia Bover. 1995. Another Look at the Instrumental Variable Estimation of Error-components Models. Journal of Econometrics 68(1):2951. Arrow, Kenneth. 1962. Economic Welfare and the Allocation of Resources for Inventions. In Richard Nelson, ed. The Rate and Direction of Technological Change. Princeton: Princeton University Press. Barba Navaretti, Giorgio, and Anthony Venables. 2004. Multinational Firms in the World Economy. Princeton and Oxford: Princeton University Press. Blundell, Richard, and Stephen Bond. 1998. Initial Conditions and Moment Restrictions in Dynamic Panel Data Models. Journal of Econometrics 87(1):11543. Boone, Jan. 2000. Competitive Pressure: The Effects on Investments in Product and Process Innovation. RAND Journal of Economics 31(3):549569. Boone, Jan. 2001. Intensity of Competition and the Incentive to Innovate. International Journal of Industrial Organization 19(5):705726. Brandt, Loren, Johannes Van Biesebroeck, and Yifan Zhang. 2012. Creative Accounting or Creative Destruction? Firm-Level Productivity Growth in Chinese Manufacturing. Journal of Development Economics 97(2):339351. Buckley, Peter, Jeremy Clegg, and Chengqi Wang. 2002. The Impact of Inward FDI on the Performance of Chinese Manufacturing Firms. Journal of International Business Studies 33(4):63755. Caves, Richard. 1974. Multinational Firms, Competition, and Productivity in Host Country Markets. Economica 41(162):171193. . 1996. Multinational Enterprise and Economic Analysis. Cambridge: Cambridge University Press. Chinese Ministry of Science and Technology (MOST). 2006. National Guidelines for Mediumand Long-term Plans for Science and Technology Development (20062020) of China. Available at www.most.org.cn/eng/newsletters/2006/t20060213_28707.htm
72 ASIAN DEVELOPMENT REVIEW Chung, Wilbur. 2001. Mode, Size, and Location of Foreign Direct Investments and Industry Mark-ups. Journal of Economic Behavior & Organization 45(2):185211. Co, Catherine. 2001. Trade, Foreign Direct Investment and Industry Performance. International Journal of Industry Organization 19(12): 163183. Du, Luosha, Ann Harrison, and Gary Jefferson. 2011. Do Institutions Matter for Spillovers? The Implications of Chinas Special Characteristics. NBER Working Paper 16767. Cambridge, MA: National Bureau of Economic Research. . 2012. Testing for Horizontal and Vertical Foreign Investment Spillovers in China, 19982007. Journal of Asian Economics 23(3):234243. Foster, Lucia, John C. Haltiwanger, and C. J. Krizan. 2001. Aggregate Productivity Growth: Lessons from Microeconomic Evidence. In Charles R. Hulten, Edwin R. Dean, and Michael J. Harper, eds. New Developments in Productivity Analysis. Cambridge: National Bureau of Economic Research. Fu, Dahai, and Yanrui Wu. 2012. Foreign Entry and Protability of Domestic Firms: Evidence from China. Paper presented at Asian Economic Panel, 56 October, Keio University. Girma, Sourafel, and Yundan Gong. 2008. FDI, Linkages and the Efciency of State Owned Enterprises in China. Journal of Development Studies 44(5):728749. Girma, Sourafel, Yundan Gong, and Holger G org. 2006. Can You Teach Old Dragons New Tricks? FDI and Innovation Activity in Chinese State-owned Enterprises. CEPR Discussion Paper No. 5838. London: Centre for Economic Policy Research. G org, Holger, and David Greenaway. 2004. Much Ado about Nothing? Do Domestic Firms Really Benet from Foreign Direct Investment? World Bank Research Observer 19(2):171197. G org, Holger, and Eric Strobl. 2001. Multinational Companies and Productivity Spillovers: A Meta-analysis. Economic Journal 111(475):723739. Grifth, Breda. 2011. Middle Income Trap. In Raj Nallari, Shahid Yusuf, Breda Grifth, and Rwitwika Bhattacharya, eds. Frontiers on Development Policy: A Primer on Emerging Policy Issues. Washington, DC: World Bank Publications. Haller, Stefanie. 2005. The Impact of Multinational Entry on Domestic Market Structure and R&D. EUI Working Paper No. 16. Florence: European University Institute. . 2009. More Intra-region Direct Investment through Economic Integration? Modelling the Experience of the EUs Single Market. Journal of Economic Integration 25(3):427456. Jefferson, Gary, Albert Hu, Xiaojing Guan, and Xiaoyun Yu. 2003. Ownership, Performance, and Innovation in Chinas Large and Medium-size Enterprise Sector. China Economic Review 14(1):89113. Jefferson, Gary, Bai Huamao, Guan Xiaojing, Yu Xiaoyun. 2006. R and D Performance in Chinese Industry. Economics of Innovation and New Technology 15:345366. Karlsson, Sune, Nannan Lundin, Fredrik Sj oholm, and He Ping. 2009. Foreign Firms and Chinese Employment. The World Economy 32(1):178201. Konings, Jozef, Patrick Van Cayseele, and Frederic Warzynski. 2005. The Effects of Privatization and Competitive Pressure on Firms PriceCost Margins: Micro Evidence from Emerging Economies. Review of Economics and Statistics 87(1):124134. Koyama, Takeshi, and Stephen Golub. 2006. OECDs FDI Regulatory Restrictiveness Index: Revision and Extension to More Economies. Working Papers on International Investment no. 4. Paris: Organisation for Economic Co-operation and Development. Lipsey, Robert. 2004. Home- and Host-country Effects of Foreign Direct Investment. In Robert Baldwin, and Alan Winters, eds. Challenges to Globalization. Chicago: University of Chicago Press.

Lipsey, Robert, and Fredrik Sj oholm. 2005. Host Country Impacts of Inward FDI: Why Such Different Answers? In Theodore Moran, Edward Graham, and Magnus Blomstr om, eds., The Impact of Foreign Direct Investment on Development: New Measurements, New Outcomes, New Policy Approaches. Washington, DC: Institute for International Economics. Liu, Zhiqiang. 2008. Foreign Direct Investment and Technology Spillovers: Theory and Evidence. Journal of Development Economics 85(12):176193. Madariaga, Nicole, and Sandra Poncet. 2007. FDI in Chinese Cities: Spillovers and Impact on Growth. The World Economy 30(5):837862. National Peoples Congress. 2011. Chinas Twelfth Five-year Plan (20112015). Beijing: NPC. Available: http://cbi.typepad.com/china_direct/2011/05/chinas-twelfth-ve-new-plan -the-full-english-version.html Naughton, Barry. 2007. The Chinese Economy, Transitions and Growth. Cambridge, Mass.: MIT Press. Organisation for Economic Co-operation and Development (OECD). 2002. Foreign Direct Investment for Development: Maximising Benets, Minimising Costs. Paris: OECD. . 2006. OECD Science, Technology and Industry Outlook. Paris: OECD. Roberts, Mark, and James Tybout. 1997. The Decision to Export in Colombia: An Empirical Model of Entry with Sunk Costs. American Economic Review 87(4):545564. Roeger, Werner. 1995. Can Imperfect Competition Explain the Difference between Primal and Dual Productivity Measures? Estimate for US Manufacturing. Journal of Political Economy 103(2):316330. Sauvant, Karl. 2009. FDI Protectionism is on the Rise. World Bank Policy Research Working Paper No. 5052. Washington, DC: World Bank. Scherer, Frederic, and David Ross. 1990. Industrial Market Structure and Economic Performance. Boston: Houghton Mifin Company. Sj oholm, Fredrik, and Nannan Lundin. 2010a. The Role of Small Firms in the Technology Development of China. The World Economy 33(9):11171139. . 2010b. Will Science and Technology Solve Chinas Unemployment Problem? Asian Economic Papers 9(2):128. United Nations Conference on Trade and Development (UNCTAD). 1997. World Investment Report 1997: Transnational Corporations, Market Structure and Competition Policy. Geneva: UNCTAD. . 2005. World Investment Report 2005: Transnational Corporations and the Internationalisation of R&D. Geneva: UNCTAD. Veugelers, Reinhilde, and Peter Vanden Houte. 1990. Domestic R&D in the Presence of Multinational Enterprises. International Journal of Industrial Organization 8(1):115. Vickers, John. 1997. Regulation, Competition, and the Structure of Prices. Oxford Review of Economic Policy 13(1):1526. Wei, Yingqi, and Xiaming Liu. 2006. Productivity Spillovers from R&D, Exports, and FDI in Chinas Manufacturing Sector. Journal of International Business Studies 37(4):544557. World Bank. 2012. China 2030: Building a Modern, Harmonious, and Creative High-income Society. Washington, DC: World Bank. Xu, Xinpeng, and Yu Sheng. 2012. Are FDI Spillovers Regional? Firm-level Evidence from China. Journal of Asian Economics 23(3):244258.
Appendix I. Total Factor Productivity Calculation:
ln TFP ji = ln Y ji K i ln K ji l ln L jt Mi ln M ji where Y is real gross output, K is real capital, L represents number of employees, and M real material use. l is the share of wages in output and Mi is the share of material inputs in output. K i is calculated as one minus the shares for labor and material inputs. We deate output, capital, and materials by the appropriate 4-digit industry price deator. Following Foster, Haltiwanger, and Krizan (2001), we calculate the factor shares at the 4-digit industry level to minimize the effects of measurement errors. We use annual weights in our calculations.
Table A1. Variable List and Denitions
Variable PCM Market share Capital intensity Relative TFP Export intensity R&D intensity Skill share Firm size FDI penetration Herndahl index (H) Denition Firm-level Variables (Value added payroll)/Value added Sales by rm i/Total domestic sales of industry j at the 4-digit industry level Log(Capital stock/Total number of employees) TFP for rm i /Average TFP in industry j at the 4-digit industry level Export/Total sales R&D expenditure/Total sales Number of S&T personnel/Total number of employees Log (Real total sales) Industry-level Variables Sales by foreign rms/Total domestic sales at the 4-digit industry level Sum of squared rm-level (domestic) market shares at the 4-digit industry level
Table A2. Classication of Large, Medium, and Small Enterprises

Large (1) Employment (person) Turnover (yuan, million) Fixed assets (yuan, million)
Source: National Bureau of Statistics of China.
Medium (2) 3002000 30300 40400
Small (3) 300 30 40
2000+ 300+ 400+

Table A3. Ownership Classications
Code 110 141 151 120 130 142 171 172 173 174 Ownership Domestic Ownership: SOE State-owned enterprises Stated-owned, jointly operated enterprises Wholly stated-owned enterprises Domestic Ownership: Collective Collective-owned enterprises Shareholding cooperatives Collective-owned, jointly operated enterprises Domestic Ownership: Private Private wholly owned enterprises Private-cooperative enterprises Private limited liability enterprises Private shareholding enterprises
Foreign Ownership: Hong Kong, China; Taipei,China; and Macau, China Invested 210 Overseas joint ventures 220 Overseas cooperatives 230 Overseas wholly owned enterprises 240 Overseas shareholding limited companies 310 320 340 330 Foreign Ownership: Foreign-invested Joint Ventures Foreign joint ventures Foreign cooperatives Foreign shareholding limited companies Foreign Ownership: Foreign Invested Foreign wholly owned enterprises
Source: National Bureau of Statistics of China.
Dynamics of Household Assets and Income Shocks in the Long-run Process of Economic Development: The Case of Rural Pakistan
TAKASHI KUROSAKI
This paper analyzes the dynamics of assets held by low-income households facing various types of income shocks in pre-independence and post-independence Pakistan. Focusing on the province of Khyber Pakhtunkhwa (formerly known as the NorthWest Frontier Province or NWFP), the paper rst investigates long-run data at the district level beginning 1902. Results show that the population of livestock, the major asset of rural households, experienced a persistent decline after crop shocks due to droughts, but did not respond much to the Great Depression. In the post-independence period, crop agriculture continued to be vulnerable to natural disasters, although less substantially so, while the response of livestock to such shocks was indiscernible from district-level data. To examine microeconomic mechanisms underlying such asset dynamics, I analyze a panel dataset collected from approximately 300 households in three villages in the NWFP during the late 1990s. Results show that the dynamics of household landholding and livestock are associated with a single long-run equilibrium. When human capital is included, the dynamics curve changes its shape but this is not sufciently nonlinear to produce statistically signicant multiple equilibriums. The size of livestock holding was reduced in all villages hit by macroeconomic stagnation, while land depletion was reported only in a village with inferior access to markets. The patterns of asset dynamics established from historical and contemporary analyses are consistent with limited but improving access to consumption smoothing measures in the study region over the century. Keywords: asset dynamics, natural disaster, buffer stock, poverty trap, Pakistan JEL codes: O13, O44, N55
I. Introduction
Natural and man-made disasters, such as oods, droughts, earthquakes, depressions, hyperinations, epidemics, etc., have affected the local and household
Takashi Kurosaki is Professor, Institute of Economic Research, Hitotsubashi University. The author would like to thank two anonymous reviewers of this journal and participants at the Asian Development Review Conference on Development Issues in Asia held in Manila on 7 November 2012, particularly Natalie Chun, Charles Horioka, and Masahiro Kawai, for their productive comments on the earlier versions of this paper. The author is also grateful to S. Hirashima, Yukinobu Kitamura, Motoi Kusadokoro, Ken Miura, Tetsuji Okazaki, Hiroshi Sato, Masahiro Shoji, Yoshifumi Usami, and Koji Yamazaki for their useful comments on related papers. All remaining errors are the authors. This paper was funded by a JSPS Grant-in-Aid for Scientic Research-S (22223003).
DYNAMICS OF HOUSEHOLD ASSETS AND INCOME SHOCKS 77
economy worldwide and throughout modern history. Households in contemporary, low-income developing countries are particularly vulnerable for several reasons. First, their initial welfare levels are already close to the poverty line. Second, institutional arrangements to cope with disasters are lacking. Third, early warning systems are absent. Similar reasons are applicable to households in developed countries before these countries experienced modern economic growth. This is because of the presence of numerous symptoms associated with absolute poverty in such economies. To compound issues, according to the emergency events database (EM-DAT), there appears to be an increase in the number of natural disasters globallyfrom fewer than 100 per year in the mid-1970s to approximately 400 per year during the 2000s.1 Thus, it is of critical importance to understand how households are affected by such disasters, how they recover from them, and how policies and market environments affect the dynamic process of recovery in the context of long-run economic development. Under incomplete markets, particularly with underdeveloped credit markets and missing insurance markets, poor households need to save as a precaution against downturn risk such as natural and man-made disasters. As a result, the asset choices of poor households may be excessively sensitive to risk avoidance, causing them to miss the opportunity to enhance expected income. In development economics, numerous theoretical and empirical studies focus on households ability to cope with these shocks (Fafchamps 2003, Dercon 2005). Furthermore, if the asset dynamics are highly nonlinear and hence associated with low and high long-run equilibriums, farmers may reduce consumption substantially after a disaster to preserve asset levels and avoid a low equilibrium (Carter and Barrett 2006). In an extreme case, households may nd themselves in a poverty trap in the aftermath of disasters. These theoretical predictions have been investigated quantitatively for several developing countries, but there is no consensus regarding the shape of the asset dynamics curve.2 Among recent studies, McKay and Perge (2011) tested for evidence of the existence of an asset-based poverty trap mechanism across seven panel datasets in developing countries, but they did not nd evidence for this mechanism. In contrast, the number of quantitative analyses on asset dynamics applied to historical contexts is small, mostly due to the nonavailability of suitable data. As an exception, the case of prewar Japanese farmers has been analyzed in several studies. For example, Kusadokoro, Maru, and Takashima (2012) analyzed asset accumulation behavior of farm households in rural Japan based on panel data from 1931 to 1941, the years of reconstruction following the Great Depression. They showed that
1 Available online: http://www.emdat.be/natural-disasters-trends (accessed on 20 October 2012). It is possible that the reported increase is partially due to an increased tendency to reportnot necessarily an increase in the occurrence ofdisasters. 2 For example, see Naschold (2005); Adato, Carter, and May (2006); Carter et al. (2007); Mogues (2011); McKay and Perge (2011); and Miura, Kanno, and Sakurai (2012).
households accumulated liquid assets such as cash, quasi-money, livestock animals, and stocks in kind, suggesting the existence of a precautionary saving motive. Analyzing the same period but with a different data source, Fujie and Senda (2011) showed that farm households maintained the amount of arable land, increased nonfarm labor supply, and decreased the use of fertilizer in response to the depression. They indicated that aggregate shocks from the Great Depression led to stagnation of agricultural growth in Japan. Both these studies applied microeconometric approaches similar to those reviewed in the previous paragraphs. Since the data requirement is high, application of these approaches to other historical cases is not straightforward. This paper attempts to ll these gaps in the literature by combining empirical analyses using long-run historical data at the macro or semi-macro levels and contemporary data collected at the household level. The analysis focuses on the province of Khyber Pakhtunkhwa (formerly known as the NorthWest Frontier Province or the NWFP) of Pakistan.3 Using these different data sources, the paper addresses the question of how assetsparticularly livestock, which is the core asset in the study arearespond to natural and man-made disasters. Long-run historical (semi-) macro data, combined with data from historical reports prepared by the government, is employed to speculate on the microeconomic mechanism underlying asset dynamics in response to natural and man-made disasters. Contemporary household-level data are then used to shed light on the speculation from a different angle. The micro panel data, which form the basis for contemporary analysis, were collected from approximately 300 households in three villages during the late 1990s, the period associated with overall macroeconomic stagnation and not with major natural disasters. Using household panel data, this paper estimates the shape of the asset dynamics curve using both nonparametric and parametric analyses. The parametric analysis additionally reveals how each type of asset responds to village-level and idiosyncratic shocks. Thus, the major contribution of this study is to demonstrate the complementarity of using both historical and contemporary analyses in understanding household vulnerability and resilience in the context of long-run economic development. To the best of the authors knowledge, there is no attempt in the literature to combine these two types of data in the manner adopted in this study. The remainder of the paper is organized as follows. Section II reviews the related microeconomic literature and outlines the empirical methodology. Section III describes the rural NWFP economy with a focus on assets, agricultural technology, and livelihood. A descriptive analysis of long-run changes using district-level and
3 Khyber Pakhtunkhwa is one of the four provinces that comprise Pakistan. In April 2010, the constitution of Pakistan was amended and the former NWFP renamed to Khyber Pakhtunkhwa. This paper identies the province as NWFP since most of the data were taken under this name.
province-level historical data on crop production and livestock is then given in Section IV . Section V examines the two-period panel data collected during the late 1990s, while Section VI provides an interpretation combining the two types of analysis. Finally, conclusions are presented in Section VII.
II. Literature and Analytical Framework
The empirical analysis in this paper is motivated by two strands of development economics literature. The rst is the consumption smoothing literature, which focuses on low-income households ability to cope with exogenous shocks (Fafchamps 2003, Dercon 2005). These studies have shown that poor households are likely to suffer not only from low levels of welfare on average but also from uctuations in welfare due to limited coping ability. The inability of households to avoid declines in welfare can be called vulnerability. Currently, there is a substantial amount of literature on the measurement of vulnerability (Ligon and Schechter 2003; Dercon 2005; Kurosaki 2006; Dutta, Foster, and Mishra 2010). In developing countries, studies on household vulnerability found that the ability to avoid declines in welfare improves with an increase in the amount of assets that can be used as a buffer. The other strand of literature relates to the asset poverty trap hypothesis by Carter and Barrett (2006). In the standard consumer theory of assets as a buffer (Deaton 1992), the next periods asset is a linear function of the current periods asset multiplied by a factor of one plus the real interest rate minus the depreciation rate. Under this condition, assets can be used to smooth consumption in response to income shocks. In this framework, a disaster that partially damages assets can be interpreted as an unexpected and transient increase in the depreciation rate. In the context of low-income developing countries, however, asset dynamics may be nonlinear. Suppose that the expected value of an asset in the next period is an S-shaped function of the initial asset that has three intersections with a 45-degree line (Figure 1). The long-run dynamics of assets are then characterized by a middle and unstable equilibrium (called the Micawber threshold, A in Figure 1) and two stable equilibriums. The lower of the two stable equilibriums ( A L in Figure 1) corresponds to the poverty trap if the welfare level associated with this level of assets falls below or settles around the poverty line. Given the existence of multiple equilibriums, Carter and Barrett (2006) argue that only those households well above the Micawber threshold can afford to use assets as a buffer to smooth consumption. They further contend that those households close to the Micawber threshold when hit by a negative income shock may rationally attempt to protect their assets to avoid falling into the asset poverty trap rather than sell these assets to smooth consumption. Therefore, we may observe asset smoothing behavior instead of consumption smoothing behavior in such cases.
80 ASIAN DEVELOPMENT REVIEW Figure 1: The Asset Dynamics Curve
Source: Figure 4. Carter, Michael, and Christopher Barrett. 2006. The Economics of Poverty Traps and Persistent Poverty: An Asset-Based Approach. Journal of Development Studies 42(2): 178199.
Although empirical support for the Micawber threshold is mixed, the concept is an attractive one.4 Even when the said threshold is not found, the shape of the asset dynamics curve and the location of the equilibriums are informative for understanding household responses to shocks. Therefore, this paper tries to estimate asset dynamics curves for several types of assets using both nonparametric and parametric analyses. The parametric analysis additionally allows the identication of the impact of exogenous shocks on asset changes. Such shocks may include permanent and transient shocks on the one hand and aggregate and idiosyncratic shocks on the other. This type of microeconometric analysis is usually conducted using micro panel data of households in developing countries in isolation from historical (semi-) macro data. Section V of this paper follows this tradition. However, by combining a statistical analysis with a historical one, we can benet from complementarity. Hence, before presenting the microeconometric analysis, Section IV of this paper provides a descriptive analysis of the long-run changes using district-level and province-level data on livestock assets. If the asset poverty trap hypothesis describes the historical data well, an exogenous shock that destroys the assets of majority of households should have a persistent impact. This persistence could possibly lead to an overall decline in assets in the district in the long run. On the other hand, if asset returns are linear and assets are used as a buffer, such a shock should have only a temporary impact and the economy should eventually revert to the initial trend. If the exogenous shock
4
See studies listed in footnote 2.
destroys the assets of a majority of households, however, the reversion to initial trend may take time even under the buffer stock hypothesis such that empirically distinguishing the two hypotheses could be difcult. The theoretical prediction regarding the impact of an exogenous shock that decreases the income of the majority of households but does not directly affect household assets also differs between the two hypotheses. Under the poverty trap hypothesis, the size of assets is not affected much by such a shock, while the size of assets decreases according to the buffer stock hypothesis. When the impact of such an exogenous shock on household income is heterogeneous, however, the shock may not affect the aggregate asset level even under the buffer stock hypothesis as effects cancel out. As a result, empirically distinguishing the two hypotheses could likewise be difcult. Meanwhile, whether a specic type of asset, such as livestock, is used more as a buffer or as productive capital compared to other types of assets depends on the availability of other consumption smoothing measures and agricultural technology (e.g., the substitutability of draft animals in farm production). The next section thus describes the means of livelihood in the rural NWFP economy, with its focus on assets and agricultural technology.
III. The Study Area
Economic development in South Asia is characterized by moderate success in economic growth and substantial failure in human development such as basic health, education, and gender equality (Dr` eze and Sen 1995). This characteristic is most apparent in the NWFP. Furthermore, the scope for economic growth based on crop agriculture is limited since the province is land-scarce and crop production is riskier than in other parts of Pakistan due to low development of irrigation. These additional hardships make the NWFP case study an interesting one in terms of investigating the relationship between asset dynamics and disasters.
A. Rural Livelihood and the Role of Livestock
The NWFP as a whole is a rural province. According to the population census conducted in 1998, 83% of its population lived in rural areas (Government of Pakistan 2012).5 Until the 1980s, a majority of rural residents were engaged in agriculture, both crop cultivation and animal husbandry. However, since the late 1980s, employment in the nonagricultural sector has been growing. Two types of nonagricultural activities are noteworthy: short-term migration (both domestic and foreign) and rural nonagricultural activities in villages. In both
5
The 1998 population census is the latest conducted by the Government of Pakistan.
types, semi-skilled work such as transport and construction work dominates in terms of employment creation. The availability of skilled or professional jobs has been limited in the province, although it has been increasing gradually in recent years. Furthermore, the average household size is larger than in other parts of Pakistan, partly reecting the norm of the Pakhtunthe ethnic majority of people in the NWFPwho highly respect family-based reciprocity and bravery in defending their land, property, family, and women from incursions, etc. (Ahmed 1980). The major crops in the NWFP are wheat in the rabi (winter) season and maize in the kharif (monsoon) season. Both these crops are cultivated as staple food, although large farmers tend to sell the surplus to the market. Sugarcane is the most important cash crop. Fodder crops to feed livestock animals also occupy a signicant share of cropped land in both kharif and rabi seasons. A particular mention must be made of the role of livestock. Most farmers in the NWFP are engaged in mixed farming, combining livestock raising and crop cultivation on a single farm. Large livestock animals include cows and female buffaloes for milk. Bullocks were once an important productive asset used for plowing and transportation. However, tractors gradually replaced draft animals, thereby decreasing the role of livestock as draft animals.6 As shown in the next section, the livestock portfolio in the NWFP has been changing from draft to milk animals. Small livestock animals, such as goats, sheep, and poultry, are common means of saving. This implies several interactions between crop farming and livestock husbandry in the study area (Kurosaki 1995). The direct interactions can be explained in the following manner: Fodder crops and dry fodder (e.g., grain straws) are fed to animals, animal excrements are processed into farmyard manure used in crop cultivation, draft animals are used in plowing and crop transportation, while crop rotations including leguminous fodder crops improve the soil fertility. Meanwhile, the indirect interactions between the livestock sector and crop farming through the household economy can be described as follows: Milk animals provide milk for consumption and cash from selling surplus milk, family labor is utilized throughout the year for taking care of animals, and livestock serve as a liquid form of asset that can be used as a buffer in a bad year. In the following sections, I analyze how these complicated interactions result in a reduced-form relation between asset dynamics and disasters (identifying each of these interactions, however, is beyond the scope of this paper). Direct interactions mentioned above are relevant only for households that operate farmland for crop cultivation (i.e., farm households).7 However, indirect interactions are important for nonfarm households as well. Income sources of
6 As the tractor service rental market is well developed, the majority of farmers who use tractors for land preparation do not own a tractor. 7 Because of a social distinction between land-operating households and other households in Pakistan (Hirashima 2008), I employ the standard categorization in which such households are called farm households.
nonfarm households could include livestock activities, nonagricultural activities, net rental receipt, transfers, etc. This implies that in the study area, agricultural assets (land and livestock) are the key assets that constitute rural livelihood. It also implies that human capital (size of labor force, education, etc.) and transport and agricultural equipment (e.g., tractors and vehicles) are also gaining importance. Hence, all these assets were included in the microeconometric analysis. Because of data availability, the historical analysis focuses on livestock animals since they wereand arethe most important asset supporting the livelihood of both farm and nonfarm households.
B.
The NWFP during the Colonial Period and the Change after Independence
In October 1901, the British government carved out the NWFP from Punjab as a separate province. The word frontier in the name implied the frontier against Russian inuence. As one of the British provinces of the Indian Empire, the NWFP was divided into districts for the purpose of its administration. During the colonial period, there were ve districts (Hazara, Peshawar, Kohat, Bannu, and Dera Ismail Khan) until 1937, when a new district of Mardan was carved out of Peshawar District. The property right of land was established under the British rule of the NWFP, where ownership was given to cultivators, as was the case in Punjab (Khalid 1998). The cultivators included village-based landlords who operated a part of their land and rented out the remainder. During the colonial period, social development such as education was highly limited in the province, while infrastructure development such as roads and irrigation canals progressed gradually. The British rule respected the local norm of self-governance in the NWFP, particularly the institution called jirgaan assembly of elders taking decisions by consensus (Ahmed 1980)as long as such decisions did not violate British rulings. Agricultural innovation such as the introduction of chemical fertilizers and improved methods of cultivation was facilitated in the province, although at a slower pace than in the post-independence period. For example, systematic agricultural research began in the NWFP in 1908 at a government research institute in Tarnab, Peshawar District. Furthermore, the colonial government introduced a modern credit facility for farmers called taccavi loans. However, the credit facility was not utilized by the majority of farmers due to limited access and high requirements of land collateral, and informal credit prevailed in villages (Malik 1999). Pakistan and India obtained independence from the British in August 1947 (the so-called Partition of the Indian Subcontinent). The NWFP now belonged to the new state of Pakistan. The basic administrative structure remained intact and the list of six NWFP districts remained the same until 1970. However, since then, the
subdivision of districts has continued due to the growth in population. At the end of 2012, there were 25 districts in Khyber Pakhtunkhwa.8 After independence, the pace of public investment in infrastructure and agricultural innovation accelerated. At the time of Partition, the percentage of cultivated area with irrigation was 38% in the NWFP. After 50 years, the corresponding gure grew to 50%. The Green Revolution technology for wheat was introduced to the province in the late 1960s. However, land property institutions remained more or less the same. Laws and regulations related to land reforms were enacted to put ceilings on land holdings, but they did not have much impact on the province as the number of large landlords had been small (Khalid 1998). Furthermore, in the post-independence period, government credit for agricultural production was expandedfor example, the Agricultural Development Bank of Pakistan was established in 1961. Nevertheless, the dependence of rural households on informal credit continued, partly because of the Islamic norm of banning interest payments and partly due to the limited resources of the public sector (Malik 1999).
IV. District-level and Province-level Analysis of Crop Production and Livestock
This section conducts long-run historical analysis using district-level and province-level data on crop production and livestock. Province-level data indicates data aggregated at the NWFP, regarded as the macro level in this study. Each district is viewed as the semi-macro level. The analysis in this section is descriptive in nature. It rst examines the time series plots for crop production and livestock and extracts statements from government reports. It then interprets the descriptive results based on theoretical predictions summarized in Section II.
A. Data
Considering the changes in district borders described in Section III, this study adopted the following geographical demarcation. For the colonial period, the paper compiled a balanced panel dataset of ve districts (after 1937, data for Peshawar and Mardan were merged to form the initial district of Peshawar). For the post-colonial period, it compiled a balanced panel dataset of six districts (Hazara, Peshawar, Mardan, Kohat, Bannu, and Dera Ismail Khan) that correspond to the district borders at the time of Partition. Original data sources and the data compilation procedure were the same as those adopted in the authors ongoing attempt to construct long-term agricultural statistics for South Asia under the Asian Historical Statistics Project at Hitotsubashi
8
The household surveys analyzed in Section V were conducted in the current district of Peshawar.
University, Tokyo (Kurosaki 2003 and 2011). For the colonial period, various issues of Season and Crop Reports published by the NWFP Government, Peshawar, were used as the main data source. The rst issue was published for the agricultural year 1902/1903 and the last for the agricultural year of 1944/1945.9 Each Season and Crop Report presents an overview for the year with regard to rainfall, agriculture, and the rural economy, with statistical tables at the district level.10 From this source, the paper compiled district-level and province-level annual data of areas under crops and output of major crops.11 The same source reports statistical tables for the district-level agricultural stock (livestock, plows, etc.) based on quinquennial livestock census. Thus, the paper obtained livestock information for the years 1903, 1904, 1909, 1914, 1920, 1925, 1930, 1935, 1940, and 1945. In compiling the dataset, typographical errors were corrected, and denitional changes were adjusted to improve comparability across years. The data source for the post-colonial period is the ofcial statistics compiled by the Government of Pakistan (Crops Area Production by Districts and Pakistan Livestock Census, with the names differing slightly depending on the publication year). Regarding the district-level crop data, the rst year for which data had been made available was 1947/1948. Therefore, a data gap of 2 years exists between the pre-1947 and post-1947 periods. Compared to the colonial period, districtlevel livestock data became less frequent after the Partitionthere were only six observations taken from the Agricultural Census (1960 and 1972) and the Livestock Census (1976, 1986, 1996, and 2006). Three types of crop variables are investigated in this paper to infer the shocks that occurred in the crop sector. Since the area planted to major crops declines when monsoon rainfall is less than normal, the rst type includes the total area sown with kharif crops (kharif_a), rabi crops (rabi_a), and wheat (wheat_a) for the pre-independence period; and further, the area sown with maize (maize_a) and wheat (wheat_a) for the post-independence period.12 As natural disasters such as droughts, oods, and hailstorms affect yield per acre, the second type incudes peracre yield of wheat (wheat_y) and maize (maize_y). Since such information is not available for the early part of the colonial period, this investigation is limited to the post-independence period. Meanwhile, the third type includes the total area of failed crops (fail_a) as a direct measure of crop production shocks. Since the information
9 An agricultural year begins in July and ends on 30 June the following year. The year 1902/1903, for instance, covers kharif crops planted in the mid-1902 and harvested in the later months of 1902, rabi crops planted in late 1902 and harvested in AprilJune 1903, and sugarcane harvested in late 1902 to early 1903. In gures with limited space, this period is represented as 1903. In Pakistan, a scal year constitutes the same periodi.e., from 1 July to 30 June the next year. 10 The sections in Season and Crop Reports on agricultural stock, agricultural deterioration, and condition of agricultural population were particularly useful for understanding shocks that occurred during the year. Unfortunately, comparable information was not available for the post-colonial period. 11 The information on the output of major crops became available from 1906/1907 onward. 12 The reason for the difference is the unreliability of maize data in the pre-independence period and the inconsistency in reporting the total kharif area in the post-independence period.
on fail_a is not available for the post-independence period reecting negligible areas under this category, the investigation only applies to the pre-independence period. Changes in these crop variables are then associated with changes in the population of livestock animals, which are the major assets of rural households in the NWFP. The livestock variables analyzed include the number of adult bulls and bullocks (bull), adult cows (cow), and adult she-buffaloes (buf_f).13
B. Impacts of Disasters before Independence
Figure 2 plots the time series of crop production and livestock population for the colonial period. Panel A shows the result for the province. First, there was no longrun trend in crop production and livestock population. This is in sharp contrast to the Punjab area of Pakistan before Partition where there had been sustained agricultural growth (Kurosaki 2003). Since there had been population growth in the NWFP during the rst half of the century, the relative stagnation of agriculture in this area implies that its dependence on agriculture in Punjab for food increased.14 Second, crop production uctuated substantially from year to year. Third, area uctuations in kharif crops and in wheat were not synchronized. In certain years, only one of the two experienced a fall while the other experienced a rise, while in other years, both of them moved in the same direction. Fourth, livestock population experienced an increase until 1914, then declined in two subsequent censuses in 1920 and 1925, after which the population remained stable. The gure clearly suggests that the agricultural year 1920/1921 was a particularly bad year, followed by a substantial decline in the livestock population. From 1920 to 1925, the population of adult bulls and bullocks in the province declined by 5.4%, adult cows by 5.3%, and adult she-buffaloes by 9.3%. The time series plot for Peshawar District (panel B, Figure 2) is similar to that for the entire province. Of particular importance had been the decline in the livestock population from 1920 to 1925 and the substantial crop production shock in 1920/1921. The similarity in the time series plot is expected since the colonial district of Peshawar was the most important district in the NWFP in terms of agriculture, accounting for approximately one-third of the cropped area in the entire province, and the extent of spatial specialization was weak due to lack of infrastructure and low level of urbanization. On the other hand, panel B differed noticeably from panel A in terms of trends after the mid-1920s. Bottoming in the years around 1925, cropped areas and the livestock population in Peshawar District grew, albeit
13 Since a more detailed classication of animals is available for the period after Partition and the distinction among animals according to purpose quite important in the study area, the paper considers only adult bullocks used as draft animals, adult cows in milk, and adult she-buffaloes in milk after Partition. Therefore, the absolute level of livestock population is not comparable between pre-independence and post-independence periods. 14 The population of the NWFP districts analyzed in this paper grew from 2.04 million in the 1901 census to 3.25 million in the 1951 census. The corresponding gure in the 1998 census was 13.48 million.

Figure 2: Crop Production and Livestock in the NWFP before Partition, 19031945
A. The NWFP 2,250 2,000 1,750 1,500 1,250 1,000 750 500 250 1903 1905 1907 1909 1911 1913 1915 1917 1919 1921 1923 1925 1927 1929 1931 1933 1935 1937 1939 1941 1943 1945 B. Peshawar District 800 700 600 500 400 300 200 100 0 1903 1905 1907 1909 1911 1913 1915 1917 1919 1921 1923 1925 1927 1929 1931 1933 1935 1937 1939 1941 1943 1945 wheat_a bull kharif_a cow rabi_a buf_f fail_a
NWFP = NorthWest Frontier Province. Note: (1) Crop variables (wheat_a, kharif_a, rabi_a, and fail_a) are plotted on the left axis (in 1,000 acres). (2) Livestock numbers (bull, cow, and buf_f) are plotted on the right axis (in 1,000 heads). (3) Peshawar District corresponds to the district borders until 1937. It comprises the current districts of Peshawar, Charsadda, Nowshera, Mardan, and Swabi. Source: Authors representations using the dataset described in the text.
450 400 350 300 250 200 150 100 50
160 140 120 100 80 60 40 20 0
gradually, until the year of independence, while the crop failure rate declined. This could be attributed to agricultural innovation facilitated by systematic agricultural research. From Season and Crop Reports, several statements are extracted below regarding agriculture in the NWFP. For example, with regard to the livestock decline in 1920, the publication stated that the The recent cattle census [February 1920] came at rather an unfortunate time following, as it did, a year of war and frontier disturbances, and also a severe season of drought (in the barani [rainfed] tracts) in 1918. Widespread epidemics of cattle disease followed which the already greatly
debilitated stock was unable to withstand.15 Regarding the decline in livestock in 1925, it was stated that Bulls, cows and cow-buffaloes decreased by 6, 2 and 9 per cent, respectively. The drought of 192022 and consequent scarcity of fodder, cattle diseases and plague, which prevailed more or less in all districts during the period under report, were mainly responsible for this.16 In sharp contrast, I nd no such statements for the other years. With regard to the interaction between crop and livestock sectors, notable descriptions in the NWFP Season and Crop Reports include the following: Two poor harvests (except in the Hazara District) combined with a very serious epidemic in the autumn [Spanish u] have occasioned a passing check to agricultural prosperity.17 The abnormally severe drought experienced during the year under report has been a great trial to the agricultural population who have had to dispose of their plough cattle in many tracts in order to raise money to buy food. Seed stocks have mostly been consumed as food . . . The condition of the agricultural population was generally very unsatisfactory throughout the Province as both the Kharif and Rabi harvests were poor and the supply of water and fodder was insufcient on account of prolonged drought.18 There were also statements regarding natural disasters other than droughts, for example, regarding hailstorms and oods. With regard to oods, all the statements found in the reports are related to local oods that affected only a particular portion within a district.19 This is in sharp contrast to the nationwide oods that hit Pakistan in JulyAugust 2010 (Kurosaki and Khan 2011). Moreover, no statement was found in which the livestock population change was associated with crop shocks due to hailstorms and oods. With regard to the Great Depression, the most detailed description was given in the 1930/1931 edition of the publication (p. 9): The fall in prices and the resulting contraction in the credit of the cultivator, the repression in trade, and the shortage of moneyall aspects of the same phenomenonhave caused the greatest inconvenience to the agricultural community in the Peshawar District where money has to be raised to pay cash rents and Government dues particularly for water-rate. The result is that very large arrears are outstanding in spite of the general remissions
Page 5 of the 1919/1920 edition, with brackets added by the author. Page 7 of the 1924/1925 edition. 17 Page 5 of the 1918/1919 edition, with brackets added by the author. 18 Pages 5 to 6 of the 1920/1921 edition. 19 See page 2 of the 1903/1904 edition, page 2 of the 1908/1909 edition, pages 1 to 2 for the 1910/1911 edition, and page 6 of the 1921/1922 edition.
16 15
and reductions designed to counter the fall in prices. Similar but shorter statements were found in the report in the years that followed, until 1938/1939. However, it is rather difcult to nd an impact of the depression on either crops or livestock in Figure 2. The absence of an impact on crops could be due to the difculty in clearly designating the year (or years) of the disaster or to the indirect nature of the disasters impact on crop production.
C.
Crop Production and Livestock Population after Independence
Figure 3 plots the time series of areas and per-acre yield of wheat and maize. It also plots the livestock population found in six agricultural/livestock censuses after independence. Panel A presents the result for the province. First, all four time series for crops show sustained and continuous growth. This is similar to the case of Punjab Province after Partition (Kurosaki 2003). Second, crop production seemingly uctuated from year to year. However, signicant reductions were less frequently observed after Partition than before Partition except for a sudden drop in per-acre yield of wheat in 2000/2001. In 2010/2011, when unprecedented oods hit Pakistan (Kurosaki and Khan 2011), maize yield was adversely affected (direct effect of oods), but maize area was not, as the crop had already been planted when oods came. Moreover, wheat area was also adversely affected due to farmers preoccupation with reconstruction and oods destruction of irrigation and other facilities, but wheat yield improved since the oods fertilized the soil. Overall, the impact of the 2010 oods does not seem substantial from the macro viewpoint regarding crops. Third, two trends are evident in livestock population: a continuous decrease in the number of bullocks and a continuous increase in the number of cows and she-buffaloes in milk. As data are available only for 6 years with 10-year intervals on average, this paper could not examine how the livestock population responded to crop shocks in the short to medium run. However, the gure clearly shows that there has been no discernible instance of livestock damage due to crop shocks that persisted for over a decade. The time series plot for Peshawar District (panel B, Figure 3) appears rather different from that for the entire province, the major dissimilarity being the sustained growth observed only in the per-acre yield of wheat. The area under wheat and under maize and the per-acre yield of maize have all been stagnant since the early 1970s. There was growth in spatial specialization during the post-independence period owing to the development of infrastructure and cities. As a result, agriculture in Peshawar underwent a transformation to include high-value activities such as horticulture, plant nursery, and livestock husbandry. Because of this, the shape of the time series plot in Peshawar District after independence deviates from that at the provincial level. On the other hand, the improvement in per-acre yield of wheat in the late 1970s was substantial (the late arrival of the Green Revolution).
90 ASIAN DEVELOPMENT REVIEW Figure 3: Crop Production in the NWFP after Partition, 19482011
A. The NWFP
300 2,100
250
1,750
200
1,400
150
1,050
100
700
50
350
0
1948 1950 1952 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
B. Peshawar District
300 300
250
250
200
200
150
150
100
100
50
50
NWFP = NorthWest Frontier Province. Note: (1) Crop variables (wheat_a, wheat_y, maize_a, and maize_y) are plotted on the left axis (unit: index with 1948 = 100. The absolute level in 1948 for the four variables are 410 (1000 ha), 695 (kg/ha), 190 (1000 ha), 995 (kg/ha) for the NWFP and 68 (1000 ha), 828 (kg/ha), 36 (1000 ha), 1550 (kg/ha) for Peshawar. (2) Livestock numbers (bull, cow, and buf_f) are plotted on the right axis (unit: 1,000 heads). (3) Peshawar District corresponds to the district borders in 1948. It comprises the current districts of Peshawar, Charsadda, and Nowshera. Source: Authors representations using the dataset described in the text.
Regarding the livestock population, trends similar to those for the NWFP are observed in Peshawar District, but with steeper slopes for the decrease in the number of bullocks. The diversication toward milk animals in Peshawar District thus occurred at a faster pace than at the provincial level.
D. Interpreting the Historical Patterns in Assets from the Viewpoint of Microeconomics
The descriptive analysis presented above showed that agriculture in the NWFP, particularly before Partition, had been affected by several natural disasters, mostly
1948 1950 1952 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
wheat_a wheat_y maize_a maize_y bull cow buf_f
droughts. This led to a decline in the livestock population that persisted for over 5 years. On the other hand, such persistent declines in the livestock population were not observed in district-level data after independence. This indicates that during the post-1947 period, persistent declines in the livestock population due to such shocks had been avoided by the province. In this subsection, I provide a speculative interpretation of this contrast based on the theoretical predictions summarized in Section II. The pre-independence observations could be consistent with both the asset poverty trap hypothesis and the buffer stock hypothesis. The descriptive analysis showed that in 1920/1921, natural disasters damaged the livestock population (which already suffered from small disasters in the preceding years) so intensively that the livestock population level did not recover to the 1920 level even in 1925. The persistence of the damage could be more consistent with predictions under the poverty trap hypothesis than under the buffer stock hypothesis. However, as the droughts killed a number of animals directly, the recovery to the initial trend could have taken a long time even under the buffer stock hypothesis. The statement from the 1920/1921 Season and Crops Report quoted above (agricultural population . . . had to dispose of their plough cattle in many tracts in order to raise money to buy food) also supports the view that livestock had been used as a buffer. The descriptive analysis also showed that the Great Depression failed to affect the trends in livestock population. On one hand, this could be interpreted under the poverty trap hypothesis as the absence of a direct impact of the shocks on assets. On the other, this could be interpreted under the buffer stock hypothesis as income shocks being heterogeneous among households, where some farmers sold livestock to cope with the negative shocks while others purchased them, resulting in nonresponse of the livestock population at the district level. The post-independence observations seem more consistent with the buffer stock hypothesis than with the poverty trap hypothesis. The descriptive analysis showed the absence of persistent declines in the livestock population despite several instances of crop shocks that should have reduced the livestock population directly in the short run. If asset returns are almost linear and the assets are used as a buffer, such a shock would have only a temporary impact and the economy would revert quickly to the initial trend. This theoretical prediction is consistent with Figure 3, as it does not show any disturbance in the trends in the livestock population. However, this observation could also be consistent with the poverty trap hypothesis with nonlinear asset returns if farmers had sufciently diversied portfolios so that several instances of crop shocks shown in Figure 3 did not actually reduce the livestock population substantially. Unfortunately, the unavailability of more frequent and/or more disaggregated data on the livestock population does not allow us to explore this possibility further. The historical description of the study areas in Section III could provide additional support to the interpretation that post-independence livestock dynamics
seem more consistent with the buffer stock hypothesis. As shown in the section, the post-independence period was characterized by better infrastructure, greater availability of formal credit in villages, and agricultural technology where draft animals were substitutable with tractor services. My speculation is that the combination of changing agricultural technology and better opportunities for villagers to spread risk across and within villages had been responsible for the contrast between pre-independence and post-independence periods.
V. Household-level Asset Dynamics in 19961999
Although suggestive, the empirical results given in the previous section were at the aggregate level, and hence not indicative of asset dynamics at the household level. The speculations discussed above need to be supplemented by micro-level evidence. Therefore, in this section, the dynamics are examined using a detailed panel dataset of households collected from three villages in Peshawar District during the late 1990s.
A. Data
The panel dataset was compiled from the baseline survey conducted in the scal year 1996/1997 and the resurvey conducted 3 years later (Kurosaki 2006, Kurosaki and Khan 2006).20 The baseline survey covered 355 households randomly chosen from three villages in the Peshawar District. Sample villages were purposely chosen so that they would be similar in terms of size, historical background, and tenancy structure, but different in terms of irrigation level and access to the main market (Peshawar, the provincial capital of NWFP). The intention for this method of choosing villages was to infer long-run development implications by comparing the three villages. Using a detailed questionnaire, information was collected on household roster, agricultural production (corresponding to the agricultural year of 1995/1996), employment, assets, etc. in the baseline survey (referred to as the 1996 survey). The resurvey, conducted 3 years after, collected crop information for the agricultural year 1998/1999 (the 1999 survey). Out of 355 households surveyed in 1996, 304 households were resurveyed successfully. Among those resurveyed, three were divided into multiple households, while two had incomplete information on consumption. Therefore, a balanced panel dataset of 299 households with two periods
20 Existing papers using the same dataset, including Kurosaki (2006) and Kurosaki and Khan (2006), did not analyze the dynamics of assets.

Table 1. Characteristics of the Sample Households in the 19961999 Panel Survey, the NWFP of Pakistan
Variable 1. Village Characteristics Agriculture Distance to main roads (km) Population (1998 Census) Adult literacy rates in % (1998 Census) 2. Characteristics of Households in the Panel Number of households Average of initial characteristics in the 1996 surveya Household size Literacy rate of working age adults (%) Average farmland owned (acres) Per capita value of farmland asset (PRs100,000)b,c 1996 survey 1999 survey Per capita value of livestock asset (PRs1,000)b,c 1996 survey 1999 survey Per capita income (poverty line units)b,d 1996 survey 1999 survey Per capita consumption (poverty line units)b,d 1996 survey 1999 survey Village A Rainfed 10 2,858 25.8 83 10.75 16.8 5.51 0.475 0.306 1.525 1.087 1.074 0.856 0.743 0.773 Village B Rain/Irrigated 4 3,831 19.9 111 8.41 17.6 1.28 0.804 0.417 1.181 0.727 1.278 0.953 0.868 0.828 Village C Irrigated 1 7,575 37.5 105 8.95 31.3 1.43 0.376 0.394 1.884 1.042 1.861 1.225 1.110 1.148
PRs = Pakistan rupees. a Averages over all sample households. b Averages based on individuals with the number of household members as weights. c In 1996 prices, adjusted for ination. d The same poverty line adjusted for ination was used for each period (1996 and 1999) for both income and consumption and for all villages to convert per capita income (consumption) into poverty line units. Source: Authors computations based on 19961999 panel data described in the text.
was utilized for the analysis.21 After the 1999 survey, I revisited the villages several times, observing changes in a casual manner. Table 1 summarizes the characteristics of sample villages and households. Village A is rainfed and located at a considerable distance from the main roads, serving as an example of the least developed villages. Village C is fully irrigated and located close to a national highway, serving as an example of the most developed
21 To infer the potential bias due to attrition, I rst regressed the attrition dummy on village dummies and household initial characteristics. The probit result shows that attrition occurred more among households living in Village A than in Villages B and C and among households whose heads were more educated. Other household attributes were not statistically signicant. As the probit result shows that attrition was not completely random on observables, I conducted a test developed by Becketti et al. (1988). The welfare ratio in the rst survey was regressed on the baseline characteristics of households, an attrition dummy, and the attrition dummy interacted with the other explanatory variables. None of the coefcients corresponding to an additional slope for the attrition households was signicant and the joint signicance test did not reject the null hypothesis at the 20% level. Therefore, there is unlikely to be a signicant attrition bias in the estimates provided in this paper. Detailed results are available on request.
villages. Village B falls in between Villages A and C. Average household sizes are larger in Village A than in Villages B and C, thereby reecting the stronger prevalence of an extended family system in the village. Average landholding sizes in acres are also larger in Village A than in Villages B and C. Since the productivity of purely rainfed land is substantially lower than that of irrigated land, effective landholding sizes are comparable among the three villages as shown in the statistics for per capita values of land assets in Table 1. Household income and consumption were calculated by including the imputed values of nonmarketed transactions. Average income and consumption per capita were lowest in Village A and highest in Village C, in line with the survey objective of selecting villages with different levels of economic development. In terms of education, Village C had higher achievement levels than the other two villages. As shown in Kurosaki and Hussain (1999), nonagricultural income constituted a larger share in the total household income than agricultural income in 1996, regardless of the land operation status of households in all three villages. In all three villages, nonagricultural, unskilled wage work accounted for approximately one-third of household income. Among other nonagricultural sources, migration income was important in Village A, while self-employed business was important in Village C. Self-employment income from livestock activities accounted for approximately 10% of household income, including nonfarm households. As shown in Table 1, the average household income per capita declined substantially from 1996 to 1999. This was mostly due to the macroeconomic stagnation of Pakistans economy associated with political turmoil, which affected the NWFPs economy most severely among the four provinces. As shown in Figure 3 in the previous section, there was no province-wide agricultural shock that affected both the rabi and kharif crops during 19961999. Therefore, the analysis in this section is intended to capture the asset dynamics in years with a man-made disaster but without major natural disasters. On the other hand, Table 1 shows that the average consumption in 1999 remained similar to the level in 1996. As will be shown below, sample households sold assets, mostly livestock, to supplement the reduced income. In addition to the macroeconomic shock, household-level consumption was also subject to idiosyncratic shocks, thereby resulting in substantial uctuation. Kurosaki (2006) presented a transition matrix of consumption poverty with ve categories of poverty status in each year and indicated a highly frequent perturbation of the poverty status at the micro level. This variation is utilized in this section to assess the asset dynamics. A note needs to be provided to justify the use of a dataset that is somewhat dated. The advantage of having historical semi-macro data that cover the period both before and after the micro panel survey of households is the main reason for using this micro dataset. In addition, because of the village selection strategy, the economic conditions in Village A during the panel survey appear to correspond to the semi-macro picture during the 1960s to 1970s in panel B of Figure 3, those in
Village B to the semi-macro picture during the 1980s to 1990s, and those in Village C to the semi-macro picture during the 2000s. Based on this observation, Section VI combines the microeconometric analysis in this section and the historical analysis in the previous section. Furthermore, because of the low economic growth rates and the slow pace of social transformation in Pakistans economy in recent decades, the basic economic behavior of households during the second half of the 1990s is of relevance to development issues in Pakistan currently. For example, the importance of livestock in the asset portfolio of households was conrmed in a recent resurvey of the study region (Kurosaki and Khan 2011). For these reasons, I regard the analysis of this panel dataset as highly relevant for the purpose of this paper.
B. Empirical Strategy
Motivated by the asset poverty trap hypothesis given by Carter and Barrett (2006), this paper follows the empirical strategy of McKay and Perge (2011). First, the shape of the asset dynamics curve is estimated, with Yi ,1999 the asset level of household i in 1999regressed on Yi ,1996 , the corresponding value in 1996, Yi ,1999 = f (Yi ,1996 ) + u i , (1)
where f(.) is the unknown function and ui is a zero-mean error term. To let the data determine the shape of the function, I employ a nonparametric approach to estimate the function. If the expected value of the asset in 1999 is an S-shaped function of the initial asset with three intersections with the 45-degree line, as shown in Figure 1, the long-run dynamics of assets are characterized by the middle and unstable equilibrium (the Micawber threshold) and two stable equilibriums. The lower of the two stable equilibriums may correspond to the poverty trap. A parametric model is then estimated that controls for various shocks and initial conditions as the asset dynamics curve is likely to be affected by these factors. As semi-parametric analyses are computationally expensive in general and not feasible in this case due to the small sample size, I adopt a completely parametric model in which the function f(.) in equation (1) is proxied by a polynomial function. Thus, the following model is estimated,
3 4 Yi ,1999 Yi ,1996 = a1 Yi ,1996 + a2 Yi2 ,1996 + a3 Yi ,1996 + a4 Yi ,1996
+ a5 Yi5 ,1996 + X i b1 + Dv dv + Z i + u i ,
(2)
where Xi is a vector of household-level variables that might affect the asset dynamics; Dv is a vector of village dummies; Zi is a vector of variables that characterizes shocks experienced by each household; and a1 , a2 , a3 , a4 , a5 , b1 , dv , and are the parameters to be estimated. Note that the effect of household-level time-invariant factors on the
asset level is controlled cleanly as the rst difference is employed as the dependent variable. The empirical specications for Xi , Dv , and Zi are discussed in the next subsection. Using specication (2), I rst examine the shape of the polynomial 3 functiondened as the tted values of (1 + a1 )Yi ,1996 + a2 Yi2 ,1996 + a3 Yi ,1996 + 5 4 a4 Yi ,1996 + a5 Yi ,1996 as an asset dynamics curve conditional on Xi , Dv , and Zi . Its shape is then compared with the unconditional counterpart estimated from equation (1). A similar parametric approach is adopted in the literature, for example, by Naschold (2005) and McKay and Perge (2011), although they use a fourth-order polynomial. If the null hypothesis of a2 = a3 = a4 = a5 = 0 is not rejected, the linear specication is supported. I then examine the coefcient vectors of dv and . By comparing dv across different types of assets, one can characterize how each asset responds to villagelevel aggregate shocks. By comparing for different types of shocks and assets, it is possible to infer which asset is more responsive to a particular type of idiosyncratic shock. Although any empirical measure of household-level shocks may contain aggregate components, the inclusion of village xed effects absorbs the effects of the aggregate components so that the coefcient can be interpreted as a measure of the asset response to idiosyncratic components of an observed measure of householdlevel shocks. Thus, equation (2) is a parameterized version of equation (1) that focuses on the asset response to shocks.22 One can refer to Mogues (2011) and Kusadokoro et al. (2012) for other empirical attempts where both equations (1) and (2) are estimated, with focus on the asset response to shocks. With regard to the type of assets, equations for livestock and land assets are rst estimated separately. An analysis of a composite asset (called the livelihood asset below) follows. This aggregates the vector of various types of human capital, social capital, and physical assets that contribute to the well-being of the household. Following the empirical methodology by Adato, Carter, and May (2006), the livelihood asset is estimated in the following steps. First, for each household in each year, per capita consumption expenditure is calculated, which includes the imputed values of in-kind transactions. Second, per capita expenditure is divided by the poverty line in each year that corresponds to the ofcial poverty line. This measure is called the welfare ratio and is reported in Table 1. Third, the welfare ratio is regressed linearly on various types of assets. The vector of assets includes village xed effects; demographic variables (household size, female ratio, dependency ratio, female head dummy, and age of household head); the literacy rate of working-age adults; monetary assets; machinery and equipment
22 Although equation (2) allows us to examine the different responses of assets to aggregate versus idiosyncratic shocks, I cannot examine their different responses to transient versus permanent shocks due to data limitationsthe two-period panel data is too short for the latter analysis.
(agricultural, nonagricultural, and consumption durables); value of owned land and livestock animals; and income sources (access to nonfarm income and remittance receipts). The tted value of this regression is the resulting estimate for the livelihood asset. The coefcients on assets used in the aggregation give the marginal contribution to livelihood of the j different assets (Adato, Carter, and May 2006, p. 233).
C. Shape of the Asset Dynamics Curves
Figure 4 shows the estimation results using the locally weighted scatter plot smoothing (LOWESS) methodology. The black curve represents the LOWESS t, while the grey, straight line represents the 45-degree line. The shape and corresponding equilibrium values remained qualitatively the same when the fractional polynomial t was used instead, or f(.) in equation (1) replaced by a polynomial function up to the fth degree as in equation (2). Panels A and B show that the dynamics curves for livestock and farmland have a single long-run equilibrium. As the curve intersects the 45-degree line from above, the single equilibrium is stable. The exact level of land or livestock equilibrium is close to the household average. Since the majority of the households are poor, this appears to indicate that the long-run equilibrium is associated with poverty.23 The shape of the asset dynamics curve changes slightly when various types of assets are aggregated into a scalar of the livelihood asset following the methodology given by Adato, Carter, and May (2006).24 Panel C of Figure 4 shows the results when the LOWESS method is applied to the livelihood asset. The gure depicts an S-curve with two stable equilibriums. The lower of the two corresponds to the poverty trap dened by Carter and Barrett (2006), since it is at the level around the poverty line and the other (the highest intersection) could correspond to a middle-class income level, far beyond the poverty line. At the same time, however, observations are scattered over the tted curve with a large variance, indicating that actual asset dynamics are subject to substantial stochastic shocks.25 A large unexplained variance is evident from panels A and B as well. To explain some of this unexplained variance, it would be useful to control for shocks and initial conditions. For this reason, I estimate the parametric model of
23 For example, when the ofcial poverty line was applied to the dataset, the poverty headcount ratio among the sample households stood at 67% and the poverty gap ratio at 20%. 24 See Appendix Table for the regression results used to construct the livelihood asset. In estimation, the larger sample including attrition or split households was used to fully utilize the cross-section information. The qualitative results remained the same when the subsample of 299 households was used. 25 Although not shown in the gure, the 95% condence interval zone estimated by bootstrapped standard errors contains the 45-degree line for almost the entire distribution of the livelihood asset in 1996. Therefore, the three equilibrium points are not statistically signicant. In this sense, Figure 4 conrms what is indicated in existing literature that multiple equilibriums are not found in Pakistan (Naschold 2005).
98 ASIAN DEVELOPMENT REVIEW Figure 4: Nonparametrically Estimated Asset Dynamics Curves in the NWFP of Pakistan, 19961999
A. Livestock per Capita (PRs1,000 at 1996 prices) 9 8 7
1999 Livestock
6 5 4 3 2 1 0 0 1 2 3 4 5 6 7 1996 Livestock 4.0 3.5 3.0 B. Owned Farmland per Capita (PRs100,000 at 1996 prices)
1999 Farmland
2.5 2.0 1.5 1.0 0.5 0 0 0.5 1.0 1.5 2.0 2.5 1996 Farmland 3.0 C. The Livelihood Asset (the poverty line)
1999 Asset Index (poverty line units)
2.5 2.0 1.5 1.0 0.5 0 0 0.5 1.0 1.5 2.0 2.5 1996 Asset Index (poverty line units)
Notes: Nonparametrically estimated by LOWESS with a bandwidth of 0.8. The scatter plots on the vertical axis are raw observations without any control. Source: Authors computations.
equation (2). As controls for the household characteristics, Xi in equation (2), three demographic variables are included: the initial number of household members, the change in the number of household members, and the literacy rate of working-age adults in the baseline survey. These variables, together with a polynomial function of the lagged asset variables, control for households activities and available consumption smoothing measures. Another reason for including the two variables regarding household size is that asset variables are dened such that they are affected by demographic changes by construction. As proxy variables for household-level shocks, Zi in equation (2), the dataset includes 14 dummy variables collected in the 1999 survey with respect to shocks that hit the household during the 3 years. From these 14 variables, I created three indicator variables that take a positive value if the household was hit by shocks that decreased its income and welfare. The three variables are shocks in farming, off-farm wage work, and others.26 The denition and summary statistics of these household-level shock variables are provided in the footnotes to Table 2. The regression results are reported in Table 2. In all three cases, the null hypothesis of linearity is rejected at the 5% level. All three coefcients on the linear lagged asset variable are between 1 and 0, thereby suggesting local convergence evaluated at the mean.27 As the null hypothesis that slopes of explanatory variables are the same across villages was not rejected except for the intercept, I report the results based on equation (2) assuming village-specic intercepts. Based on the results in Table 2, I plot the estimated asset dynamic curves in Figure 5. The black curve represents the tted value of (1 + a1 )Yi ,1996 + a2 Yi2 ,1996 + 5 4 a3 Yi3 + a Y + a Y , while the scatter plot is replaced by the observed 4 i ,1996 5 i ,1996 ,1996 value minus the tted value of Xi b1 + Dv dv + Zi . Because of the contribution of these controls, observations net of the controls are scattered over the tted curve with a smaller variance than that depicted in Figure 4. However, what is striking is the similarity of the asset dynamics curves. Panels A and B of Figure 5 show that the dynamics curves for livestock and farmland are associated with a single long-run equilibrium. The precise level of land or livestock equilibriums is close to the level shown in Figure 4. The shape of the asset dynamics curve for the livelihood asset appears to be an S-curve as well (panel C, Figure 5). However, the tted curves and the 45-degree lines are very similar in the wide range of the asset level that corresponds to the welfare level that ranges between 1 to 1.75 poverty line units.
26 The major portion of variation in these three variables is idiosyncratic. The ANOVA decomposition suggests that the between-village components explain only 0.60% of the total variation for Agricultural shock, 0.68% for Off-farm work shock, and 1.05% for Other shocks. The authors observations in the eld also support this view. For example, farmers were subject to highly idiosyncratic agricultural shocks, such as plot-specic wild animal/pest attacks, farmer-specic unfavorable selling prices, etc. 27 The convergence test was conducted at different quartiles of the lagged asset distribution. The results were also consistent with local convergence at these evaluation points. Detailed results are available on request.
100 ASIAN DEVELOPMENT REVIEW Table 2. Response of Assets to Village-level and Household-level Shocks
Dependent Variable: Change in Assets from 1996 to 1999 D_livestock Initial Level of Each Asset Linear Squared Cubic Fourth degree Fifth degree Demographic Controls Initial household size 0.801 (0.210) 0.023 (0.085) 0.049 (0.053) 0.007 (0.006) 0.000 (0.000) D_farmland 0.340 (0.085) 0.300 (0.203) 0.155 (0.072) 0.021 (0.009) 0.001 (0.000) D_livelihood asset 0.421 (0.186) 0.233 (0.637) 1.179 (0.749) 1.325 (1.062) 0.269 (0.803) 0.007 (0.007) 0.023 (0.005) 0.106 (0.136) 0.020 (0.050) 0.009 (0.054) 0.100 (0.029) 0.033 (0.032) 0.055 (0.046) 0.047 (0.040)
Continued.
0.019 0.012 (0.015) (0.008) 0.021 Change in household size 0.031 (0.017) (0.012) Literacy rate of working age adults 0.329 0.429 (0.360) (0.131) Response to a Village Level Shock (coefcient on the xed effect) 0.272 Village A 0.656 (0.137) (0.088) 0.109 Village B 0.699 (0.178) (0.083) 0.097 Village C 0.446 (0.164) (0.078) Response to a Household Level Shock Agricultural shock 0.003 0.021 (0.122) (0.080) Off-farm work shock 0.072 0.065 (0.233) (0.088) Other shocks 0.200 0.138 (0.151) (0.061)
Thus, Figures 4 and 5 suggest that the dynamics of household landholding and livestock are associated with a single long-run equilibrium. When human capital is added, the dynamics curve changes its shape but this is not sufciently nonlinear to produce statistically signicant multiple equilibriums. Therefore, the tentative conclusion is that the poverty trap hypothesis a ` la Carter and Barrett (2006) does not explain the behavior of household assets in the NWFP during the late 1990s.
D. Response of Assets to Village-level and Household-level Shocks
This subsection discusses the estimated coefcients related to the shocks presented in Table 2. First, the coefcients on village dummies, dv in equation (2),

Table 2. Continued.
Dependent Variable: Change in Assets from 1996 to 1999 D_livestock R-squared F-stat for zero coefcient on nonlinear terms F-stat for homogenous village xed effect F-stat for zero coefcient on household-level shocks 0.784 7.50 0.87 0.65 D_farmland 0.389 2.97 3.35 2.06 D_livelihood asset 0.314 3.73 1.85 1.48
= signicance at the 1% level, = signicance at the 5% level, = signicance at the 10% level. Notes: 1. Figures in parentheses are Huber-White robust standard errors. 2. Weighted least squares (WLS) regression with village xed effects (no intercept) is employed, where weights are the number of household members inated by village-specic ination factors. 3. To make the coefcients on village xed effects readily interpretable, the initial asset level, household size, and literacy rates are replaced by their deviations from the mean. Table 1 and Appendix Table provide summary statistics of these variables. 4. The following are the denitions and statistics of household-level shocks: Agricultural shock, represented by an index of agricultural shocks experienced by the household between 1996 and 1999, with a value of +2 if the household experienced both crop failure and output prices lower than the market rate, +1 if it experienced either, 0 if it experienced neither, 1 if it experienced either a bumper crop or output prices higher than the market rate, and 2 if it experienced both. The mean is 0.074 and the standard deviation 0.721; Off-farm work shock, represented by an index of shocks to off-farm work conditions of the household between 1996 and 1999, with a value of +2 if the household experienced both the loss of employment and a decrease in wage rates, +1 if it experienced either, 0 if it experienced neither, 1 if it experienced either a gain in employment or an increase in wage rates, and 2 if it experienced both. The mean is 0.036 and the standard deviation 0.532; and Other shocks, represented by an index of other shocks experienced by the household between 1996 and 1999, such as unexpected deaths and funerals and discontinuation of remittances from family members living outside the village, with a value of +1 for a negative shock, 0 if there was no such shock, and 1 for a positive shock. The mean is 0.098 and the standard deviation 0.565. Source: Authors computations based on 19961999 panel data described in the text.
show an interesting contrast across the three types of assets. All three of the village xed effects are negative and statistically signicant when the dependent variable is the change in livestock assets. This indicates that sample households sold livestock to supplement the reduced income when the three villages were hit by macroeconomic stagnation. On the other hand, there was a signicant reduction in farmland in Village A only. In the farmland asset regression, the null hypothesis of homogenous village xed effects is rejected at the 1% level. My interpretation is that this reects the cost of inferior access to markets in Village A. Because of isolation, farm households in Village A had to sell or mortgage part of their farmland to cope with aggregate negative shocks. In contrast, farm households in Villages B and C did not need to use their land since they had access to other smoothing measures. Another interesting intervillage difference is that the livelihood asset increased slightly in Village C, while it remained at the same level in the other two
102 ASIAN DEVELOPMENT REVIEW Figure 5: Parametrically Estimated Asset Dynamics Curves in the NWFP of Pakistan, 19961999
A. Livestock per Capita (PRs1,000 at 1996 prices) 9 8 7 1999 Livestock 6 5 4 3 2 1 0 0 1 2 3 4 5 6 7 1996 Livestock B. Owned Farmland per Capita (PRs1,000 at 1996 prices) 4.0 3.5 3.0 1999 Farmland 2.5 2.0 1.5 1.0 0.5 0 0 0.5 1.0 1.5 2.0 2.5 1996 Farmland C. The Livelihood Asset (the poverty line) 3.0 1999 Asset Index (poverty line units) 2.5 2.0 1.5 1.0 0.5 0 0 0.5 1.0 1.5 2.0 2.5 1996 Asset Index (poverty line units)
Notes: Parametrically estimated by OLS. The scatter plots on the vertical axis are raw observations net of asset changes predicted by other controls included in Table 2. Source: Authors computations.
villages. During the 3 years spanning the two surveys, I observed in the eld that there was a rapid diversication of the economy in Village C, with growth in new activities such as the plant nursery business and commuting to the city of Peshawar. In such circumstances, the livelihood asset in this village increased because the livelihood asset is a positive function of human capital (see Appendix Table) and the human capital level increased in this village during the 3 years that the survey was conducted.28 However, these are speculations without solid evidence, as village xed effects can capture any unobservable factor. The expectation was that the coefcients on household-level negative shocks, in equation (2), would be negative. Estimations revealed that eight out of nine coefcients were indeed negative (Table 2). However, only one of them (the impact of other shocks on the change in farmland) was statistically signicant. The signicant coefcient suggests that there was a depletion in farmland when the household was hit by a shock that was not related to agriculture or off-farm wage work. The overall insignicance of these idiosyncratic shocks suggests that, on average, such shocks did not directly reduce assets, and households did not need to reduce their assets after these shocks. Thus, the estimation results reported in Table 2 regarding coefcients on shock variables are consistent with the behavior in which households use assets as a buffer. These results were robustly supported through other specications.29 For example, when the list of household initial characteristics in Xi of equation (2) was expanded, the additional variables had insignicant coefcients and other coefcients remained highly similar to those in Table 2. This is probably because the lagged asset value on the right-hand side of equation (2) already controls for most of the impact of such variables on the asset dynamics. I also attempted several alterations for the denition of household-level shocks and different weights used in regression. Regardless of the alterations, the estimation results are qualitatively the same as those reported in Table 2.
VI. Combining Microeconometric and Historical Analyses A. Interpreting the Microeconometric Results from the Historical Perspective
The microeconometric results discussed in the previous section could be interpreted in several ways if the analysis were conducted in isolation. As historical
28 This interpretation is consistent with the nding by Kurosaki and Khan (2006) using the same panel dataset that education investment had high economic returns if associated with nonagricultural employment. With new opportunities for poverty reduction through human capital investment, the livelihood asset can be increased during adverse macroeconomic conditions. Households that moved out of poverty through human capital accumulation could then settle at the higher equilibrium in the S-shape curve in panel C of Figure 4. 29 The estimation results under alternative specications are available on request from the author.
semi-macro data are available encompassing the period both before and after the panel data collected, the information derived from these data can be utilized to narrow down the interpretation. First, with regard to the shape of the asset dynamics curve, the microeconometric results for land and livestock suggest an absence of multiple equilibriums. This is further supported by the historical nding of an absence of persistent declines in livestock population after independence. When human capital was included, the results were ambiguous due to statistical insignicance of multiple equilibriums. Therefore, without other indirect evidence, this paper concludes that no evidence is found for multiple equilibriums. If historical data were available on the average level and distribution of education at the district level, it would be possible to provide further support or refutation to this tentative conclusion. As speculated in the previous subsection, my eld impression is that it is possible that multiple equilibriums existed in recent years when human capital became the key component of the livelihood asset. This possibility could be investigated in further research with other datasets. Second, regarding the response of household assets to shocks, the results of the microeconometric analysis reveal how livestock declined rapidly in all villages when these villages were hit by macroeconomic shocks. Since there was no natural disaster that caused the death of livestock during 19961999, the negative coefcients cannot be interpreted in the same way as for 1920/1921 when livestock animals died due to droughts. Therefore, this is evidence that livestock had been used as a buffer against negative aggregate shocks during the 1990s. The change in the livestock portfolio over the century (Figures 2 and 3) is worth attention in this regard. Since draft animals were an indispensable part of crop production in the old days, it was difcult for farmers to reduce the stock even in difcult years. In contrast, the number of milk animals can be reduced more easily and the increasing share of such animals in the livestock portfolio of households has facilitated the effectiveness of livestock as a buffer. Thus, the microeconometric results in Table 2 can be better understood with the help of long-term historical evidence. Furthermore, the microeconometric results regarding household idiosyncratic shocks showed that assets declined on average, but the decline was not statistically signicant. Among the shocks, the adverse impact of other shocks such as unexpected deaths and funerals and discontinuation of remittances from family members living outside the village was statistically signicant in the land regression. The results could be interpreted as showing heterogeneity among villagers and among the type of shocks in terms of the extent of insurance against idiosyncratic shocks. This interpretation is indirectly supported by the historical analysis if the cross-sectional difference is compared with changes over time. The historical analysis showed that the livestock population at the district level became less responsive to crop shocks in more recent years with the development of infrastructure, agricultural technology, and intertemporal resource allocation opportunities.
B.
Reinterpreting the Historical Patterns
It was speculated in Section IV that: (i) pre-independence livestock dynamics were consistent with both the asset poverty trap hypothesis and the buffer stock hypothesis, while the post-independence livestock dynamics were more consistent with the buffer stock hypothesis; and (ii) the contrast could be attributable to different levels of infrastructure, formal credit facilities, and agricultural technology. The microeconometric ndings in Section V are broadly supportive of these conjectures. They reveal that during the late 1990s assets were indeed used as a buffer. The between-village contrast also supports the contention that as an economy develops, the function of assets in smoothing consumption against shocks strengthens. As the buffer stock hypothesis was supported even for the least developed village in the microeconometric analysis, it appears more likely that asset dynamics during the pre-independence period was also more consistent with the buffer stock hypothesis than with the poverty trap hypothesis. As a nal remark on combining the two types of analysis, let us consider a prediction regarding the district livestock population, which is theoretically derived from the conclusion that the asset dynamics of livestock follow the buffer stock hypothesis. As discussed in Section II, the theoretical prediction is that the decline in the livestock population witnessed at the micro level during the 19961999 period should be temporary. Panel B of Figure 3 indeed supports thisthe change in the livestock population from 1996 to 2006 is connected with the change from 1986 to 1996 without a signicant discontinuity in the growth rates.30 However, it must be noted that to infer district-level dynamics from microlevel analysis, one needs to specify how initial assets are distributed across households and villages over the entire district and how livestock markets behave in response to district-level changes. The previous prediction is based on a simple assumption that the initial livestock distribution is the multiple of the three villages analyzed and that there is no market equilibrium effect in the livestock markets. Further research is necessary to replace this assumption by a numerical model based on hard data and an appropriate microeconomic model.
VII. Conclusion
This paper analyzed asset dynamics held by low-income households in the NWFP area in Pakistan over a period from 1902 to 2011. I rst investigated the longrun data at the district and province levels. The results showed how the population of
30 The following are the annual growth rates of the livestock population in Panel B of Figure 3 (the rst number shows the growth from 1986 to 1996 and the second from 1996 to 2006): bullocks as draft animals (6.4% and 2.0%), adult cows in milk (+1.2% and +2.7%); and adult she-buffaloes in milk (+7.5% and +4.3%).
livestockthe major asset of rural householdsdeclined with crop shocks due to droughts, but did not respond much to the Great Depression. The decline in livestock due to droughts was persistent. In the post-independence period, crop agriculture continued to be vulnerable to natural disasters, although less substantially so, while the response of livestock to such shocks was indiscernible in district-level data. I then analyzed a panel dataset collected from approximately 300 households in three villages in the NWFP during the late 1990s. Results showed that the dynamics of household landholding and livestock were associated with a single equilibrium. When human capital was included, however, the dynamics curve changed its shape, though this was not sufciently nonlinear to produce statistically signicant multiple equilibriums. On the other hand, the response of household assets to village-level and household-level shocks showed several interesting patternslivestock assets were depleted widely when the village economy was affected by macroeconomic stagnation, land assets were depleted only in a village with inferior access to markets, and idiosyncratic agricultural and off-farm work shocks did not substantially affect household-level asset dynamics. To understand these patterns revealed from historical and contemporary analyses, the paper suggested the possibility that the contrast could be attributable to the different levels of infrastructure, formal credit facilities, and agricultural technology. Considering the long-run historical data, the household panel data during the 1990s appear to show that there had been an improvement in access to consumption smoothing measures such as asset sale markets, credit institutions, and reciprocitybased transfers. However, the improvement was not homogenous, leaving pockets of villages and households with inferior access. In this regard, the role of livestock as liquid assets was found to be important in smoothing consumption while the role of less liquid assets, particularly land, was more limited. The reduction in the number of draft animals in the livestock portfolio in the long-term, and their replacement by milk animals, facilitated the effectiveness of livestock as a buffer against negative shocks. On the other hand, throughout the period since the early 20th century, nancial markets existed in cities in the NWFP, and villagers had a network of credit transactions. However, the actual use of modern nancial markets and formal credit institutions did not prevail widely in the early stage of development. The highly unequal distribution of land in Pakistan could have accentuated the disparity, as land is often used as collateral in formal credit transactions. These interpretations imply that improving the intertemporal smoothing ability of households through the development of assets and credit markets is key to mitigating the adverse effects of natural disasters. It is also expected that investment in infrastructure such as for transport and communication could contribute to higher resilience against natural disasters as it would facilitate the movement of labor and improve the level of efciency of risk sharing and credit transactions.
It must be noted that these interpretations and policy implications are merely speculations. A limitation of this paper is that the attempt to demonstrate the complementarity of combining historical and contemporary analyses is incomplete. Because of data limitations, the paper was unable to investigate the asset dynamics during the pre-independence period in a microeconometric way. This is left for further research. Nevertheless, as a policy-oriented research, this paper shows the potential benet of empirical analyses that combine both contemporary and historical information. Economic development is, by denition, a long-term process. A microeconometric test of a particular structure of incomplete markets needs to be aligned with the historical context.
References
Adato, Michelle, Michael Carter, and Julian May. 2006. Exploring Poverty Traps and Social Exclusion in South Africa Using Qualitative and Quantitative Data. Journal of Development Studies 42(2): 226247. Ahmed, Akbar. 1980. Pukhtun Economy and Society: Traditional Structure and Economic Development in a Tribal Society. London: Routledge & Kegan Paul. Becketti, Sean, William Gould, Lee Lillard, and Finis Welch. 1988. The Panel Study of Income Dynamics after Fourteen Years: An Evaluation. Journal of Labor Economics 6(4): 472492. Carter, Michael, and Christopher Barrett. 2006. The Economics of Poverty Traps and Persistent Poverty: An Asset-Based Approach. Journal of Development Studies 42(2): 178199. Carter, Michael, Peter Little, Tewodaj Mogues, and Workneh Negatu. 2007. Poverty Traps and Natural Disasters in Ethiopia and Honduras. World Development 35(5): 835856. Deaton, Angus. 1992. Understanding Consumption. Oxford: Clarendon Press. Dercon, Stefan, ed. 2005. Insurance against Poverty. Oxford: Oxford University Press. Dr` eze, Jean, and Amartya Sen. 1995. India: Economic Development and Social Opportunity. Delhi: Oxford University Press. Dutta, Indranil, James Foster, and Ajit Mishra. 2010. On Measuring Vulnerability to Poverty. IED Discussion Paper no. 194. Boston: Institute for Economic Development. Fafchamps, Marcel. 2003. Rural Poverty, Risk and Development. Cheltenham: Edward Elgar. Fujie, Takeshi, and Tetsushi Senda. 2011. How Do Farm Households Cope with Aggregate Shocks? Evidence from the Great Depression in Prewar Japan. Journal of Rural Economics 83(1): 1527. In Japanese. Government of Pakistan. 2012. Pakistan Economic Survey 201112. Islamabad: Economic Advisers Wing. Hirashima, S. 2008. The Land Market in Development: A Case Study of Punjab in Pakistan and India. Economic and Political Weekly 43(42): 4147. Khalid, A. K. 1998. The Agrarian History of Pakistan. Lahore: Allied Press Ltd. Kurosaki, Takashi. 1995. Risk and Insurance in a Household Economy: Role of Livestock in Mixed Farming in Pakistan. Developing Economies 33(4): 464485. . 2003. Specialization and Diversication in Agricultural Transformation: The Case of West Punjab, 19031992. American Journal of Agricultural Economics 85(2): 372386. . 2006. Consumption Vulnerability to Risk in Rural Pakistan. Journal of Development Studies 42(1): 7089.
108 ASIAN DEVELOPMENT REVIEW . 2011. Compilation of Agricultural Production Data in Areas Currently in India, Pakistan, and Bangladesh from 1901/02 to 2001/02. PRIMCED Discussion Paper No. 6. Tokyo: Hitotsubashi University. Kurosaki, Takashi, and Anwar Hussain. 1999. Poverty, Risk, and Human Capital in the Rural NorthWest Frontier Province, Pakistan. IER Discussion Paper Series B No. 24. Tokyo: Hitotsubashi University. Kurosaki, Takashi, and Humayun Khan. 2006. Human Capital, Productivity, and Stratication in Rural Pakistan. Review of Development Economics 10(1): 116134. . 2011. Floods, Relief Aid, and Household Resilience in Rural Pakistan: Findings from a Pilot Survey in Khyber Pakhtunkhwa. The Review of Agrarian Studies 1(2): 79107. Kusadokoro, Motoi, Takeshi Maru, and Masanori Takashima. 2012. Asset Accumulation Behavior of Rural Households in the Reconstruction Period following the Showa Depression: A Panel Data Analysis Using the Third Period MAF Survey of Farm Household Economy. Paper Presented at the Asian Historical Economic Conference, 1315 September, Hitotsubashi University. Available: http://ahec2012.org/papers/ Ligon, Ethan, and Laura Schechter. 2003. Measuring Vulnerability. Economic Journal 113: C95C102. Malik, Sohail. 1999. Poverty and Rural Credit: The Case of Pakistan. Islamabad: Pakistan Institute of Development Economics. McKay, Andrew, and Emilie Perge. 2011. How Strong is the Evidence for the Existence of Poverty Traps? A Multi-country Assessment. CPRC Working Paper No. 180. London, UK: Chronic Poverty Research Centre. Miura, Ken, Hiromitsu Kanno, and Takeshi Sakurai. 2012. Rainfall Shock and Livestock Transactions in Rural Zambia: An Empirical Examination Using High Frequency Panel Data. Paper presented at the Asian Historical Economic Conference, 1315 September, Hitotsubashi University. Available: http://ahec2012.org/papers Mogues, Tewodaj. 2011. Shocks and Asset Dynamics in Ethiopia. Economic Development and Cultural Change 60(1):91120. Naschold, Felix. 2005. Identifying Asset Poverty Thresholds: New Methods with an Application to Pakistan and Ethiopia. Paper presented at the American Agricultural Economics Meeting. 2427 July. Providence, Rhode Island.
Appendix Table: Estimation of the Livelihood Asset through Regression Analysis

Dependent Variable: Welfare Ratio 1996 1999
Summary Regression Summary Regression Statistics Results Statistics Results Village-specic Intercept Village A Village B Village C 0.201 (0.402) 0.270 (0.444) 0.529 (0.500) 1.003 (0.115) 1.104 (0.117) 1.197 (0.123) 0.206 (0.405) 0.251 (0.434) 0.544 (0.499) 0.937 (0.131) 0.948 (0.120) 1.123 (0.121)
Continued.

Appendix Table: Continued.
Dependent Variable: Welfare Ratio 1996 1999
Summary Regression Summary Regression Statistics Results Statistics Results Household Wealth Characteristics Number of household members Ratio of females in the household Ratio of dependent members in the household Dummy for a female-headed household Age of the household head Literacy rate of working-age adults Per capita value of household assets such as agricultural machinery, transport equipment, durable goods, etc. (PRs1,000 in 1996 prices) Per capita outstanding credit including informal lending to others (PRs1,000 in 1996 prices) Per capita value of farmland owned by the household (PRs100,000 in 1996 prices) Per capita value of livestock owned by the household (PRs1,000 in 1996 prices) Dummy for households with workers employed in nonagricultural employment on a permanent basis Dummy for households that regularly receive remittances from family members living separately Mean of the dependent variable/R-squared Standard deviation of the dependent variable/ F-stat for zero slopes Number of observations

11.843 (6.854) 0.484 (0.129) 0.488 (0.184) 0.005 (0.072) 52.259 (16.411) 0.278 (0.257) 2.494 (6.259) 0.967 (5.778) 0.488 (1.500) 1.531 (2.656) 0.551 (0.498) 0.086 (0.281) 0.957 (0.473) 354
0.012 (0.006) 0.017 (0.143) 0.481 (0.136) 0.261 (0.121) 0.001 (0.001) 0.349 (0.101) 0.015 (0.007) 0.017 (0.003) 0.081 (0.020) 0.039 (0.011) 0.013 (0.047) 0.164 (0.097) 0.893 14.28 354
12.222 (7.606) 0.483 (0.126) 0.492 (0.189) 0.014 (0.118) 54.735 (15.809) 0.323 (0.258) 0.597 (0.963) 0.719 (3.605) 0.366 (0.919) 0.934 (1.297) 0.583 (0.494) 0.184 (0.388) 0.994 (0.512) 351
0.006 (0.005) 0.070 (0.182) 0.644 (0.133) 0.007 (0.128) 0.001 (0.001) 0.176 (0.092) 0.219 (0.043) 0.013 (0.013) 0.120 (0.024) 0.060 (0.034) 0.089 (0.043) 0.127 (0.075) 0.902 15.71 351
= signicance at the 1% level, = signicance at the 5% level, = signicance at the 10% level, PRs = Pakistan rupees. Notes: 1. In the summary statistics column, weighted means are reported with standard deviations given in parentheses. 2. In the regression results column, gures in parentheses are Huber-White robust standard errors. 3. WLS regression with village xed effects (no intercept) is employed. 4. Weights are the same as those described in Table 2. 5. In the 1996 regression, the number of observations was 354 as one observation had been excluded due to nonavailability of consumption data. In the 1999 regression, the number of observations was 351 as split and replacement households had been included. Source: Authors computations based on 19961999 panel data described in the text.
Inequality of Human Opportunities in Developing Asia

HYUN H. SON
This paper analyzes the equity of opportunity in basic education and infrastructure services in seven developing countries, Bangladesh, Bhutan, Indonesia, Pakistan, the Philippines, Sri Lanka, and Viet Nam. The analysis applies a method developed by the World Bank called the Human Opportunity Index, which measures the total contribution of individual socioeconomic and demographic circumstances to inequality of opportunity in accessing basic services. The new and major contribution of the paper, however, is the development of a methodology that quanties the relative contribution of each circumstance variable to the inequality of opportunity. This contribution is crucial in identifying which underlying inequalities matter mostwhich can have important policy implications, for instance, in terms of developing better-targeted interventions. Results of the empirical analysis indicate that more needs to be done to improve the distribution of economic benets. Opportunities to access basic education and infrastructure services in the seven countries vary widely in terms of availability and distribution. The study also nds that inequality of opportunity is driven mainly by per capita household expenditure. This suggests that household poverty plays a crucial role in determining equitable access to basic services. Keywords: inequality, education, health, basic infrastructure, human opportunity, poverty, access to water and sanitation, access to electricity, urbanization, agglomeration, developing Asia JEL codes: I14, I24, I25, I38, O15, O57
I. Introduction
Inequality remains a persistent challenge in many economies today. In Asia and the Pacic, inequality has risen over the last decade despite growth rates that have lowered poverty incidence (ADB 2007a). In the 16 countries that make up developing Asia, the Gini coefcient increased from 46.8 in 1993 to 52.4 in 2003 (ADB 2007a). Inequality is usually measured in terms of income or consumption, but the concept is now being extended to cover many other standard of living dimensions such as inequality of outcomes in health, education, and basic infrastructure, among
The author is a Principal Evaluation Specialist at the Independent Evaluation Department of the ADB. She would like to acknowledge insightful comments from two anonymous referees, who helped improve an earlier draft of the paper.
INEQUALITY OF HUMAN OPPORTUNITIES IN DEVELOPING ASIA 111
others. In a recent study, Zhang and Kanbur (2005) and Tandon and Zhuang (2007) have demonstrated that disparities in health outcomes in the Peoples Republic of China (PRC) have gotten worse. Although any societys ultimate objective is to eliminate or reduce inequality of outcomes, the 2006 World Development Report has argued that it is not appropriate to focus on this alone when assessing the fairness of a social system. Inequality of opportunity, not of outcome, should inform the design of public policy. According to this view, public policies need not necessarily eliminate or reduce all outcome inequalities but may instead focus on reducing inequalities that arise from unequal opportunity. Thus, a just society is one that provides equal opportunity to all. Governments usually provide people opportunities in education, health, nutrition, security, and basic infrastructure. However, not all citizens can equally avail of these opportunities. For instance, many school-age children in developing countries are unable to attend school due to family circumstances. Similarly, many of those children have no access to clean water, electricity, and sanitary toilets. Measuring the inequality of opportunity in such basic services is therefore essential prior to designing policies aimed at providing universal provision of these basic opportunities. To measure the inequality of opportunity contributed by individual socioeconomic and demographic circumstances, the World Bank (2006) has developed the Human Opportunity Index (HOI). Inequality of opportunity due to differences in circumstances is considered unjust and should be of concern to society. When a child is unable to get proper education because her family belongs to a low social group, for instance, it is deemed as gross injustice. In its study on Latin America, the World Bank considered six circumstance variables: (i) urban or rural area, (ii) gender, (iii) number of siblings, (iv) parents education, (v) per capita income, and (vi) presence of parents. The number of circumstance variables included was determined by the availability of data in 19 Latin American countries. This paper seeks to measure for the rst time the inequality of opportunity in seven developing countries: Bangladesh, Bhutan, Indonesia, Pakistan, the Philippines, Sri Lanka, and Viet Nam.1 The HOI is measured for a set of opportunities related to education and basic infrastructureschool attendance among children aged 611 years for primary school and 1217 years for secondary, as well as access to safe water, electricity, and sanitation. The HOI measures the total contribution of individual socioeconomic and demographic circumstances to inequality of opportunity in accessing basic services.
1 Selection of the seven countries was based purely on the availability of data. Although these countries are not representative of developing countries in Asia, they are at different stages of economic development and thus provide a nice contrast. This is the rst study that has applied the World Banks method to Asian countries. The household surveys used in this study include: the Household Income and Expenditure Survey 2000 for Bangladesh; the Bhutan Living Standard Survey 2007; SUSENAS 2009 for Indonesia; the Household Integrated Economic Survey 20072008 for Pakistan; the Annual Poverty Indicator Survey 2002 for the Philippines; the Household Income and Expenditure Survey 20092010 for Sri Lanka; and the Household Living Standards Survey 2008 for Viet Nam.
The new and major contribution of this paper, however, is the development of a methodology that quanties the relative contribution of each circumstance variable to the inequality of opportunity. This contribution is crucial in identifying underlying inequalities that matter most, which can have important policy implications for instance in terms of developing better-targeted interventions. Results of the empirical analysis indicate that more needs to be done to improve the distribution of economic benets. Opportunities to access basic education and infrastructure services in the seven countries vary widely in terms of availability and distribution. The study also nds that inequality of opportunity is driven mainly by per capita household expenditure. This suggests that household poverty plays a crucial role in determining equitable access to basic services. This paper is outlined as follows. Section II briey outlines the HOI methodology. Section III discusses the method of quantifying the relative contribution of each of circumstance variable to inequality of opportunity. Section IV provides a crosscountry comparison of inequalities in opportunity in the seven developing countries considered in this study. Section V summarizes the major ndings emerging from the study and presents the corresponding policy implications.
II. Human Opportunity Index
Let us dene a variable z i which takes a value of 1 if the ith individual has access to an opportunity (such as education) and takes a value of 0 if the ith individual lacks access to the opportunity. It can be easily seen that E (z i ) = i = P (z i ), where i is the probability that the ith individual has access to a given opportunity. A distinction is made between circumstance and effort variables (Roemer 1998). Circumstance variables are exogenous variables in the sense that an individual has no control over them. Effort variables, meanwhile, reect an individuals efforts and capacity to innovate and take risk. Inequality caused by differences in effort is deemed acceptable, while inequality caused by circumstances is considered unjust and unacceptable, and should thus be reduced. The HOI measures the contribution of inequality of opportunities given the circumstance variables. Therefore, we estimate i by means of a logit model using a set of k circumstance variables xi 1 , xi 2 , . . . , xik . Accordingly, we have a logit model: i = e
k j =1
j xi j
k j =1
1+e
j xi j
(1)
This model can be estimated using the maximum likelihood method. i , the maximum likelihood estimate of i , gives the estimate of the probability of access to a given opportunity that is explained by the circumstance variables. Any measure of inequality of i will be the inequality of opportunity that is explained by the
circumstance variables. The World Bank (2006) uses the relative mean deviation dened as D= 1 2
n
wi | i |
i =1
(2)
where n is the number of sample households, wi is the population weight attached to the ith sample household, and is the proportion of the population with access to a given opportunity.2 Note also that may be called level or coverage.3 D measures the degree of inequality of opportunity that is explained by the individuals circumstances. As such, (1 D ) may be interpreted as equity of opportunity. D takes values between 0 and 1. D = 0 implies that every individual in a society enjoys the same opportunities, while D = 1 implies that only one person in the society enjoys all opportunities. The HOI is then dened as HOI = (1 D ) (3)
which is a composite index of two factors: (i) the level or coverage, and (ii) equity of opportunity.4 Since 0 D 1, HOI will always be less than or equal to , which means that there will always be loss of average opportunities available to the society because opportunities are not equally enjoyed by all members of the society. HOI can be interpreted as an inequality-adjusted coverage rate. The policymakers objective will be to maximize HOI, which can be achieved by enhancing total opportunity (coverage), increasing equity of opportunity (more equitably distributing opportunity), or increasing both coverage and equity.
III. Contribution of Individual Circumstance Variables
The relative mean deviation dened in equation (2) measures the total contribution of all circumstance variables to inequality of opportunity. Although it is useful to determine the total impact of all circumstance variables on inequality of opportunity, determining the impact of each circumstance variable would be more useful to policymakers. These individual contributions will identify the circumstance variables having the most impact on inequality of opportunity. In this section, we present a method of calculating the relative contributions of individual circumstance variables to the inequality of opportunity.
2 3
D is also referred to in the literature as the dissimilarity index, which is widely used in sociology. Note that is the mean of i across all individuals. 4 This methodology was developed by Paes de Barros et al. (2008).

i A variable yi = (1 is the ratio of the odds of z i = 1 against z i = 0. Thus, i ) the larger is yi , the greater are the odds that the ith person will have access to an opportunity. A special feature of the odds ratio is that, in utilizing equation (1), it can be written in natural logarithmic form as k
ln( yi ) =
j =1
j xi j
(4)
The maximum likelihood estimate of yi is then given by

k
i ) = ln( y
j =1
j xi j
(5)
j is the maximum likelihood estimate of j derived from the logit model in where i is the estimate of the ith persons odds ratio that is explained by equation (1), and y the circumstance variables. Since yi is a monotonically increasing function of i , there is one-to-one relationship between them. This implies that inequality of i will be equivalent i will be equivalent to to inequality of yi . This in turn suggests that inequality of y inequality of i which, as shown above, is also equal to the inequality of opportunity explained by the circumstance variables. We can thus measure the inequality of opi . portunity explained by the circumstance variables by measuring the inequality of y We may measure inequality of opportunity by any of the inequality measures that have been proposed in the literature. As discussed above, the World Bank (2006) used the relative mean deviation to measure inequality of opportunity. In this study, we use the log variance measure of inequality, which has an attractive feature of decomposability. Following Fields (2003), we take the variance of both sides of equation (5) to obtain
k
i )) = 2 (ln ( y
j =1
i )) j cov (xi j , ln ( y
(6)
which decomposes the inequality in opportunity (measured by the log variance) in terms of the contributions made by each of the individual circumstance variables.5 i )) gives the percentage contribution Dividing both sides of equation (6) by 2 (ln ( y of individual circumstance variables as
k
100% =
j =1
Sj
(7)
Equation (6) follows a straightforward application of calculating the variance of a sum of random variables.
where Sj = j cov (xi j , ln ( y i )) 100 i )) 2 (ln ( y (8)
is the percentage contribution of the jth circumstance variable to the total inequality of opportunity. The larger is S j , the greater is the contribution of the variation in the jth circumstance variable to total inequality of opportunity. The idea is similar to analysis of variance that is widely used in statistics to measure contributions of various factors to the total variance. Further, it should be noted that S j measures the net contribution of the jth circumstance variable after accounting for all interj actions between circumstance variables. S j will generally be positive because i )) are expected to be of the same sign. However, due to interacand cov (xi j , ln ( y tions among circumstance variables we may obtain a negative value of S j , which is largely expected to be statistically insignicant. In that case, the contribution of the jth circumstance variable may be deemed as neutral to total inequality. The decomposition presented in equation (7) is based on the log variance as a measure of inequality. This may appear to be a restricted result but is in fact not. Using the famous Shorrocks (1982) theorem, we can easily show that this result holds for a wide variety of inequality measures including the Gini index, the Atkinson index, the generalized entropy family, and the coefcient of variation.
IV. Empirical Analysis
In this section, the methodologies outlined in the previous sections are applied to seven developing countries in Asia: Bangladesh, Bhutan, Indonesia, Pakistan, the Philippines, Sri Lanka, and Viet Nam. The section provides analysis of the inequality of opportunity related to basic education and infrastructure. There are ve outcome variables used in our analysis: (i) primary school attendance among children aged 611 years, (ii) secondary school attendance among children aged 1217 years, (iii) access to safe water, (iv) access to electricity, and (v) access to sanitation.6 Similarly, we used a set of circumstance variables required to estimate the D-Index and the HOI. This comprised (i) gender, (ii) location of household (urban or rural area), (iii) education of household head, (iv) per capita household expenditure, (v) age of household head, (vi) gender of household head, and (vii) household size. Circumstances, as used here, consist of personal or family
6 School attendance for children aged 611 years and 1217 years was not restricted to primary levels and secondary education levels, respectively (i.e., some in the 611 year age group may be in secondary levels). This denition was followed uniformly across all countries. Similarly, same denitions were adopted for access to safe water, access to electricity, and access to sanitation across all seven countries studied.
116 ASIAN DEVELOPMENT REVIEW Table 1. Inequality of Opportunity in Primary Education, 611 Years
Country Indonesia Philippines Viet Nam Bangladesh Bhutan Sri Lanka Pakistan Survey Year 2009 2002 2008 2000 2007 20092010 20072008 Average Opportunity 94.29 93.92 96.31 75.59 83.05 99.39 74.59 D-Index 0.92 1.80 1.29 3.53 4.98 0.12 8.71 Human Opportunity Index 93.42 92.22 95.07 72.92 78.91 99.27 68.09
Source: Authors calculations based on household surveys.
socioeconomic and demographic characteristics over which an individual has no direct control. These seven circumstance variables are available in household datasets for the seven countries selected for the study.7
A. Inequality of Opportunity in Basic Education
The distribution of opportunity for children to access basic primary education is highly variable across countries in Asia. As indicated by the high value of HOI in Table 1, the playing eld is level for primary-school-age children in Sri Lanka, where 99.27% of primary education services are available and equitably allocated. In contrast, only 68.09% of the basic services in Pakistan are available and distributed inequitably among children. Countries in Southeast Asia, such as Indonesia, the Philippines, and Viet Nam are moving towards universal access of basic primary education. For each of these countries, the estimated HOI is higher than 90%, suggesting that more than 90% of primary education services required for universal coverage are available and distributed equitably. Three countries are at the bottom of the ranking, with HOIs lower than 80%. These are Bangladesh, Bhutan, and Pakistan (Figure 1). Compared to their younger cohorts, children in the secondary-school-age group (1217 years old) in developing Asia are more likely to have lower levels of equitably allocated education services. The HOI for primary school attendance is far higher than the corresponding gure for secondary school attendance across the seven countries. As shown in Table 2 and Figure 2, the HOI for secondary education services ranges from a high of 84.49 for Sri Lanka to a low of 47.64 for Pakistan. These ndings suggest that countries in the region face greater challenges in equitably ensuring that all children aged 1217 attend school than ensuring that all children of primary school age attend school. This result could be expected because
7 For Bangladesh and the Philippines, relatively old datasets were used2000 and 2002, respectivelypossibly resulting in unfair comparisons for the two countries. That is, they may perform worse than they actually do as compared to countries with more recent data.

Figure 1. Human Opportunity Index for Primary Education for Selected DMCs
Sri Lanka Viet Nam Indonesia Philippines Bhutan Bangladesh Pakistan 0 10 20 30 40 50 60 70 80 90 100
Table 2. Inequality of Opportunity in Secondary Education, 1217 Years

Figure 2. Human Opportunity Index for Secondary Education for Selected DMCs
Sri Lanka Philippines Viet Nam Indonesia Bhutan Bangladesh Pakistan 0 10 20 30 40 50 60 70 80 90 100
the opportunity costs of sending children to school are higher at the secondary than at the primary level. This also implies that nancial incentives such as conditional cash transfer programs could be more effective in targeting older children if the main objective is to improve school enrollment.
118 ASIAN DEVELOPMENT REVIEW Table 3. Inequality of Opportunity in Access to Safe Water
Table 4. Inequality of Opportunity in Access to Electricity

B.
Inequality of Opportunity in Basic Infrastructure
Basic infrastructure services make signicant contributions to well-being. Basic services such as safe water and sanitation (e.g., ushing toilets) have a direct impact on health status and overall well-being. Access to services such as electricity helps households increase their productivity for income generation. A number of studies reveal that a households access to basic infrastructure services is highly and signicantly correlated with a lower probability of being poor. Compared to basic education services, results for the HOIs suggest that Asia faces a greater challenge in providing basic infrastructure services. As presented in Tables 35, the HOIs for access to basic infrastructure services such as safe water, electricity, and sanitation show lower values for all countries and higher dispersion across countries than those for access to basic education services, highlighting the uneven rates of progress in expanding opportunities for basic infrastructure services in the region. As seen in Table 3, Bhutan takes the lead in the provision of access to safe water, with an HOI equal to 86.91. In contrast, Bangladesh and Viet Nam have HOIs lower than 20 for this service. In the area of electricity provision, Viet Nam and Sri Lanka lead with HOIs higher than 90, in contrast to an HOI of about 20 for Bangladesh (Table 4). In sanitation, three of the seven countries examined in this study have an HOI higher than 50, while Bangladesh and Bhutan have HOIs lower than 20 (Table 5). These ndings suggest that less than one out of ve people in

Table 5. Inequality of Opportunity in Access to Sanitation
Bangladesh have equal opportunity to live in households with access to safe water, electricity, and sanitation.
C. Do Circumstance Variables Matter for Inequality of Opportunity?
This section quanties the relative contribution of each of the seven circumstance variables to inequality of educational opportunity for both primary and secondary levels, as well as inequality of opportunity to access basic infrastructure such as safe water, electricity, and sanitation. As pointed out, different circumstance variables may interact with each other. As such, the percentage contributions calculated in this section are the net contributions after taking into account all interactions among circumstance variables. For primary education, the most important circumstance variable that inuences whether or not a child has fair access to education opportunities is per capita household expenditure. Its contribution to inequality of opportunity for primary education ranges from 60.6% in Pakistan to more than 95% in countries such as Bangladesh and the Philippines (Table 6). This suggests that overall standards of living of households play a major role in inuencing the ability of a child to improve his or her situation over time and achieve intergenerational mobility through education. With regard to equal opportunity for primary education, variables such as education of household head, urban or rural location (where a child lives), and household size are also important circumstance conditions in Asia. In Pakistan, the household heads level of formal education accounts for more than 20% of the inequality of educational opportunity for children of primary school age (611 years). This suggests a direct association between the household heads education and his perception of education. A recent study by Lodhi, Tsegai, and Gerber (2011) found that parents with less education in Pakistan are more likely to view education as a trivial factor to future income. Parents with these perceptions are signicantly more likely to send their children to madrassahs8 or let them nd paid work.
8 Madrassah refers to any type of religious school or college for the study of the Islamic religion, though this may not be the only subject studied.
Table 6. Percent Contribution of Circumstance Variables to Inequality of Opportunity for Primary Education, 611 Years
Country 12.4 1.3 10.5 2.7 16.1 5.1 10.5 0.0 0.2 0.4 0.7 0.9 2.1 2.4 74.6 95.2 72.8 97.1 77.6 76.4 60.6 4.5 0.6 0.0 0.0 1.1 3.8 0.0
Survey Year 3.4 0.3 6.3 2.1 6.8 0.0 20.5
Gender
Area of Residence (urban/rural) Per Capita Household Expenditure Age of Household Head Gender of Household Head
Education Level of Household Head
Household Size 4.0 1.7 10.7 0.0 0.5 12.1 0.2
Indonesia Philippines Viet Nam Bangladesh Bhutan Sri Lanka Pakistan
2009 2002 2008 2000 2007 20092010 20072008
0.9 0.8 0.1 2.8 0.2 0.4 5.8
= estimated coefcient found to be statistically signicant at the 5% level in the logit regression model of the probability of school attendance among the primaryschool-age children, 611 years. Source: Authors calculations based on household surveys.
Similarly, location circumstance (i.e., urbanrural residence) substantially contributes to inequality of opportunity for primary education in countries like Bhutan (16.1%) and Indonesia (12.4%). In addition, a signicant proportion of children aged 611 years old in Sri Lanka and Viet Nam are often deprived of basic opportunities to help them gain access to education due to large household size, which accounts for 12.1% and 10.7%, respectively, of inequality of opportunity for primary education. Controlling for other factors such as household expenditure and parents education, households with more members tend to invest less in education of school-age children (Dang and Rogers 2008). Results in Table 6 suggest that circumstance variables such as age and gender of household head have little inuence on whether or not a child of primary school age has fair access to education opportunities. In terms of school attendance for children aged 1217, the inequality of educational opportunity is also driven mainly by their per capita household expenditure. Table 7 shows that per capita household expenditure yields a higher level of contribution to inequality of opportunity for secondary education compared with the other six circumstance variables in the seven selected countries. The contribution of household expenditure to inequality of opportunity for secondary education ranges from 61% in Pakistan at the minimum to 96% in Sri Lanka at the maximum. Apart from per capita household expenditure, the gender of the child, urbanrural residence, and education of household head make a sizable contribution to the inequality of educational opportunity for children of secondary school age. In Bangladesh, a childs gender accounts for 20.8% of the inequality of opportunity for secondary education and about 10% in Pakistan. These ndings call for strategic government policies that could redistribute available education services toward female children to help achieve equality of opportunity in secondary education. In Bangladesh, a major hindrance to girls attendance in secondary school is early marriage and fertility, prompting the government to introduce the highly successful Girls Stipend Program aimed at encouraging girls to continue their schooling (Raynor and Wesson 2006). However, despite recent increases in enrollment, girls still face inequity in achieving education outcomes especially in secondary school (Hossain and Zeitlyn 2010). A similar story of gender bias can also be found in Pakistan. For children of secondary school age, households exhibit a pro-male bias both in the decision to enroll children and amount to spend on education conditional on enrollment, while for the primary-school-age group the bias is only in the decision to enroll (Aslam and Kingdon 2008). The results in Table 7 also show that the ruralurban divide in terms of residence affects whether or not the child has access to opportunities for secondary education. The contribution of this location circumstance is particularly prominent for the Bhutan case where urbanrural residence accounts for 42.4% of the total inequality of opportunity. In Bhutan, urbanrural residence is the next most important
Table 7. Percent Contribution of Circumstance Variables to Inequality of Opportunity for Secondary Education, 1217 Years
Country 11.6 2.4 6.6 2.6 42.4 0.8 5.2 69.1 90.7 65.7 76.9 54.7 96.0 61.0 0.4 0.3 1.2 0.4 4.1 0.6 0.2 1.5 0.5 0.4 2.5 1.4 0.7 1.2
Survey Year 17.2 0.0 11.0 0.5 4.5 0.0 24.0
Gender
Area of Residence (urban/rural) Per Capita Household Expenditure Age of Household Head Gender of Household Head
Education Level of Household Head
Household Size 0.1 0.6 12.0 2.5 1.6 0.6 1.1
Indonesia Philippines Viet Nam Bangladesh Bhutan Sri Lanka Pakistan
2009 2002 2008 2000 2007 20092010 20072008
0.2 6.4 4.0 20.8 3.4 1.3 9.5
= estimated coefcient found to be statistically signicant at the 5% level in the logit regression model of the probability of school attendance among the secondaryschool-age children, 1217 years. Source: Authors calculations based on household surveys.

Table 8. Percent Contribution of Circumstance Variables to Inequality of Opportunity for Access to Safe Water
Area of Education Residence Per Capita Age of Gender of Level of (urban/ Household Household Household Household Household rural) Expenditure Head Head Head Size 67.7 8.5 82.8 82.0 55.6 89.0 70.8 29.5 89.0 11.9 17.5 32.4 9.1 19.0 0.7 2.6 2.5 0.4 0.5 0.2 0.0 0.3 0.6 2.6 0.0 0.2 0.2 0.0 0.5 0.5 0.1 0.1 13.8 0.0 10.2 1.3 1.2 0.1 0.1 2.1 1.5 0.1
Country
Survey Year
Indonesia 2009 Philippines 2002 Viet Nam 2008 Bangladesh 2000 Bhutan 2007 Sri Lanka 20092010 Pakistan 20072008
= estimated coefcient found to be statistically signicant at the 5% level in the logit regression model of the probability of having access to safe water. Source: Authors calculations based on household surveys.
circumstance variable after per capita household expenditure (54.7%). Children living in rural areas in Bhutan have limited access to schools due to their remote and mountainous location. According to the Ministry of Education of Bhutan (2004), a continuing problem is the tendency for children to drop out of schools if these are distant, and the unwillingness of teachers to be assigned to remote areas. A study conducted by the World Bank (2006) found that teachers are a critical constraint to improving quality of education in Bhutan. In particular, it is more difcult to recruit and motivate teachers for deployment to rural and remote areas. Furthermore, the results also reveal that the education level of the household head has a signicant inuence on whether or not a child of secondary school age has fair access to education opportunities. More importantly, the relative contribution of parental education to inequality of opportunity among children is far higher for secondary education than for its primary level counterpart: its contribution jumps to 17.2% for secondary school from 3.4% for primary school in Indonesia, and to 11% for secondary school from 6.3% for primary school in Viet Nam. As indicated by Table 7, other circumstance variables such as the age and gender of the household head and household size seem to have relatively negligible or statistically insignicant effects on the inequality of opportunity for secondary education. For access to safe water, electricity, and sanitation, the inequality of opportunity is driven mainly by per capita household expenditure and where an individual lives (urban or rural residence). As can be seen from Tables 810, location circumstance dominates in six out of seven countries in the case of safe water access, while per capita household expenditure is the most important circumstance in ve countries in the case of access to electricity and sanitation. In the case of water and sanitation, access is generally higher in urban areas than rural areas (WHO and UNICEF 2010). This phenomenon can also be observed
124 ASIAN DEVELOPMENT REVIEW Table 9. Percent Contribution of Circumstance Variables to Inequality of Opportunity for Access to Sanitation
Country
Survey Year
= estimated coefcient found to be statistically signicant at 5% level in the logit regression model of the probability of having access to sanitation. Source: Authors calculations based on household surveys.
Table 10. Percent Contribution of Circumstance Variables to Inequality of Opportunity for Access to Electricity
Country
Survey Year
= estimated coefcient found to be statistically signicant at the 5% level in the logit regression model of the probability of having access to electricity. Source: Authors calculations based on household surveys.
in Tables 8 and 9. In rural areas, the main challenge deals with the relatively higher cost of building water and sanitation infrastructure as well as the presence of rural poverty. Given this, rural areas often lack an enabling environment that encourages public or private investment in water services leading to low provision of these services (WHO and UNICEF 2010). This is a particular problem in South Asia where there is low overall public or private investment in such infrastructure particularly in Bangladesh, India, and Pakistan (WaterAid 2011). Moreover, even if investments are made in these countries, poor maintenance in rural areas still persists due to poor planning and lack of support. Even in Sri Lanka, which is on track to meet its Millennium Development Goal commitments on water and sanitation, rural areas are relatively underserved because the National Water Supply and Drainage
Board has concentrated its efforts on providing services to densely populated areas (ADB 2007b). On top of nancial constraints to providing water and sanitation services to rural areas, the perceptions and behavior of people in rural areas relating to water use and sanitation also pose challenges. Many rural households do not see the need to invest in tap water or sanitary toilets in their households since there are free options available. This leads to low demand and further decreases the nancial viability of such projects. Such behavior has been observed in Indonesia where low consumer demand and community acceptance for water and sanitation services in rural areas are deemed important constraints to investment (AusAID 2009, Yuerlita and Saptomo 2008). Similarly, lack of community participation in water and sanitation projects and poor hygienic behaviors in rural areas, which contribute to low demand, are seen as constraints to improving health and sanitation outcomes in Bhutan (Collett 2010). In Viet Nam, this problem is exacerbated by the decentralized structure of delivering water and sanitation services. Households and communes in Viet Nam are expected to pay for the construction, use, and maintenance of these structures. However, in rural areas, there is less interest among households to invest in such costly infrastructure (Sijbesma, Truong, and Devine 2010). In the Philippines, household poverty is a more important constraint to accessing water and sanitation services than residing in a rural area. As seen in our results in Tables 8 and 9, per capita household expenditure is the main contributor to inequality of opportunity to access water and sanitation services in the Philippines. Although the rural population in general still has less access to safe water or sanitation services than urban dwellers, poor people in rural and urban areas suffer the most deprivation and thus tend to bear higher burdens of disease or economic costs arising from unsafe water and inadequate sanitation. Thus, investments in water and sanitation in the country should be focused on rural areas and urban slums. As shown in Table 10, the inequality of opportunity for access to electricity is largely dependent on two circumstance variables: location (i.e., whether living in urban or rural areas) and economic status (as measured by per capita household expenditure). The rst circumstance is straightforward to explainthere are high costs associated with building an electricity grid in rural areas (World Bank 2010). As such, people living in rural areas are expected to be less likely to have access to electricity than their urban counterparts. However, this is not the entire story, as there are other constraints to achieving universal electrication in rural areas. In the case of Viet Nam, early attempts at rural electrication were hampered by inefcient coordination and a lack of regulatory framework (World Bank 2010). The high cost of building a rural grid and lack of central coordination also hampered electrication in Sri Lanka. In response to this, the government adopted a decentralized approach by encouraging off-grid electrication such as the use of solar panels (Independent Evaluation Group 2008).
Particular problems for rural electrication in Bhutan and Indonesia are their terrain, remoteness, and scattered settlements. Indonesias many islands are so sparsely populated that electrifying them is not nancially viable (World Bank 2005). In fact, it is unlikely that the Indonesian State Electricity Company will achieve its electrication targets outside the islands of Java and Bali. In Bhutan, a major challenge is the ruggedness and remoteness of the mountainous terrain, compounded by the fact that the rural population is scattered in small settlements (Kumar 2011). Similarly, community remoteness is an important factor that explains a lack of access to electricity in Pakistan (Mirza and Kemp 2011). Thus, even rich households in rural areas can be considered living in energy poverty. On the other hand, in Bangladesh and the Philippines, it is household-level constraintsi.e., povertythat is a major constraint to electrifying rural areas (World Bank 2010). Poor rural households are unlikely to be able to pay for connection fees or electricity consumption, which in turn makes it less nancially viable to invest in rural electrication. However, the recent success of Bangladesh shows the importance of central planning to map out subsidies and investments in rural electrication as well as the need to provide rural households with nancial assistance through cooperatives (Barnes 2007). As for the Philippines, household poverty is a particularly binding constraint to access to electricity because of high costs as conrmed by studies showing that the country has among the highest electricity rates in Asia due to inefciencies in energy production and transmission (Department of Energy 2008, Woodhouse 2005).
V. Conclusions
Inequality has become a major item on the development agenda in recent years. After decades of rapid economic growth around the world, economic gains have been threatened by the global nancial crisis of 2008 and the ongoing eurozone crisis. While economic theory has always maintained that growth is a necessary but not sufcient condition for improving standards of living, the recent economic crises have reinforced this view even in developed countries. Concepts such as equity, fairness, and justice in the distribution of economic benets are no longer in the realm of philosophers and theorists. Rather, they are now in the forefront of policy design and economic reform in both developed and developing countries. This study is concerned with analyzing the equity of distribution of opportunity for basic services in education and infrastructure. The analysis was carried out using a methodology introduced by the World Bank called the Human Opportunity Index (HOI), which is the product of average opportunity and equity of opportunity. The HOI examines both the coverage and distribution of opportunity in an outcome variable such as school attendance, access to safe water, access to electricity, or access to sanitation. The methodology was applied empirically using available
household data from Bangladesh, Bhutan, Indonesia, Pakistan, the Philippines, Sri Lanka, and Viet Nam. The HOI measures the total contribution of all circumstance variables to inequality of opportunity. From the perspective of policymakers, determining the impact of individual circumstance variables would be more useful since these individual contributions will help to identify circumstance variables that have the most impact on inequality of opportunity. This study presents a method of quantifying the relative contributions of individual circumstance variables to the inequality of opportunity. The new methodology introduced in this study would be helpful in analyzing binding constraints to providing equitable opportunities across countries. Among the circumstance variables examined, household expenditure and the location of residence in rural or urban areas were found to be crucial in inuencing access to education and basic infrastructure. Household povertyas manifested in the contribution of per capita household expenditure in the inequality of opportunityplays an extremely signicant role in determining equitable access. Household poverty denes the ability of households to pay for and access these services. This study showed a signicant correlation between household resources and the demand for education and infrastructure services. Thus, policymakers may opt to explore policies that address the demand side of the provision of education and infrastructure. Targeted subsidies or loans may be used to provide incentives to households to increase their demand for education and infrastructure. For education, cash transfers have been widely used to encourage school attendance (e.g., the Philippine governments conditional cash transfer program that provides stipends to poor households whose children meet the required school attendance rate, among others). Apart from household poverty, the location of residence in urban or rural areas also plays a crucial role in shaping access to infrastructure services. Serving as economic hubs, urban areas often have access to basic infrastructure services since adequate basic infrastructure is a fundamental requirement for encouraging private investments. Basic infrastructure also helps to ensure the proximity of various economic agents such as rms and workers. With reduced transport and transactions costs given the presence of basic infrastructure, rms opt to locate near each other in urban areas, giving rise to the economies of agglomeration which allow for reduced production costs through specialization and division of labor. Thus, there are greater incentives for urban communities to invest in basic infrastructure to take advantage of the benets of urbanization and agglomeration. In contrast, there is lower demand for basic infrastructure services in rural areas. These services such as water and sanitation are often provided through free and communal means. As such, there is little willingness to pay for these services among rural settlers, discouraging private investments in the provision of water, electricity, or sanitation services. Given the varying access to basic infrastructure between urban and rural communities, policies may be implemented to compensate
for the lack of demand and incentive to provide basic infrastructure services in rural areas. For instance, subsidies may be provided to rural communities to support the provision of safe water or reliable electricity supply. Opportunities to access basic education services in the seven countries vary widely. At the primary and secondary levels, Sri Lanka is a stellar example in equitably providing opportunities to access education, with school attendance among children aged 611 years reaching nearly 100% and an HOI of more than 99. The country also has the highest attendance rate among secondary-school-age children (86.38%) and an HOI of about 85. Sri Lankas educational achievements are remarkable considering that it does not have the highest per capita gross domestic product (GDP) among the seven countries, with the Philippines per capita GDP higher in 2007. In contrast, Sri Lankas neighbors, Bangladesh, Bhutan, and Pakistan, have yet to reach 90% attendance rates among primary-school-age children, while attendance rates for children of secondary school age are still below 60% in Bangladesh and Pakistan. HOIs in these South Asian countries are also among the lowest, indicating that they need to both improve overall access to basic education services and ensure that education opportunities among children are equally distributed, even to the poorest segments of the population. In Southeast Asia, access to and distribution of opportunities for basic education services have been impressive in Indonesia and Viet Nam in recent years, pointing to the effectiveness of their governments efforts to provide basic education for all. Meanwhile, there is also wide variation in the availability and distribution of opportunities to access basic infrastructure services such as safe water, electricity, and sanitation. Bhutan has the highest level of access to safe water (89.94%), Viet Nam has the highest level of electrication (97.19%), and Sri Lanka has the highest percentage of population living in homes with sanitation (94.19%). Our ndings highlight the uneven rates of progress in expanding opportunities to access quality infrastructure services in the region as compared with education opportunities. Unfortunately, for all basic infrastructure, Bangladesh shows the lowest levels of overall opportunities available and the distribution across population. Less than one-third of the population has household access to electricity, about one-fth has access to sanitation, and only a little more than one-twentieth has access to safe water. For all these facilities, only the richest 20% of the population has access rates of 50% or higher. These ndings suggest that these basic infrastructure services are nearly luxuries to most Bangladeshis. Clearly, a lot needs to be done to improve the distribution of economic benets in developing countries in Asia. While Bangladesh may be a particularly urgent case, all countries considered in this analysis need to bolster their efforts to improve access to basic education and infrastructure services, especially among the poor and marginalized groups. Sri Lankas achievements in equitably providing
basic education opportunities demonstrate the importance and possible effectiveness of public policy in achieving equity of opportunity, particularly in education. An important factor to consider in improving the delivery of basic services could be decentralization. Many countries have decentralized the delivery of services in education, water, electricity, and sanitation mainly to improve transparency, accountability, and responsiveness in the provision of these services. However, decentralization could also exacerbate existing inequalities across various local government units. Moreover, these local government units will have their own solvency and liquidity issues, which could affect their access to nancial services. This could cause highly inequitable distribution of services across districts, with afuent areas having much better services than poorer areas. As such, national governments will have to balance the costs and benets of decentralizing the responsibility of delivering these services. More rigorous research would be required to determine the best method to deliver public services more effectively and efciently.
References
Aslam, Monnaza, and Geeta Kingdon. 2008. Gender and Household Education Expenditure in Pakistan. Applied Economics 40: 25732591. Asian Development Bank. 2007a. Key Indicators 2007: Inequality in Asia. Manila: Asian Development Bank. . 2007b. Sri Lanka Country Assistance Program Evaluation: Water Supply and Sanitation Sector. Available at: http://www.adb.org/documents/evaluation/capes/sri/cape-sri-watersupply -sanitation-sector.pdf AusAID. 2009. Independent Evaluation of Australian Aid to Water Supply and Sanitation Service Delivery in East Timor and Indonesia. ODE Working Paper. Available at: http://www.ode.ausaid.gov.au/publications/documents/aus-water-supply-and-sanitation -indonesia-working-paper.pdf Barnes, Douglas, ed. 2007. The Challenge of Rural Electrication: Strategies for Developing Countries. Washington, DC: RFF Press. Collett, John. 2010. Thirty-ve Years of Searching for Answers to Rural Sanitation and Hygiene in Bhutan. Paper presented at South Asia Hygiene Practitioners Workshop. Available at: http://www.wsscc.org/sites/default/les/publications/3_collett_ruralsanitation_hygiene _bhutan_2010.pdf Dang, Hai-Anh, and Halsey Rogers. 2008. How to Interpret the Growing Phenomenon of Private Tutoring: Human Capital Deepening, Inequality Increasing, or Waste of Resources? Policy Research Working Paper Series 4530. The World Bank. Department of Energy. 2008. Philippine Energy Summit Summary Report. Manila: Department of Energy. Fields, Gary. 2003. Accounting for Income Inequality and its Change: A New Method with Application to the Distribution of Earnings in the United States. In Solomon Polachek and Konstantinos Tatsiramos, eds. Research in Labor Economics, Vol. 29. Emerald Group Publishing Limited. Hossain, Altaf, and Benjamin Zeitlyn. 2010. Poverty and Equity: Access to Education in Bangladesh. CREATE Pathways to Access Research Monograph No. 51. Brighton and Dhaka: University of Sussex and BRAC University.
130 ASIAN DEVELOPMENT REVIEW Independent Evaluation Group (World Bank). 2008. The Welfare Impact of Rural Electrication: A Reassessment of the Costs and Benets. Washington, DC: World Bank. Kumar, Sarath. 2011. Is Electrication Welfare-Improving: Non-Experimental Evidence from Rural Bhutan. Harvard School of Public Health research paper. Lodhi, Abdul, Daniel Tsegai, and Nicolas Gerber. 2011. Determinants of Participation in Childs Education and Alternative Activities in Pakistan. ZEF Discussion Papers on Development Policy No. 159. Mirza, Bilal, and Rene Kemp. 2011. Why the Rural Rich Remain Energy Poor. The Journal of Sustainable Development 6(1): 133155. Ministry of Education of Bhutan. 2004. National Report on the Development of Education. Paper presented at the 47th International Conference on Education, Geneva, 811 September. Paes de Barros, Richard, Francisco Ferreira, Jos e R. Molinas Vega, and Jaime Saavedra Chanduvi. 2008. Measuring Inequality of Opportunity in Latin America and the Caribbean. Conference Edition. Washington, DC: World Bank. Roemer, John. 1998. Equality of Opportunity. Cambridge, MA: Harvard University Press. Raynor, Janet, and Michael Wesson. 2006. The Girls Stipend Program in Bangladesh. Journal of Education for International Development 2:2. Available at http://www.equip123.net/ JEID/articles/3/GirlsStipendPrograminBangladesh.pdf Shorrocks, Anthony. 1982. Inequality Decomposition by Factor Components. Econometrica Vol. 50(1): 193211. Sijbesma, Christine, Truong Truong, and Jacqueline Devine. 2010. Case Study on Sustainability of Rural Sanitation Marketing in Viet Nam. Water and Sanitation Program technical paper. Available at http://www.wsp.org/wsp/sites/wsp.org/les/publications/ WSP_SustainabilityCaseStudy_TSSM.pdf Tandon, Ajay, and Juzhong Zhuang. 2007. Inclusiveness of Economic Growth in the Peoples Republic of China: What Do Population Health Outcomes Tell Us? ERD Policy Brief Series No. 47. Manila: Asian Development Bank. WaterAid. 2011. South Asian Peoples Perspective on Sanitation: Synthesis Review. London: WaterAid. World Bank. 2005. Electricity for All: Options for Increasing Access in Indonesia. Jakarta: World Bank Ofce Jakarta. . 2006. World Development Report 2006: Equity and Development. Washington, DC: World Bank. . 2010. Addressing the Electricity Access GapBackground Paper for the World Bank Group Energy Sector Strategy. Washington, DC: World Bank. World Health Organization and UNICEF. 2010. Progress on Sanitation and Drinking Water 2010 Update. WHO/UNICEF Joint Monitoring Programme on Water Supply and Sanitation. Woodhouse, Erik. 2005. The Philippines Electricity Market Investment Context. Paper presented at PESD Seminar, Stanford University, 23 June. Yuerlita, Rudi Febriamansyah, and Ade Saptomo. 2008. Peoples Participation in Rural Water Supply and Sanitation Project: A Case Study in Jorong Kampung Baru, Solok, West Sumatra, Indonesia. Paper presented at 3rd WEPA International Forum on Water Environmental Governance in Asia. Available at: http://www.wepa-db.net/pdf/0810forum/paper09.pdf Zhang, Xiaobo, and Ravi Kanbur. 2005. Spatial Inequality in Education and Health Care in China. China Economic Review, 16(2): 199204.
Political Connection and Firm Value

JAMES S. ANG, DAVID K. DING, AND TIONG YANG THONG
We study the effect of political connection (PC) on company value in an environment where low PC is due to better institutions and not confounded by favorable social/cultural factors. We nd that in Singapore, the only country that ts this description, PC in general adds little to the value of a company. However, in industries that are subject to more stringent government regulations, PC appears to be somewhat important. Robustness checks show that alternative PC variables give rise to similar results, and the addition of control variables do not drastically change the ndings. Politically connected rms have higher managerial ownership and tend to be smaller than non-PC rms, rendering them more susceptible to poorer governance practices. We show that the presence of politically connected directors somewhat neutralizes such potential negative effects. PC rms are associated with good governance practices such as nonduality in their chairman and chief executive ofcer positions and fewer executive directors. Keywords: political connection, corporate governance, rm value, Singapore JEL codes: G32, G34, O53
I. Introduction
While the value of political connection to rms has received considerable research interest (Goldman, Rocholl, and So 2009; Imai 2006; Khwaja and Mian 2004; Ang and Boyer 2007), the question of whether political connection enhances rm value has mixed ndings.1 When the value of political connection is found to be high, they are often in countries with higher levels of ofcial corruption. We do not know whether political connection is as valuable in the absence of political corruption, the question being, is political corruption a precondition for political connection to be valuable? To test this hypothesis, we examine the impact
James S. Ang is Professor of Finance at the College of Business, Florida State University, USA. David K. Ding is Professor of Finance at the School of Economics and Finance, Massey University, New Zealand and at the Lee Kong Chian School of Business, Singapore Management University. Tiong Yang Thong is Lecturer of Finance at the Lee Kong Chian School of Business, Singapore Management University. 1 Examining 47 countries, Faccio (2006) nds a positive relation between political connection and rm value. Specically, she nds that political connection is common in countries that are highly corrupt. Goldman, Rocholl, and So (2009) show positive abnormal stock returns following the announcement of a politically connected individual nominated to the board. Fan, Wong, and Zhang (2007), however, nd a negative relation between politically connected CEOs and post-IPO performance in the Peoples Republic of China.
of politically connected directors on the value of rms operating in an environment where the perceived level of corruption is comparatively low.2 Singapore was selected for the study because it is unique among low corruption countries. Unlike other low corruption countries where the results may be attributable to shared social and/or cultural factors with neighboring countries, Singapore has a distinctive need for much stronger institutions as it is surrounded by countries that are perceived to be inherently more corrupt. For example, the 2010 edition of the global corruption perceptions index released by Transparency International lists Denmark, New Zealand, and Singapore jointly as the worlds least corrupt countries. New Zealands closest neighbor, Australia, is ranked eighth on the same index. Denmark, together with neighbors Sweden and Finland, are ranked within the top ve in 2010. By contrast, Singapores immediate neighbors in Southeast Asia include Malaysia (ranked #56), Thailand (ranked #78), Indonesia (ranked #110), Viet Nam (ranked #116), and the Philippines (ranked #134). Furthermore, the social and cultural background of Singaporeans, as derived from their ethnic background, also does not favor low corruption. Of the countries of ancestral origin among Singapores three major ethnic groups (Chinese, Malays, and Indians), the Peoples Republic of China (PRC) ranked #78, Malaysia ranked #56, and India ranked #87. Thus, even if we nd political connection not beneting private rms with connections in both Scandinavian countries and Singapore, the underlying causes are inherently different. Singapore has to rely on having strong institutions to achieve low corruption, one that is not confounded by social and cultural factors. These unique factors allow for a natural experiment to be conducted on how institutions may limit the role of political inuence in businesses. Specically, we examine this issue within the context of post-initial public offering (IPO) rms in Singapore to show that rent-seeking through politically connected directors is not viable in countries viewed to be very clean, such as in Singapore, and where the legal and political institutions are fair, transparent, and effective.3 Previous studies did not specically investigate low corruption countries nor did they conclusively argue that the effect of political connection is independent of corruption. We investigate the value of political connection under a political regime with low perceived corruption and document that rms operating under such a political environment benet littleexcept for rms in highly regulated industriesfrom their political connection. This new nding, in contrast to those in previous studies that nd strong value in political connection under a corrupt
2 According to Transparency International, an international nongovernment organization addressing corruption including but not limited to political corruption, Singapore together with Denmark and New Zealand ranks as the cleanest country out of 178 in the world in 2010, with a corruption perceptions index of 9.3. Finland and Sweden complete the top ve least corrupt countries that year. 3 It has been well documented in the existing literature (e.g., Ritter 1991, Ritter and Welch 2002) that rms with a higher level of underpricing during an initial public offering (IPO) tend to underperform over the long run. In the present study, politically connected rms in the sample have a lower, but not statistically signicant, level of underpricing than nonpolitically connected rms.
POLITICAL CONNECTION AND FIRM VALUE 133
political regime, claries the role of political corruption and the channel through which values are created for such rms and/or their managers. A corrupt political environment increases the probability that rms connected politicians are willing and able to extract rents from the public and competitors on behalf of their rms. Such an environment also gives politically connected individuals certain incentives, as they may receive a share of the extracted rents in the form of personal payoffs or campaign contributions with low perceived personal risks. Politically connected rms may benet through easier access to debt nancing, lower taxes, or stronger market power. Such benets are usually greater when the rm operates in a country with a high level of corruption among its ofcials, low protection of property rights, a highly interventionist government, or a nondemocratic government (Faccio 2006). In addition to gaining economic benets, some rms may appoint politically connected directors for their knowledge and experience with government procedures, their insights into government actions, their ability to enlist the government for the rms interest at the expense of competitors, or to forestall government action inimical to the rm (Agrawal and Knoeber 2001). Goldman, Rocholl, and So (2009) nd that companies connected to the United States (US) Republican Party experience an increase in value following the Republican Partys win in the 2000 presidential election, while companies connected to the Democratic Party saw a stock price plunge. In Singapore, rms may appoint politically connected directors to their board to signal stronger corporate governance. This argument is in line with the nding of Ang and Ding (2006) that government-linked companies in Singapore are associated with stronger corporate governance and higher rm valuations. Political connection may add value to either the connected rms and/or their managers. An example is the contrast between Indonesia and the PRC. Managers of Indonesian rms are often the largest shareholders, where 84.6% of management is afliated with the controlling owners (see Claessens, Djankov, and Lang 2000). The extant literature highlights evidence from Indonesia where investors view political connection to the countrys president as valuable, accounting for one-fourth of a rms value and adding 33% to rm value (Fisman 2001). Thus, because managers stakes in the rms are large, a substantial share of rent extraction accrues to the rm. In the PRC, however, the average management ownership of Chinese rms at the time of IPO is a mere 0.298% (Li et al. 2007) and rms with political connection underperform their counterparts that have no political connections by 37% over a 3-year post-IPO period (Fan, Wong, and Zhang 2007). This nding is consistent with the view that managers with a low personal ownership in their rms mainly divert the rent extracted from political connection to themselves and the connected politicians.4
4 Other forms of political connection related to value reduction include having lower quality political appointees as managers and running the business as a political bureaucracy. It is also possible that Chinese investors fail to understand that it takes ownership alignment for rent extracted from political connection to ow to the rm and not the managers.
The issue addressed in this article is whether political connection enhances rm valuation when political corruption is low. Underlying the hypothesis is the conjecture that signicant rent extraction is made possible only under an environment of high political corruption. The alternative hypothesis is that political corruption may not be necessary for politically connected directors to help create value for their rms. For instance, politically connected directors who are not corrupt may be able to inuence their rms governance structure by aligning it with the governments policy initiatives, which may lead to higher investor condence and resultant higher rm value in some cases.5 Singapore is in a unique situation being the only low corruption country in a region dominated by high corruption. In contrast, Denmark and New Zealand, which are jointly ranked with Singapore as the least corrupt, have neighboring countries that are also low on corruption. Denmark and New Zealand are less suitable for studying the issue raised here as their environment of political corruption, the willingness of politicians to abuse their powers, and the likelihood of rms or managers engaging in political rent extraction, may be related to an unspecied common factortheir shared social and/or cultural environment with neighboring countries. Thus, Singapore provides a more ideal natural experiment for our study.6 In this study, we dene a company to be politically connected if at least one member on its board of directors is: (i) a former cabinet minister of the Singapore government, (ii) a serving or former Member of Parliament (MP), or (iii) a current or former senior civil servant of the Singapore government.7 We investigate the relation between political connection and rm valuation within the context of newly listed companies during 19982006. We employ Tobins Q as a proxy for the value of newly listed rms in Singapore for each of the 3 years after their issuance and compare the differences in value between companies with and without political connection. We also explore the relationship between political connection, corporate governance, and rm value. The results of this paper show that in a country such as Singapore, where political corruption is relatively low, political connection adds little to the value or performance of the company. However, upon further careful investigation, we nd that in certain industries, political connection appears to be more important than in others. These are industries that tend to be subject to more stringent government regulation such as electrical and electronic equipment (SIC 36), holding and other investment ofces (SIC 67), general building contractors (SIC 15), food and kindred
5 An example is from Ferguson and Voth (2008), who report that rms politically linked to Hitlers Nazi Germany outperformed the market by 5%10%. However, this should be properly viewed as a consequence of the skewed policy of fascism in favor of a few companies. 6 The International Monetary Fund (IMF), in June 2004, indicated in its nancial system stability assessment of Singapore that a competent judiciary is one of the cornerstones of Singapores legal system, giving top marks for the reliability of the countrys legal, supervisory, and institutional framework. 7 Currently serving ministers of the government are not permitted to sit on corporate boards. However, no such restrictions are imposed on MPs and senior civil servants.
products (SIC 20), and rubber and miscellaneous plastics products (SIC 30). There is evidence that companies in such industries may nd that having directors who are politically connected could have a positive and signicant impact on their rms value. We further examine the characteristics of these rms and nd that, among these rms, those with a politically connected chairman/chief executive ofcer (CEO) or senior civil servant on their board benet the most in terms of a positive effect on rm value. The remainder of this paper is organized as follows. In Section II, we present a brief introduction of corporate governance in Singapore. Section III reviews the prior literature on political connection and develops the hypotheses. The sample data and research methodology are described in detail in Section IV . Research outcomes and results are documented in Section V . Section VI summarizes the study and provides some concluding remarks.
II. Background of Corporate Governance in Singapore
Following loosely the Anglo-American model, Singapores corporate governance system revolves around capital market controls of managerial behavior (Prowse 1998). The capital market in Singapore is thin (less than 500 listed companies were on the Singapore Stock Exchange in 2002, growing to about 800 rms in 2008), and equity is rmly held among a small group of investors including the Government of Singapore, multinational and regional corporations, wealthy individuals, and entrepreneurial families. Government-linked corporations (GLCs) account for approximately 24% of the stock markets total capitalization of $287 billion and control over a 10th of the countrys economic output (Ang and Ding 2006). Therefore, any study of corporate governance in Singapore would not be complete without understanding the role and governance structure of Singapores GLCs. Typically, GLC boards are populated by senior civil servants and political appointees, making board appointments an oblique method for monitoring or controlling corporate activities and business practices by the government. The government-centered corporate governance system can be potentially effective if strong governance is regarded as keeping with effective industrial policy (Phan and Yoshikawa 2005). To buttress this view, Ang and Ding (2006) compare the nancial and market performance of GLCs with non-GLCs, where each had a different governance structure, the key difference being government ownership. They show that Singaporean GLCs have higher valuations and better corporate governance than a control group of non-GLCs. Their results hold even after controlling for rm-specic characteristics such as protability, leverage, rm size, and foreign ownership. GLCs in Singapore are largely corporate investments by Temasek Holdings, a wholly owned government entity that prides itself for its ability to make investment decisions strictly on a commercial basis. Temaseks articulated policy with respect to GLCs is to play a key monitoring role in commercially viable and nancially
independent companies. Due to Temaseks stake in GLCs, it is possible that some politicians are appointed on behalf of Temasek. It may then be argued that these appointees, in protecting the interests of Temasek, also maximize share value, as long as: (i) Temasek does not inuence the government to show favoritism at the expense of competitors, and (ii) it is not possible for Temasek or its appointees to share in the private benets of control. There are a few anecdotes to reinforce the view that politically connected rms in Singapore do not enjoy favoritism from the government. For example, since April 2007, the postal sector has witnessed the entry of new players in both domestic and international mail services after a 15-year monopoly held by SingPost.8 Also, the corporate tax rate is applied uniformly across all businesses and industries with the expectation that no rm is specially protected by the government to reap economic benets.
III. Literature Review and Hypotheses
Political connection, or the lack thereof, is a double-edged sword; it can either enhance or jeopardize a rms value. In the PRC, Xu, Zhu, and Lin (2002) show that when political control is curtailed, rm performance improves. This happens when there is a resulting increase in a rms exibility in labor deployment and in the enforcement of more effective corporate governance mechanisms. It is recognized, however, that some political appointees have conicting objectives, e.g., maximizing employment or minimizing social costs. Likewise, politically connected CEOs, instead of being helpful to rm performance, may have a deleterious effect. Fan, Wong, and Zhang (2007) report that rms with politically connected CEOs underperform those without political connection by 37% when measured by their rms 3-year post-IPO stock returns. In addition, performance measures such as market-to-book value and return on assets of statecontrolled rms are found to be negatively related to the level of state ownership (Fan, Wong, and Zhang 2007). Similarly, evidence from 47 countries shows that politically connected rms underperform nonpolitically connected rms on an accounting basis, notwithstanding the fact that they are able to derive considerable benets from their political connections (Faccio 2006). One possible explanation for the underperformance of politically connected rms is that the channeling of resources by politicians toward favored rms can lead to a distortion of incentives, misallocation of investment, and increase in corrupt activities (Shleifer and Vishny 1994). On the other hand, with political connection, a rm may increase in value if it manages to extract unfair economic rents at the expense of competitors and consumers (e.g., Faccio 2006). However, when all or more of the increase in rm value is consumed by politicians and their connected
8 Lee Hsien Yang, the brother of Prime Minister Lee Hsien Loong, serves as an independent director on SingPosts board.
managers, less of any remaining value would be available to shareholders. This is a form of agency problem that proper governance structures can help contain. A subtle point to note is that good governance, which aims to increase share value, does not equate to good citizenship. It is perfectly consistent for good governance rms to take advantage of political connection to increase share value under corrupt regimes. If political connection were used as a prime determinant of protability, it would induce distorted investment decisions (Faccio 2006) and consequently lead to lower rm value. There is evidence in Indonesia of rms facing difculties in building connections with a new government when their patron falls from power, causing those rms to underperform under the new regime and subsequently to turn to foreign nancing (Leuz and Oberholzer-Gee 2006). On the other hand, politically connected companies may benet from easier access to debt nancing, lower taxes, and stronger market power (Faccio 2006). For example, Friedman (1999) reports that bankers are often compelled to extend loans for projects undertaken by politically connected rms even when they are forecast to be unprotable, thus extracting rents from the banks. Such evidence, together with others (e.g., Johnson et al. 2006, Khwaja and Mian 2004, Sapienza 2004) provide further support that the discrepancy in the lending behavior of state-owned banks is affected by the electoral results of the party afliated with the bank. Such actions represent a wealth transfer from citizens or consumers to the rm, leading to an increase in rm value. Besides easier access to credit, politically connected rms may enjoy other benets. Some public ofcials and politicians may clandestinely sell underprovided goods and a spectrum of rent-generating advantages to individual rms, often allowing rms to shape the rules of the game to their advantage at considerable social cost (Hellman, Jones, and Kaufmann 2000). Alternatively, directors may be appointed for their knowledge, experience with government procedures, insights in government policy, and the ability to persuade the government in favor of the rms interest or forestall governmental action pernicious to the rm (Agrawal and Knoeber 2001). Imai (2006) shows that powerful business groups strive to directly hold inuential public ofces in order to change economic policies to their favor as companies with political connection are more likely to win a project tender because of the protection from corrupt politicians or bureaucrats. Thus, the award of government contracts to companies without basic qualications, resources, and expertise is, not surprisingly, often linked to political parties (della Porta and Vannucci 1997).9
9 It is noted that companies that form political connections with government ofcials in order to obtain perks or preferential treatment are characteristic not only of countries traditionally labeled as corrupt, but may also be found in countries known for their transparent systems. In the US, about half of those who leave government jobs, including some who have served in Congress, end up working as lobbyists, commanding a higher remuneration. Oftentimes, as in Japan, they end up working for corporations that try to inuence the decisions of government agencies they had left. These examples demonstrate just how imperfectly political corruption is measured. Many countries that have no corruption by bureaucrats and policemen, etc., nevertheless have politicians inuenced by money spent on lobbying and campaign contributions by interested groups and rms.
It is evident that politically connected directors have the power and ability to bring benets to rms by inuencing the laws under which their rms operate, as well as increasing the possibility of winning government contracts for their rms and, in the process, enhance rm value. Crudely put, politically connected board members can either inuence the transfer of wealth from competitors or consumers to their rms (Faccio 2006) or corruptly extract economic rents from the rm for their personal gain. We test the following null hypothesis. Hypothesis 1: In the absence of political corruption, political connections of rms have little or no effect on rm value. The alternative to this hypothesis is that political connections of rms could lead to private gains to the rms shareholders. In Singapore, offenses from and sanctions on corruption are set out in the Prevention of Corruption Act, which is vigilantly enforced. Punishment on corruption includes a ne of up to S$100,000 or imprisonment for a term not exceeding 7 years or both, in addition to other potential related criminal charges. With such stringent regulations in place to uphold the integrity of the business environment, it is therefore relatively difcult for directors to corruptly exploit their connections to bring economic benets to their rms or to abuse their position in the rm for their own personal economic, social, or political interests. Thus, we hypothesize that rms are generally not expected to gain from their political connections. However, rms operating in a more highly regulated industry may deem it advantageous to appoint directors who are politically connected, but not corrupt, to their board.10 A low-corruption government has two means of limiting the value of political connections to rms. One, it is more likely to appoint, or allow to be appointed, ofcials who are not corrupt onto corporate boards. Two, it would set up mechanisms to identify and punish corrupt ofcials. The following hypothesis is tested. Hypothesis 2: Politically connected directors in rms operating in a more highly regulated industry are associated with a higher rm value.
IV. Data and Methodology A. Data and Sample Design
Underpricing during an IPO is a common phenomenon that is followed by abnormally low returns in the long run (see Ritter 1991). IPOs allow us to investigate
10 Highly regulated industries include those with oversight by a government department that requires the company to make regular submissions on compliance issues, and have guidelines, standards, or legislation to meet. These companies are required to fulll certain compliance and licensing requirements set by a government authority.
the wholesale introduction of all politically connected directors the rst time the rms are introduced to the capital market. The design eliminates the need to adjust for the timing of an anticipation effect and the incremental value of a single appointee in later years. Thus, IPOs provide an ex ante expectation of the effect of political connection on rm value, while the value of political connections in seasoned rms is captured in ex post pricing. The IPO data are obtained from the Securities Data Corporation (SDC) Global New Issue database and cover all IPOs during 19982006. We use the SDC database to obtain basic information on offer dates, offer prices, the number of shares issued, net proceeds, and the number of lead and co-lead managers. Prior to the IPO, we extract the latest nancial information such as total assets, total liabilities, net income, and the debt-to-asset ratio from the individual rms IPO prospectus. PostIPO, nancial data such as total assets, current assets, current liabilities, long-term debt, preferred stock, and market value of common equity are obtained from the Thomson One Banker database. From the IPO prospectus, we obtain the prole of each director. A total of 2,540 directors are covered in the sample of 387 listed companies. We manually identify directors that are politically connected according to the denition described in Section I. Table 1 presents summary descriptive statistics of the entire sample of newly listed companies. Panels A, B, and C contain their offer, rm, and corporate governance characteristics, respectively. On average, the sample rms offer 74.18 million shares in their IPO, garnering net proceeds of $21.06 million each. We note that rms with political connection (PC rms) on average issue 63.6 million shares which is less than those issued by non-PC rms. The median difference of these two numbers is statistically signicant at the 10% level. The net proceeds raised by PC rms also tend to be lower than those of non-PC rms. The average offer price is $0.47 with mean offer-to-close returns of 24%. On average, there are 1.33 underwriters per issue. Firm characteristics are described in Panel B. We note that the market value of each rms equity averages $183.75 million with an average of 355.79 million shares outstanding. PC rms have average total assets valued at $124.5 million whereas the corresponding gure for non-PC rms is higher at $211.8 million. The market value of PC rms at the time of IPO is also lower than that of non-PC rms. In Panel C, we report that the management of PC rms on average own 44.5% of their rms compared to 41% of non-PC rms. The average number of executive directors among PC rms is 2.78 compared to 2.90 for non-PC rms. PC rms report a signicantly lower number of chairmen who also serve as CEOs than nonPC rms. The average age of the directors is 46.6 years with those of PC rms being slightly older. Overall, compared to non-PC rms, PC rms tend to be of a smaller size with a higher degree of management ownership and have older directors. The evidence from Table 1 shows that PC rms have a higher (though not statistically signicant) percentage of management ownership of the rms equity
Table 1. Descriptive Statistics of Politically and Nonpolitically Connected Companies

Total Median Mean Median Mean Median Mean Politically Connected Nonpolitically Connected Difference (Politically Nonpolitically) Median
Mean
Panel A: Offer Characteristics 44.400 6.937 0.260 0.114 1.000 63.618 18.343 0.621 0.229 1.330 39.100 6.466 0.260 0.068 1.000 77.706 21.963 0.422 0.241 1.331 46.013 7.049 0.260 0.116 1.000 14.087 3.620 0.199 0.013 0.001 6.913 0.583 0.000 0.048 0.000
Shares offered (million) Net proceeds (US$ million) Offer price (S$) Offer-to-close return Number of underwriters
74.175 21.055 0.472 0.238 1.331
Panel B: Firm Characteristics 37.620 34.457 18.218 0.236 0.130 2.930 68.901 202.948 80.490 124.460 58.977 0.244 0.173 3.689 162.729 283.573 40.825 35.279 18.731 0.225 0.131 2.995 69.163 206.975 107.456 211.818 100.789 0.300 0.306 4.999 214.713 399.145 38.675 35.936 19.520 0.254 0.127 2.915 71.058 219.853 26.966 87.358 41.812 0.057 0.133 1.311 51.984 115.572 2.150 0.658 0.788 0.030 0.005 0.080 1.895 12.878
Continued.
Net sales (S$ million) Total assets (S$ million) Total liabilities (S$ million) Debt-to-assets ROA EPS (S$) Market value (S$ million) Shares outstanding (million)
84.804 178.414 85.714 0.282 0.285 4.403 183.748 355.785
Table 1. Continued.
Total Median Mean Median Mean Median Mean Median Politically Connected Nonpolitically Connected Difference (Politically Nonpolitically)
Mean
Panel C: Corporate Governance Characteristics 0.486 0.426 6.563 2.416 1.000 3.000 47.000 0.000 0.505 2.777 47.859 0.040 1.000 3.000 48.000 0.000 0.623 2.896 46.382 0.039 1.000 3.000 46.000 0.000 0.118 0.119 1.477 0.001 0.536 0.4141 6.000 2.000 0.475 0.445 6.515 2.417 0.510 0.447 6.000 2.000 0.496 0.410 6.684 2.472 0.550 0.375 6.000 2.000 0.021 0.035 0.169 0.055 0.040 0.072 0.000 0.000 0.000 0.000 2.000 0.000
Institutional ownership Management ownership No. of board members No. of independent directors No. of ofcials with chairmanCEO functions (duality) No. of executive directors Age of directors GLC
0.615 2.920 46.576 0.040
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test), EPS = earnings per share, GLC = government-linked corporations (i.e., number of IPOS that have government ownership), ROA = return on assets. Notes: 1. This table presents the mean and median values of the offer, rm, and corporate governance characteristics of the politically and nonpolitically connected companies in panel A, B, and C, respectively. 2. In panel A, we report the offer characteristics, i.e., shares offered, net proceeds, offer price, offer to close return, and number of underwriters. 3. In panel B, we report the rm characteristics, i.e., net sales, total assets, total liabilities, debt-to-assets, return on assets, earnings per share, market value, and shares outstanding. 4. In panel C, we report institutional ownership, management ownership, number of board members, number of independent directors, number of individuals with chairman-CEO functions (duality), number of executive directors, age of directors, and number of IPOs that have government ownership. 5. t-statistics and z-statistics based on Wilcoxon signed ranks test were used for the test of differences in the mean and median, respectively. Source: Securities Data Corporation (SDC) Global New Issue database and IPO prospectus.
and are of a smaller size (again not statistically signicant) in terms of IPO proceeds, asset size, and market value at the time of IPO. Firms with such characteristics may typically be expected to have a lower valuation compared to larger rms and those that have a lower percentage of management ownership. However, the presence of PC directors appears to have somewhat neutralized such effects with their association with better corporate governance practices such as nonduality in chairman and CEO positions and fewer executive directors. The sample consists of 97 politically connected companies and 290 nonpolitically connected companies that are newly listed over a 9-year period from 1998 to 2006. This translates to 33.45% of the sample IPO companies that are deemed to have political connection. In addition, from the prole of the 106 politically connected directors identied in this study, we document in Table 2 that 89% of these act as independent directors on the board. Moreover, 36% are current political appointees, 91% are former political appointees, 13% are politically connected chairmen, 22% are former cabinet ministers in the Singapore government, and 16% are senior civil servants of Singapore. Members of the Parliament account for 89% of politically connected directors. It is noted that current serving ministers are not permitted to sit on corporate boards. Table 3 presents the background of the politically connected directors, including their age, business-related education, work experience, and the number of outside directorships that they hold at and before the time of IPO. We note that current (former) political appointees on corporate boards have worked a median total of 24 (27) years, 11 (16) of which were spent in government or in the senior civil service. This implies that politically connected directors have had a total of 1113 years of business and professional experience which is not insignicant.11 Such experience may be attractive to rms operating in certain highly regulated industries. In Table 3, we nd that politically connected directors generally hold multiple outside directorships indicating that they are busy board members.
B. Firm Valuation Using Tobins Q
Tobins Q, the ratio of the market value of a companys assets (measured by the market value of outstanding stock and debt) to the replacement costs of the companys assets, is an important and widely accepted measure of corporate performance. As such, we make use of it as a proxy for rm value. It can be approximated by: Q= MV(CS) + BV(PS) + BV(CL) + BV(LTD) BV(CA) BV(TA) (1)
11 The corresponding background experience of directors of nonpolitically connected rms is not complete nor consistently available. Thus, comparisons between the two types of rms cannot be properly made,

Table 2. Types of Politically Connected Directors
Types of Political Connection Current political appointees Former political appointees Politically connected chairmen Politically connected independent directors Members of Parliament Current Former Former ministers Senior civil servants Current Former N No. of Directors (%) 38 (36%) 96 (91%) 14 (13%) 94 (89%) 94 (89%) 37 (35%) 34 (32%) 23 (22%) 17 (16%) 1 (1%) 16 (15%) 106 (100%) No. of Firms (%) 35 (36%) 67 (69%) 14 (14%) 80 (82%) 81 (84%) 31 (32%) 30 (31%) 23 (24%) 17 (17%) 1 (1%) 16 (16%) 97 (100%)
Notes: 1. This table reports the number of directors and number of rms for different types of politically connected directors based on 387 IPO rms from 1998 to 2006. 2. Current political appointees are those currently serving in government. Former political appointees are those who previously served in government. 3. Politically connected chairmen and independent directors are those who are related to the government. 4. Members of Parliament (MPs) include directors who are current or former MPs elected as the peoples representative in the parliament. 5. Former ministers are directors who previously served in a cabinet position. 6. Senior civil servants include senior military personnel, permanent secretaries, and parliament secretaries working in their respective ministries. 7. The percentage values in parentheses measure the ratios of each type of director to the total number of politically connected directors or the ratios of each type of rm to the total number of politically connected rms. Source: Securities Data Corporation (SDC) Global New Issue database and IPO prospectus.
where MV(CS) is the market value of common shares, BV(PS) the book value of preferred shares, BV(CL) the book value of current liabilities, BV(LTD) the book value of long term debt, BV(CA) the book value of current assets, and BV(TA) the book value of total assets. This simplied Q measure has been shown to account for at least 96.6% of the variability of Tobins Q (Chung and Pruitt 1994). A Tobins Q that is greater than one indicates that the company has a market value greater than its recorded assets, which can be attributed to intellectual capital or positive
Table 3. Background of Politically Connected Directors

Current Appointees Former Minister Mean 20.11 27.73 17.99 16.34 60.09 Median 21.00 30.00 14.00 9.00 61.00 No. of Firms 19 15 Mean 10.97 23.45 12.40 9.28 46.76 No. of Firms 14 21 45 38 50 50 No. of Firms No. of Firms 11.00 23.50 11.00 7.00 47.00 16.79 24.88 15.72 13.05 57.39 16.00 26.50 12.00 6.00 58.75 14.79 23.48 15.24 11.60 53.03 16.00 23.00 12.00 6.00 53.00 Median Mean Median Mean Median Former Appointees Current & Former MPs Senior Civil Servants Mean 21.06 30.65 9.76 10.18 62.41 Median 28.00 30.00 8.00 6.00 64.00 No. of Firms 13 12
Years served in politics Years of work experience No. of present outside directorships (at IPO) No. of past outside directorships (prior to IPO) Age of director
Business-related education Business-related industry experience
IPO = initial public offering, MP = member of parliament. Note: This table reports the backgrounds of the different types of politically connected directors, with background information collected from the IPO prospectus of rms. Source: Securities Data Corporation (SDC) Global New Issue database and IPO prospectus.
market sentiment of the company. As such, a higher Tobins Q value is associated with superior rm value. We compute the industry-adjusted Tobins Q. Specically, we subtract the industry median Tobins Q based on the two-digit SIC code from the sample rms Tobins Q. By doing this, we are able to detect whether the sample rm outperforms the industry and, at the same time, control for any industry-related biases.
C. Methodology
We employ both univariate and multivariate analyses to test our hypotheses. We examine the association between rm value and political connection by performing linear regressions of Tobins Q for each of the 3 years after an IPO and for the median of the 3 post-IPO years against alternative political connection denitions. We run the following main regression: Value = 0 + 1 Polconnect + 2 Duality + 3 Independent + 4 Dirage + 5 Underpricing + 6 Leverage + 7 ROA + 8 Exchange + 9 GLC + 10 Mktval + (2)
where the dependent variable, Value, refers to the industry-adjusted Tobins Q of the rm 1, 2, and 3 years after IPO listing and the median industry-adjusted Tobins Q over the 3 years. Polconnect refers to political connection measured by the ratio of the number of politically connected directors to the total number of board members. Control variables used in the regression models are as follows. Duality refers to the presence of a CEO who is also the chairman. Independent refers to the percentage of directors on the board who are outside directors. Dirage is the average age of directors. Underpricing is the offer-to-close return computed as the difference between the rst trading days closing and offer price as a percentage of the offer price. Leverage is measured by the debt-to-asset ratio. ROA is the return on assets computed as net income divided by total assets. Exchange is a dummy variable that takes on the value of one if a rm is listed on the main board of the Singapore Exchange (SGX) and zero otherwise. GLC is a dummy variable with a value of one if it is a subsidiary of Temasek and zero otherwise.12 Mktval refers to the natural logarithm of market value dened as the rst days closing price multiplied by the number of shares outstanding after the IPO.
12 These are government-linked companies (GLC) in which Temasek Holdings, the investment holding arm of the Singapore government, has at least a 20% stake. Examples of GLCs include some of the largest companies in Singapore such as Singtel, DBS Bank, Singapore Airlines, PSA International, SMRT Corporation, Singapore Power, and Neptune Oriental Lines. In our IPO sample, the GLC rms include Chartered Semiconductor Manufacturing, Singapore Airport Terminal Services, Singapore Post Ltd., Olam International Ltd., etc.
In addition to the inclusion of several control variables, robustness checks are performed using a number of alternative denitions of political connection. These include using political connection as a dummy variable, which takes on the value of one when at least one director is politically connected, and taking the natural logarithm of the number of politically connected directors. Additionally, in place of these variables, we employ dummy variables to reect the status of directors as current or former political appointees. The control variables included in this study are supported by previous research (e.g., Ang and Ding 2006, Chong and Lopez-de-Silanes 2006, Leuz and OberholzerGee 2006), and can be grouped into two categories. In the rst category are rmspecic control variables such as rm size, leverage, and protability.13 These include Mktval, which is the natural logarithm of a rms rst day post-IPO market value, used as a measure of rm size. Leverage is used as a proxy for mapping the risk prole of a company. ROA is used to control for a rms protability. The second category includes control variables that capture differences in corporate governance among the rms. A GLC dummy takes on a value of one when a company is government-linked, and zero if otherwise. It is noted that GLCs have been shown to provide superior returns (on both assets and equity) and are valued more highly because of better management of expenses than non-GLCs (Ang and Ding 2006).14 In order to examine the effect of corporate governance on rm value while considering a rms level of political connection, two variables are used as governance proxies: the percentage of independent directors and duality. Director independence is an indicator of the presence of a strong and independent board, whereas duality illustrates an appropriate balance of power, increased accountability, and greater capacity of the board for independent decision making if the chairman and CEO positions are held by separate persons. With an effective board, the opportunities for controlling shareholders and management to expropriate funds will, hopefully, be reduced.
V. Results and Analysis A. Univariate Analysis
Using independent t-tests and Wilcoxon signed ranks tests, we examine the differences in means and medians, respectively, of the industry-adjusted Tobins Q valuation for 1, 2, and 3 years after an IPO and the median industry-adjusted Tobins Q over all 3 years after an IPO listing between politically connected rms
13 14
We thank an anonymous referee for suggesting that protability be controlled for directly. Investors in the Singaporean market appear to value the higher standards of corporate governance found in
GLCs.

Table 4. Firm Value in Post-IPO Years
Politically Connected 1-year Mean Median N 2-year Mean Median N 3-year Mean Median N Post median Mean Median N
Nonpolitically Connected 0.17 0.09 199 0.07 0.01 193 0.21 0.08 170 0.08 0.002 154
Difference 0.13 (0.75) 0.01 (0.35) 0.20 (0.77) 0.02 (0.43) 0.25 (0.72) 0.02 (0.53) 0.15 (0.23) 0.038 (0.15)
0.30 0.10 61 0.27 0.03 63 0.46 0.10 58 0.23 0.04 48
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test), IPO = initial public offering. Notes: 1. This table reports the mean and median of the industry-adjusted Tobins Q for politically and nonpolitically connected rms in the post-IPO years, specically, 1, 2, and 3 years after an IPO. Post median refers to the median industry-adjusted Tobins Q for all three post-IPO periods. 2. Tobins Q is computed based on the Chung and Pruitt (1994) method as follows Q= M V (C S ) + BV ( P S ) + BV (C L ) + BV ( L T D ) BV (C A) BV (T A)
where MV(CS) is the market value of common shares, BV(PS) the book value of preferred shares, BV(CL) the book value of current liabilities, BV(LTD) the book value of long term debt, BV(CA) the book value of current assets, and BV(TA) the book value of total assets. 3. We compute the industry-adjusted Tobins Q by subtracting the industry median from the sample rms industry-adjusted Tobins Q. 4. t-statistics and z-statistics based on Wilcoxon signed ranks test (the gures in parentheses) were used for the test of differences in the mean and median, respectively. Source: Securities Data Corporation (SDC) Global New Issue and Thomson One Banker databases.
and those that are not politically connected. The ndings for industry-adjusted Tobins Q valuation are presented in Table 4, which shows that the industry-adjusted Tobins Q is not statistically signicant between the two groups. This suggests that there is insufcient evidence to show that political connection adds to rm value in Singapore and that any benets that such connections might bring appear to be limited, supporting Hypothesis 1 above. In place of Tobins Q, the results for cumulative abnormal stock returns (the difference between daily returns and market returns) of politically and nonpolitically
148 ASIAN DEVELOPMENT REVIEW Table 5. Cumulative Abnormal Returns in Post-IPO Years
Politically Connected 1-year Mean Median N 2-year Mean Median N 3-year Mean Median N
Nonpolitically Connected 0.20 0.20 234 0.33 0.48 195 0.44 0.72 129
Difference 0.03 (0.55) 0.04 (0.49) 0.02 (0.16) 0.04 (0.20) 0.07 (0.67) 0.07 (0.91)
0.17 0.24 75 0.35 0.52 54 0.51 0.65 34
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test), IPO = initial public offering. Notes: 1. This table reports the mean and median of the cumulative abnormal returns for politically and nonpolitically connected rms in the post-IPO years. Specically, we compute the buy-and-hold abnormal returns for 1, 2, and 3 years after an IPO. 2. The abnormal return is the difference between the daily returns and the market returns. The market return is computed based on the Singapore Straits Times Index (STI). 3. t-statistics and z-statistics based on Wilcoxon signed ranks test (the gures in parentheses) were used for the test of differences in the mean and median, respectively. Source: Securities Data Corporation (SDC) Global New Issue database and Datastream.
connected rms for 1, 2, and 3 years after an IPO are shown in Table 5. The mean and median values show that abnormal stock returns of politically connected rms do not outperform those of nonpolitically connected rms up to 3 years after an IPO as both the results of the t-test and Wilcoxon signed ranks test are not statistically signicant. Again, our evidence does not support shareholders having benetted from the appointment of politically connected directors. We examine the differences in accounting performance of the politically connected and unconnected rms and report the median values of the prot margin, cash ows from operations, return on assets (ROA), return on equity (ROE), and return on invested capital (ROIC) for 1, 2, and 3 years after the IPO and for median for the 3 years in Table 6. In general, from the results in Panel A, we do not nd any signicant differences in the accounting performance between the two types of rms. However, if we limit the sample to the more regulated industries, the results in Panel B show that the prot margin and cash ow from operations in the 1-year post-IPO period is signicantly higher among PC rms than non-PC rms. These ndings indicate that politically connected directors are associated with indicators of protability and rm value within the rst year of an IPO among rms that are more highly regulated.

Table 6. Accounting Performance of Politically and Nonpolitically Connected IPOs
Politically Connected Panel A: Full Sample Prot margin 1-year 2-year 3-year Post median 1-year 2-year 3-year Post median 1-year 2-year 3-year Post median 1-year 2-year 3-year Post median 1-year 2-year 3-year Post median 7.395 5.695 4.645 5.970 1.905 2.560 2.440 1.780 8.760 5.730 4.745 4.940 14.600 9.640 7.675 7.940 11.720 7.920 5.950 6.920 7.930 4.935 3.960 5.480 1.930 2.235 2.130 1.665 7.850 6.390 5.080 5.525 13.030 10.555 7.930 8.480 10.365 8.010 7.480 7.675 0.783 0.826 0.633 0.661 0.468 1.411 0.430 0.041 0.430 0.495 0.287 0.537 0.577 0.181 0.076 0.046 0.190 0.502 0.554 0.692 1.603 0.908 0.131 0.597 2.013 0.908 0.090 0.682 0.686 0.528 0.475 0.203 0.881 0.979 1.247 1.116
Continued.
Nonpolitically Connected
Wilcoxon Signed Ranks Test
Cash ow from operations
ROA
ROE
ROIC
Panel B: Regulated Industries Prot margin 1-year 2-year 3-year Post median 1-year 2-year 3-year Post median 1-year 2-year 3-year Post median 1-year 2-year 3-year Post median 4.700 1.688 0.063 0.766 4.950 0.298 1.450 1.160 4.663 1.680 0.230 0.694 3.265 3.420 1.145 2.585 2.970 0.198 0.135 0.290 1.965 1.668 1.743 1.620 2.730 0.090 0.850 0.188 0.588 1.545 4.890 2.085
Cash ows from operations
ROA
ROE
B.
Multivariate Analysis
The initial results (not reported) of ordinary least squares (OLS) regressions using industry-adjusted Tobins Q (a proxy for rm value for each of the 3 years and
150 ASIAN DEVELOPMENT REVIEW Table 6. Continued.

Politically Connected ROIC 1-year 2-year 3-year Post median 3.495 1.290 0.108 0.410 Nonpolitically Connected 2.960 1.380 2.843 0.734 Wilcoxon Signed Ranks Test 0.227 0.825 1.030 0.475
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test), IPO = initial public offering, ROA = return on assets, ROE = return on equity, ROIC = return on invested capital. Notes: 1. This table reports the median values of the accounting variables for politically and nonpolitically connected rms in the post-IPO years, specically, 1, 2, and 3 years after an IPO. 2. Post median refers to the median for all three post-IPO periods. 3. z-statistics based on Wilcoxon signed ranks test were used for the median test between politically and nonpolitically connected rms. Source: Securities Data Corporation (SDC) Global New Issue and Thomson One Banker databases.
the median of the 3 years after an IPO listing) as the key dependent variable show that political connection, which is expressed as a percentage of politically connected directors on the board, does not have a strong predictive power in explaining rm value. Political connection is found to be not statistically signicant in explaining industry-adjusted Tobins Q over various years. We therefore cannot reject Hypothesis 1 that political connection is not associated with rm value. Similar results (not reported) are obtained when the political connection percentage variable is replaced by a political connection dummy variable and the natural logarithm of the number of politically connected directors. However, upon further investigation, when we break down the sample according to the two-digit SIC, we nd that, in certain industries, political connection appears to be more important than in others. These are the industries that tend to be subject to more stringent government regulation such as electrical and electronic equipment (SIC 36), holding and other investment ofces (SIC 67), general building contractors (SIC 15), food and kindred products (SIC 20), and rubber and miscellaneous plastics products (SIC 30).15 The ndings for these industries show that the interaction between political connection and a particular regulated industry are mostly positively signicant at the 10% level (see Model 2 of AQ_post2yr and AQ_median in Table 7). As these industries carry higher regulatory riskse.g., not knowing the appropriate regulations, their interpretation, and/or procedures to observe themPC directors may help rms alleviate/mitigate these risks by providing the appropriate advice
15 Industries that are subject to a more stringent regulatory environment include: electronic and electronic equipment (SIC 36), which is governed by the Infocomm Authority of Singapores Electronic Transactions Act; holding and other investment ofces (SIC 67), which is regulated by the various securities and nancial acts of the Monetary Authority of Singapore; general building contractors (SIC 15), regulated by the Building and Construction Authority; food and kindred products (SIC 20), by the Agri-Food and Veterinary Authority; and rubber and miscellaneous plastics products (SIC 30), by the Rubber Association of Singapore.
Table 7. Regression Results of Industry-adjusted Tobins Q

AQ_post2yr Model 1 1.155 0.119 0.241 0.820 0.484 0.278 0.199 0.057 0.226 0.172 0.126 0.063 0.033 0.524 0.482 0.214 0.059 0.132 0.130 0.192 0.191 Model 2 2.383 0.084 0.197 0.617 0.637 0.275 0.101 0.073 0.215 0.160 0.097 0.031 0.144 0.760 0.617 0.348 0.070 0.070 0.335 0.253 0.135 2.098 3.277 5.115 1.812 1.440 0.998 1.436 4.125
Continued.
AQ_post1yr Model 2 4.357 0.061 0.003 0.041 1.108 0.118 0.001 0.116 0.026 0.272 0.112 0.016 0.115 0.755 0.231 0.503 0.162 0.197 0.556 0.162 0.421 0.927 0.022 11.594 1.951 1.710 0.445 1.618 1.080 Model 1 0.691 0.181 0.135 1.945 0.277 0.171 0.122 0.167 0.041 0.334 0.156 0.159 0.239 0.266 0.805 0.171 0.150 0.019 0.422 0.349 0.116 Model 2 0.761 0.160 0.107 1.585 0.061 0.107 0.0002 0.205 0.042 0.236 0.134 0.341 0.346 0.195 0.901 0.327 0.235 0.063 0.526 0.360 0.196 3.183 3.284 16.195 1.641 1.522 1.606 1.176 2.533 Model 1 1.928 0.156 0.250 2.390 0.314 0.199 0.070 0.225 0.080 0.043 0.109 0.261 0.156 0.098 0.688 0.199 0.175 0.042 0.521 0.550 0.051 Model 2 2.814 0.454 0.220 1.948 0.168 0.036 0.041 0.271 0.102 0.291 0.052 0.593 0.399 0.625 0.527 0.344 0.3003 0.064 0.494 0.684 0.108 6.703 3.920 18.557 1.219 0.990 7.648 2.334 2.977
AQ_post3yr
AQ_median
Intercept Polconnect Duality Independent Dirage Underpricing Leverage ROA Exchange GLC Market Value Industry1 Industry2 Industry3 Industry4 Industry5 Industry6 Industry7 Industry8 Industry9 Industry10 PolconnectIndustry1 PolconnectIndustry2 PolconnectIndustry3 PolconnectIndustry4 PolconnectIndustry5 PolconnectIndustry6 PolconnectIndustry7 PolconnectIndustry8
Model 1 4.030 0.155 0.145 0.137 1.189 0.095 0.002 0.070 0.029 0.203 0.124 0.012 0.056 0.226 0.276 0.387 0.080 0.096 0.474 0.106 0.390
Table 7. Continued.
AQ_post2yr Model 1 0.026 1.34 246 0.093 1.05 219 0.008 0.92 195 Model 1 Model 1 AQ_post3yr AQ_median
AQ_post1yr
Model 1
PolconnectIndustry9 PolconnectIndustry10 Adj. R2 F-Statistic N
0.026 1.33 252
Model 2 0.643 2.089 0.066 1.57 252
Model 2 0.028 0.958 0.009 0.93 246
Model 2 1.143 1.949 0.001 1.01 219
Model 2 1.088 3.251 0.008 1.05 195
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test) using White heteroskedasticityconsistent t-statistics. Notes: 1. This table presents the regression results for the abnormal rm value. Industry-adjusted Tobins Q is measured by the sample rms Tobins Q minus the industry median Tobins Q. Industry classication is based on the 2-digit SIC code. 2. Dependent variables AQ_post1yr, AQ_post2yr, and AQ_post3yr are the industry-adjusted Tobins Q estimates 1, 2, and 3 years after an IPO. AQ_median is the median industry-adjusted Tobins Q estimate over the 3 years. 3. The independent variables were dened as follows. Polconnect is a dummy variable that takes the value of 1 if the director is politically connected, 0 otherwise. Duality occurs when the chairman is also the CEO. Independent refers to the percentage of independent directors on the board. Dirage is the average age of the rms directors. Underpricing is computed as the difference between the rst days closing price and offer price, computed as a percentage of the offer price. Leverage is total debt divided by total assets. ROA is the return on assets prior to the IPO. Exchange is a dummy variable taking the value of 1 if the rm is listed on the main board of the SGX, 0 otherwise. GLC is a dummy variable taking the value of 1 if the rm is a government-linked corporation, 0 otherwise. Market Value is the rst days closing price multiplied by the number of shares outstanding after the IPO. Industry 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 are dummy variables taking the value of 1 if the 2-digit SIC codes are 36, 67, 73, 87, 28, 15, 20, 30, 50, and 35, respectively; 0 otherwise. Source: Securities Data Corporation (SDC) Global New Issue database, IPO prospectus, and Thomson One Banker database.
while being mindful of any potential conicts of interests. Our results reveal that companies operating in these industries have a positive and signicant addition to their rm value due to their political connection based on the 3-year median postIPO industry-adjusted Tobins Q. The results using abnormal stock returns (Table 8) largely corroborate those of the industry-adjusted Tobins Q. Given the existence of stringent laws (namely the Prevention of Corruption Act) that are strictly enforced with its attendant criminal and civil penalties against corruption in Singapore, together with the countrys high ranking in Transparency Internationals Corruption Perceptions Index, it is highly probable that any corrupt ofcial will be very quickly brought to task. As documented earlier, since PC directors are associated with better governance practices, their presence on corporate boards may compel rms to better adhere to the appropriate regulations. We are not claiming that PC rms do not receive government contracts or any preferential treatment. Rather, by virtue of their being perceived as having good governance under a low corruption environment where government regulations are observed, such companies are likely to be attractive to both private businesses and governments for suitable business alliances. Our ndings provide evidence that it is possible for political connection to be independent of corruption. The results for governance show that Duality is negatively and signicantly related to industry-adjusted Tobins Q in the second year post-IPO, whereas Independent is positive and signicant for the median industry-adjusted Tobins Q in the 3 years after an IPO. Dirage is negative and signicant for the 1-year industry-adjusted Tobins Q post IPO. We also investigate the interaction effect between political connection and two governance variables (duality and board independence) and nd insignicant results (not reported). This implies that there is insufcient evidence to show that politically connected rms in a low corruption environment, regardless of their corporate governance, reduce rm values. To lend further support to our ndings, we include an additional interactive term between political connection and GLC (to reect the presence of government ownership in a rm) into our multivariate regression. The results from this interactive term are found to be not statistically signicant. For greater robustness of our results, we have allowed for ner classications of political connection (see Table 2). That is, we further divide these PC directors into the following: politically connected chairmen/CEOs, current political appointees, former political appointees, MPs, former ministers, and senior civil servants. The regression results (Tables 9 and 10) for each type of political connection show that such connections add little to rm value. In Tables 9 and 10, besides investigating the value effects of different political connection classications, we include a dummy variable (Reg Ind) that represents the ve regulated industries identied in Table 7 to have a signicant contribution to the value of rms with PC directors.
Table 8. Regression Results of Abnormal Stock Returns 154 ASIAN DEVELOPMENT REVIEW
XRET1 Model 2 0.961 0.162 0.107 0.101 0.173 0.130 0.0003 0.057 0.074 0.121 0.014 0.149 0.088 0.160 0.093 0.232 0.213 0.099 0.114 0.183 0.513 1.984 0.498 1.623 0.415 0.115 4.975 1.235 0.420 0.171
Continued.
XRET2 Model 1 2.791 0.273 0.008 0.542 0.171 0.216 0.001 0.016 0.211 0.200 0.236 0.293 0.353 0.131 0.204 0.194 0.060 0.238 0.307 0.230 Model 2 2.759 1.444 0.051 0.603 0.218 0.170 0.001 0.004 0.203 0.184 0.279 0.401 0.504 0.270 0.259 0.254 0.307 0.269 0.382 0.329 0.377 3.446 4.345 2.625 1.008 1.371 5.908 0.752 1.220 2.179 Model 1 2.414 0.738 0.173 0.250 0.140 0.257 0.016 0.248 0.703 0.185 0.372 0.390 0.307 0.493 0.201 0.499 0.280 0.170 0.176 0.238 Model 2 2.047 1.592 0.138 0.222 0.125 0.234 0.027 0.243 0.658 0.158 0.396 0.418 0.459 0.453 0.281 0.143 0.453 0.151 0.284 0.253 0.050 0.391 2.969 1.162 1.284 1.243 6.255 0.082 2.216 0.584
XRET3
Intercept Polconnect Duality Independent Dirage Underpricing Leverage Exchange GLC Market Value Industry1 Industry2 Industry3 Industry4 Industry5 Industry6 Industry7 Industry8 Industry9 Industry10 PolconnectIndustry1 PolconnectIndustry2 PolconnectIndustry3 PolconnectIndustry4 PolconnectIndustry5 PolconnectIndustry6 PolconnectIndustry7 PolconnectIndustry8 PolconnectIndustry9 PolconnectIndustry10
Model 1 0.787 0.249 0.096 0.152 0.260 0.146 0.0004 0.069 0.098 0.132 0.023 0.078 0.096 0.044 0.137 0.198 0.047 0.028 0.081 0.160
Table 8. Continued.
XRET2 Model 2 0.227 3.98 296 Model 1 0.097 2.35 241 Model 2 0.103 1.95 241 Model 1 0.059 1.51 154 XRET3 Model 2 0.024 1.14 154
XRET1
Adj. R2 F-Statistic N
Model 1 0.182 4.45 296
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test) using White heteroskedasticity-consistent t-statistics. Notes: 1. This table presents the regression results for the buy and hold abnormal stock returns. Abnormal return is measured by the sample rms stock return minus the market return proxied by the Straits Times Index (STI) in Singapore. 2. The dependent variables XRET1, XRET2, XRET3 are the buy and hold abnormal stock returns compounding for 1, 2, and 3 years, respectively, post IPO. 3. The independent variables were dened as follows. Polconnect is a dummy variable that takes the value of 1 if the director is politically connected, 0 otherwise. It is proxied by politically connected chairmen/CEOs, current political appointees, former political appointees, members of Parliament, former ministers, and senior civil servants in models 1 to 6. Duality occurs when the chairman is also the CEO. Independent is the percentage of independent directors on the board. Dirage is the average age of the rms directors. Underpricing is computed as the difference between the rst days closing price and offer price, computed as a percentage of the offer price. Leverage is total debt divided by total assets. Exchange is a dummy variable taking on the value of 1 if the rm is listed on the main board of the SGX, 0 otherwise. GLC is a dummy variable taking on the value of 1 if the rm is a government-linked corporation, 0 otherwise. Market Value is the rst days closing price multiplied by the number of shares outstanding after the IPO. Industry 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 are dummy variables taking the value of 1 if the 2-digit SIC codes are 36, 67, 73, 87, 28, 15, 20, 30, 50, and 35, respectively; 0 otherwise. Source: Securities Data Corporation (SDC) Global New Issue database, IPO prospectus, and Datastream.
Table 9. Regression Results of Industry-adjusted Tobins Q

Model 2: Current Political Appointees
Model 1: Politically Connected Chairman or CEO
Intercept Polconnect Duality Independent Dirage Underpricing Leverage ROA Exchange GLC Market Value Reg Ind PolconnectReg Ind 0.022 1.46 262 0.014 1.26 234 0.009 1.15 208 0.018 1.39 267 0.015 1.32 262 0.019 1.35 234 0.010 1.17 208
AQ_post1yr 3.154 0.318 0.191 0.217 0.950 0.160 0.002 0.063 0.020 0.196 0.120 0.040 0.008
AQ_post2yr 0.747 0.353 0.256 0.755 0.417 0.237 0.185 0.043 0.207 0.188 0.123 0.029 0.335
AQ_post3yr 0.045 0.431 0.191 1.684 0.188 0.181 0.026 0.133 0.049 0.172 0.177 0.216 0.417
AQ_median 1.238 0.274 0.295 2.175 0.286 0.144 0.069 0.182 0.078 0.053 0.124 0.179 0.487
AQ_post1yr 3.228 0.031 0.161 0.256 0.970 0.150 0.002 0.068 0.024 0.211 0.120 0.068 0.044
AQ_post2yr 1.042 0.030 0.229 0.784 0.492 0.250 0.180 0.048 0.202 0.230 0.123 0.053 0.307
AQ_post3yr 0.082 0.273 0.132 1.709 0.220 0.220 0.014 0.148 0.035 0.087 0.182 0.134 0.279
AQ_median 1.446 0.168 0.255 2.215 0.335 0.165 0.055 0.188 0.085 0.094 0.122 0.136 0.393
0.023 1.49 267
Model 3: Former Political Appointees AQ_post3yr 0.004 0.083 0.170 1.707 0.183 0.216 0.015 0.143 0.050 0.194 0.180 0.191 0.463 0.011 1.21 234 0.007 1.12 208 0.024 1.53 267 AQ_median 1.534 0.010 0.277 2.239 0.353 0.158 0.063 0.186 0.076 0.027 0.120 0.162 0.479 AQ_post1yr 3.418 0.172 0.131 0.257 1.039 0.151 0.002 0.069 0.019 0.184 0.123 0.107 0.190
Model 4: Members of Parliament AQ_post2yr 1.030 0.0004 0.227 0.783 0.491 0.252 0.182 0.049 0.204 0.218 0.123 0.057 0.296 0.015 1.32 262 AQ_post3yr 0.277 0.045 0.147 1.723 0.262 0.210 0.009 0.144 0.046 0.187 0.179 0.160 0.340 0.010 1.18 234 AQ_median 1.525 0.020 0.280 2.244 0.349 0.157 0.064 0.186 0.075 0.021 0.120 0.165 0.489 0.007 1.12 208
Continued.
Intercept Polconnect Duality Independent Dirage Underpricing Leverage ROA Exchange GLC Market Value Reg Ind PolconnectReg Ind 0.015 1.32 262
AQ_post1yr 3.403 0.107 0.148 0.295 1.034 0.157 0.002 0.068 0.025 0.225 0.123 0.091 0.131
AQ_post2yr 1.047 0.018 0.224 0.789 0.498 0.251 0.182 0.049 0.204 0.219 0.124 0.061 0.278
0.023 1.49 267
Table 9. Continued.
Model 6: Senior Civil Servants AQ_median 1.608 0.230 0.258 2.279 0.380 0.147 0.056 0.186 0.073 0.026 0.119 0.143 0.421 0.010 1.16 208 0.025 1.54 267 0.019 1.41 262 0.011 1.20 234 0.010 1.15 208 AQ_post1yr 3.252 0.402 0.173 0.268 0.977 0.147 0.002 0.072 0.029 0.188 0.124 0.082 0.096 AQ_post2yr 1.010 0.291 0.235 0.795 0.487 0.255 0.190 0.052 0.202 0.194 0.126 0.069 0.236 AQ_post3yr 0.257 0.207 0.161 1.735 0.248 0.212 0.010 0.148 0.052 0.182 0.180 0.164 0.336 AQ_median 1.641 0.281 0.281 2.246 0.378 0.162 0.069 0.191 0.072 0.051 0.123 0.148 0.421
Model 5: Former Ministers AQ_post3yr 0.190 0.020 0.155 1.730 0.235 0.209 0.008 0.144 0.049 0.199 0.179 0.171 0.376 0.009 1.17 234
Intercept Polconnect Duality Independent Dirage Underpricing Leverage ROA Exchange GLC Market Value Reg Ind PolconnectReg Ind 0.017 1.35 262
AQ_post1yr 3.271 0.073 0.167 0.247 0.973 0.146 0.002 0.068 0.025 0.221 0.119 0.059 0.013
AQ_post2yr 1.097 0.127 0.235 0.763 0.500 0.258 0.189 0.049 0.202 0.218 0.122 0.048 0.332
0.019 1.40 267
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test) using White heteroskedasticityconsistent t-statistics. Notes: 1. This table presents the regression results for the abnormal rm value. The industry-adjusted Tobins Q is measured by the sample rms Tobins Q minus the industry median Tobins Q. The industry classication is based on the two-digit SIC code. 2. Dependent variables AQ_post1yr, AQ_post2yr, and AQ_post3yr are the industry-adjusted Tobins Q estimates 1, 2, and 3 years after an IPO. AQ_median is the median industry-adjusted Tobins Q estimate over the 3 years. 3. The independent variables were dened as follows. Polconnect is a dummy variable that takes the value of 1 if the director is politically connected, 0 otherwise. It is proxied by politically connected chairmen/CEOs, current political appointees, former political appointees, members of Parliament, former ministers, and senior civil servants in models 1 to 6. Duality occurs when the chairman is also the CEO. Independent is the percentage of independent directors on the board. Dirage is the average age of the rms directors. Underpricing is computed as the difference between the rst days closing price and offer price, computed as a percentage of the offer price. Leverage is total debt divided by total assets. ROA is the return on assets prior to the IPO. Exchange is a dummy variable taking on the value of 1 if the rm is listed on the main board of the SGX, 0 otherwise. GLC is a dummy variable taking on the value of 1 if the rm is a government-linked corporation, 0 otherwise. Market Value is the rst days closing price multiplied by the number of shares outstanding after the IPO. Reg Ind is a dummy variable representing the more highly regulated industries including electrical and electronic equipment (SIC 36), holding and other investment ofces (SIC 67), general building contractors (SIC 15), food and kindred products (SIC 20), and rubber and miscellaneous plastics products (SIC 30). PolconnectReg Ind is the interaction term of Polconnect and Reg Ind. Source: Securities Data Corporation (SDC) Global New Issue database, IPO prospectus, and Thomson One Banker database.
Table 10. Regression Results of Abnormal Stock Returns

XRET3 Model 2: Current Political Appointees 0.478 0.021 0.062 0.114 0.300 0.122 0.0003 0.088 0.040 0.117 0.008 0.007 0.154 6.21 316 Model 5: Former Ministers 0.454 0.116 0.064 0.123 0.317 0.125 0.0003 0.085 0.035 0.120 0.0002 0.137 1.896 0.037 0.017 0.453 0.084 0.215 0.0004 0.002 0.061 0.193 0.137 0.118 2.090 0.097 0.210 0.317 0.071 0.311 0.016 0.212 0.230 0.222 0.289 0.107 0.087 3.21 255 0.082 2.34 165 0.181 7.34 316 0.096 3.45 255 1.906 0.014 0.010 0.452 0.094 0.213 0.0004 0.0002 0.062 0.197 0.151 0.130 2.098 0.061 0.194 0.035 0.091 0.309 0.016 0.208 0.219 0.227 0.298 0.080 0.222 0.030 0.076 0.078 0.369 0.127 0.0003 0.092 0.015 0.117 0.055 0.252 1.991 0.135 0.014 0.438 0.057 0.204 0.0004 0.021 0.065 0.192 0.193 0.334 2.458 0.225 0.222 0.274 0.041 0.299 0.009 0.198 0.235 0.219 0.315 0.120 0.093 2.53 165 Model 6: Senior Civil Servants 0.483 0.0002 0.064 0.103 0.299 0.124 0.0003 0.088 0.039 0.117 0.003 0.121 1.933 0.204 0.013 0.439 0.075 0.205 0.0004 0.011 0.063 0.192 0.147 0.207 2.105 0.036 0.208 0.285 0.102 0.328 0.038 0.220 0.237 0.232 0.304 0.528
Continued.
XRET1 Model 3: Former Political Appointees
XRET2
XRET1
XRET2
XRET3
XRET1
XRET2
XRET3
Model 1: Politically Connected Chairmen/CEOs 2.192 0.218 0.204 0.334 0.052 0.306 0.003 0.211 0.233 0.226 0.294 0.103 0.083 2.35 165
Intercept Polconnect Duality Independent Dirage Underpricing Leverage Exchange GLC Market Value Reg Ind PolconnectReg Ind
0.487 0.025 0.062 0.108 0.297 0.124 0.0003 0.089 0.040 0.117 0.008 0.004
1.962 0.102 0.016 0.456 0.071 0.210 0.0004 0.006 0.063 0.194 0.145 0.120
0.154 6.20 316
0.087 3.21 255
Model 4: Members of Parliament 2.220 0.106 0.211 0.261 0.044 0.312 0.003 0.213 0.234 0.225 0.306 0.082
Intercept Polconnect Duality Independent Dirage Underpricing Leverage Exchange GLC Market Value Reg Ind PolconnectReg Ind
0.301 0.010 0.077 0.065 0.351 0.126 0.0003 0.093 0.020 0.119 0.053 0.216
1.923 0.077 0.002 0.461 0.081 0.202 0.0005 0.009 0.063 0.195 0.194 0.293
Table 10. Continued.

XRET3 Model 5: Former Ministers 0.158 6.38 316 0.088 3.22 255 0.083 2.35 165 0.155 6.25 316 0.089 3.26 255 XRET1 XRET2 XRET3 XRET1 XRET2 XRET3 Model 6: Senior Civil Servants 0.084 2.37 165
XRET1
XRET2
Model 4: Members of Parliament 0.084 2.37 165
0.172 6.96 316
0.094 3.40 255
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test) using White heteroskedasticityconsistent t-statistics. Notes: 1. This table presents the regression results for the buy and hold abnormal stock returns. The abnormal return is measured by the sample rms stock return minus the market return proxied by the Straits Times Index (STI) in Singapore. The dependent variables XRET1, XRET2, XRET3 are the buy and hold abnormal stock returns compounding in post-IPO 1, 2, and 3 years, respectively. 2. The independent variables were dened as follows. Polconnect is a dummy variable that takes the value of 1 if the director is politically connected, 0 otherwise. It is proxied by politically connected chairmen/CEOs, current political appointees, former political appointees, members of Parliament, former ministers, and senior civil servants in Models 1 to 6. Duality occurs when the chairman is also the CEO. Independent is the percentage of independent directors on the board. Dirage is the average age of the rms directors. Underpricing is computed as the difference between the rst days closing price and offer price, computed as a percentage of the offer price. Leverage is total debt divided by total assets. Exchange is a dummy variable taking on the value of 1 if the rm is listed on the main board of the SGX, 0 otherwise. GLC is a dummy variable taking on the value of 1 if the rm is a government-linked corporation, 0 otherwise. Market Value is the rst days closing price multiplied by the number of shares outstanding after the IPO. Reg Ind is a dummy variable representing the more highly regulated industries including electrical and electronic equipment (SIC 36), holding and other investment ofces (SIC 67), general building contractors (SIC 15), food and kindred products (SIC 20), and rubber and miscellaneous plastics products (SIC 30). PolconnectReg Ind is the interaction term of Polconnect and Reg Ind. Source: Securities Data Corporation (SDC) Global New Issue database, IPO prospectus, and Datastream.
160 ASIAN DEVELOPMENT REVIEW Table 11. Board Composition of Regulated Industries
Politically Connected Board size Duality Independent directors Independent directors/board size Executive directors Executive directors/board size Director age Management ownership GLC
Nonpolitically Connected Mean 6.891 0.598 2.424 0.360 3.239 0.461 46.500 0.393 0.022 Median 6.000 1.000 2.000 0.333 3.000 0.500 46.000 0.343 0.000
Difference Mean 0.194 0.143 0.030 0.003 0.360 0.021 1.667 0.016 0.009 Median 1.000 1.000 0.000 0.000 0.000 0.100 2.000 0.013 0.000
Mean 6.697 0.455 2.394 0.363 2.879 0.440 48.167 0.409 0.030
Median 7.000 0.000 2.000 0.333 3.000 0.400 48.000 0.330 0.000
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test), GLC = government-linked corporation. Notes: 1. This table reports the descriptive statistics of board composition for politically connected and nonpolitically connected rms in the more highly regulated industries. 2. More highly regulated industries include electrical and electronic equipment (SIC 36), holding and other investment ofces (SIC 67), general building contractors (SIC 15), food and kindred products (SIC 20), and rubber and miscellaneous plastics products (SIC 30). 3. t-statistics and z-statistics based on Wilcoxon signed ranks test were used to test for mean and median differences, respectively. Source: Securities Data Corporation (SDC) Global New Issue database and IPO prospectus.
We nd from the Tobins Q results in Table 9 that rms with politically connected chairmen or CEOs are associated with a signicant positive value effect. From Table 10, senior civil servants contribute to a signicant positive abnormal return 3 years post-IPO. All other forms of political connection have either very weak or no impact on rm value. Further analysis of the ve regulated industries (see Table 11) reveals that, compared to rms that have no political connection within the same industries, PC rms have a smaller board size, less occurrence of duality in the chairman and CEO functions, a larger independent directors-to-board-size ratio, a smaller percentage of executive directors, slightly older directors, a larger percentage of management ownership, and a higher proportion of GLCs. These differences, though not statistically signicant, provide anecdotal evidence of the composition of PC boards. We further investigate the impact from a major eventa change in Singapores Code of Corporate Governance, issued 14 July 2005, requiring all listed companies to disclose their corporate governance practices and explain deviations from the Code in their annual reports for annual general meetings held from 1 January 2007 onwards. Our results on the announcement effect reveal that PC directors do not add value to their rms both before and after the imposition of the new disclosure requirement. However, as shown in Table 12, after controlling for the presence of PC directors, rms in the more highly regulated industries appear to have a statistically signicant impact on rm value as measured by the industry-adjusted Tobins Q.

Table 12. Regression Analysis under the New Corporate Governance Code
Model 1 Intercept Polconnect Duality Independent Dirage Underpricing Leverage Exchange GLC Market Value Reg Ind Event Adj. R2 F-Statistic N
Model 2 Pre-event 4.979 (1.64) 0.767 (1.07) 0.264 (1.89) 0.382 (0.63) 1.533 (2.19) 0.072 (0.45) 0.468 (1.34) 0.074 (0.53) 0.024 (0.14) 0.130 (2.05) 0.026 (0.21) 0.022 1.50 223
Model 3 Post-event 0.026 (0.01) 1.892 (0.63) 0.433 (1.17) 0.584 (0.26) 1.172 (0.92) 1.585 (5.30) 0.001 (1.08) 0.594 (1.51) 1.993 (3.51) 0.363 (2.27) 1.024 (1.96) 0.123 1.39 29
Full Period 5.553 (2.01) 0.841 (1.20) 0.190 (1.32) 0.210 (0.35) 1.758 (2.63) 0.070 (0.48) 0.002 (3.31) 0.019 (0.14) 0.161 (0.73) 0.121 (2.06) 0.048 (0.39) 0.102 (0.41) 0.027 1.63 252
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test) using White heteroskedasticity-consistent t-statistics. Notes: 1. This table presents the regression results for the abnormal rm value during the pre-event and post-event. The event is the new corporate governance code adopted in 14 July 2005. 2. The industry-adjusted Tobins Q is measured by the sample rms Tobins Q minus the industry median Tobins Q. The industry classication is based on the 2-digit SIC code. 3. The dependent variable AQ_post1yr is the industry-adjusted Tobins Q a year after an IPO. 4. The independent variables were dened as follows. Polconnect is a dummy variable that takes the value of 1 if the director is politically connected, 0 otherwise. It is proxied by politically connected chairmen/CEOs, current political appointees, former political appointees, members of Parliament, former ministers, and senior civil servants in Models 1 to 6. Duality occurs when the chairman is also the CEO. Independent is the percentage of independent directors on the board. Dirage is the average age of the rms directors. Underpricing is computed as the difference between the rst days closing price and offer price, computed as a percentage of the offer price. Leverage is total debt divided by total assets. Exchange is a dummy variable taking on the value of 1 if the rm is listed on the main board of the SGX, 0 otherwise. GLC is a dummy variable taking on the value of 1 if the rm is a government-linked corporation, 0 otherwise. Market Value is the rst days closing price multiplied by the number of shares outstanding after the IPO. Reg Ind is a dummy variable representing the more highly regulated industries including electrical and electronic equipment (SIC 36), holding and other investment ofces (SIC 67), general building contractors (SIC 15), food and kindred products (SIC 20), and rubber and miscellaneous plastics products (SIC 30). Event is a dummy variable that takes the value of 1 if the IPO was issued under the new corporate governance code regime (14 July 2005), 0 otherwise. Source: Securities Data Corporation (SDC) Global New Issue database, IPO prospectus, Thomson One Banker database, and Singapore Code of Corporate Governance 2005.
We provide results of a logistic regression in Table 13 to reect the demand for politically connected directors. The coefcient for director age is found to be positively signicant. This means that PC rms have directors that are more experienced in guiding their rms than non-PC rms. PC rms also appear to have stronger corporate governance as evidenced by their lower likelihood of duality in their chairman and CEO functions and a smaller percentage of executive directors. On the whole, the results show that, in Singapore, political connection in a non-corrupt regime in general does not affect rm value. However, rms in industries that are more highly regulated appear to receive some benet in terms of higher rm valuation from their political connections. In particular, companies with a politically connected chairman/CEO or senior civil servant on their board benet the most in terms of a positive effect on rm value. PC rms, by virtue of their being perceived as having good governance under a low corruption environment where government regulations are followed, such companies may be attractive business partners to other businesses and governments.
VI. Summary and Conclusions
Political connection is a double-edged sword. On the one hand, rm value can be jeopardized if exploiting such connections distorts incentives, misallocates investment, and increases the extent of corruption (Shleifer and Vishny 1994). On the other hand, politically connected companies may benet through easier access to debt nancing, lower taxes, and stronger market power (Faccio 2006). Although the effect of political connections has been previously investigated, our contribution is to examine for the rst time the impact of politically connected directors on the value of rms operating in an environment where the perceived level of corruption is comparatively low due to better institutions and not confounded by favorable social and cultural factors. Singapore is ideal for the study as it is not only among the countries with the lowest level of corruption, but it is also free of confounding factors such as an innate culture among countries with the lowest political corruption. We study this issue at a rms inception (i.e., at IPO), where most of the new politically connected directors are appointed at the same time and their impact, if any, has not been anticipated as would be the case at any arbitrary time after an IPO. We study the effect of political connection on the value of a company in a low political corruption environment by examining the industry-adjusted Tobins Q and the excess returns of newly-listed companies from 1998 to 2006. Firm value, within 3 years from the issue of an IPO, is found largely to be independent of a rms political connection, even after controlling for differences in corporate governance and rm characteristics. Importantly, such connections do not reduce rm value. However, when the sample is broken down in terms of industry, we nd that rms operating in a more highly regulated environment appear to receive some benet

Table 13. Logistic Regression Results of Politically Connected Directors
Model 1 Intercept Market Value ROA Leverage Underpricing GLC Exchange Board Size Duality Dirage Independent IO Industry1 Industry2 Industry3 Industry4 Industry5 Industry6 Industry7 Industry8 Industry9 Industry10 Cox & Snell R2 2 Log Likelihood N
Model 2 0.273 0.236 0.065 0.011 0.273 0.116 0.478 0.124 0.538 0.087 0.106 0.005
Model 3 0.081 0.253 0.073 0.014 0.324 0.181 0.503 0.128 0.589 0.091 0.041 0.005 0.011 0.316 0.395 0.584 0.551 0.740 0.215 0.234 0.026 1.639 0.101 297.34 429
2.129 0.193 0.043 0.012 0.114 0.295 0.063
0.017 407.62 429
0.077 304.89 429
= signicance at the 10% level (two-tailed test), = signicance at the 5% level (two-tailed test), = signicance at the 1% level (two-tailed test) using Wald z-statistics. Notes: 1. This table presents the results of the demand for politically connected directors. The dependent variable is the dummy variable that takes the value of 1 if the IPO is politically connected and 0 otherwise. 2. The independent variables were dened as follows: Market Value is the rst days closing price multiplied by the number of shares outstanding after the IPO. ROA is the return on assets prior to the IPO. Leverage is total debt divided by total assets. Underpricing is computed as the difference between the rst days closing price and offer price, computed as a percentage of the offer price. GLC is a dummy variable taking on the value of one if the rm is a government-linked corporation and zero if otherwise. Exchange is a dummy variable taking on the value of 1 if the rm is listed on the main board of the SGX, 0 otherwise. Board Size is the total number of board members. Duality occurs when the chairman is also the CEO. Dirage is the average age of the rms directors. Independent is the percentage of independent directors on the board. IO is the institutional ownership of the IPO. Industry 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 are dummy variables taking the value of 1 if the 2-digit SIC codes are 36, 67, 73, 87, 28, 15, 20, 30, 50, and 35, respectively; 0 otherwise. Source: Securities Data Corporation (SDC) Global New Issue database and IPO prospectus.
from their political connection in terms of a higher rm valuation, especially among those with a politically connected chairman/CEO or senior civil servant on their board. As PC directors have been shown to be associated with better governance practices, their presence on corporate boards may thus compel rms to better adhere to government regulations. Thus, because PC rms are perceived as having good governance under a low corruption environment and strictly observe government regulations, other rms and governments may be attracted to forming business relationships with them. We nd that rms with politically connected directors are associated with good governance practices such as nonduality in their chairman and CEO and fewer executive directors. We show that PC rms have higher managerial ownership and tend to be smaller than non-PC rms. Firms with such characteristics may typically be expected to have a lower valuation compared to those that are larger or have a lower percentage of management ownership. Our results show that the presence of PC directors appears to have somewhat neutralized any negative effects with its association with better corporate governance. Thus, having a politically connected director on the board may be used as a signal by a rm to investors that it is subject to adequate appropriate monitoring mechanisms. As the supply of good candidates for outside independent directors is often limited, rms would be inclined to invite such politicians to serve on their board, not because of their political connections and potential economic payoffs but as a means of expanding their list of good candidates for independent directors. In conclusion, the ndings of this study suggest that, despite the lack of strong economic value that political connection per se brings to rms, politically connected directors may still contribute to the rm when they serve as independent directors within a noncorrupt political environment. We provide evidence that companies in certain highly regulated industries may nd that having certain types of directors who are politically connected could have a positive and signicant impact on their rms value.
References
Agrawal, Anup, and Charles Knoeber. 2001. Do Some Outside Directors Play a Political Role? Journal of Law and Economics 44(1): 17998. Ang, James, and Carol Boyer. 2007. Finance and Politics: The Wealth Effects of Special Interest Group Inuence during the Nationalization and Privatization of Conrail. Cambridge Journal of Economics 31(2): 193216. Ang, James, and David Ding. 2006. Government Ownership and the Performance of GovernmentLinked Companies: The Case of Singapore. Journal of Multinational Financial Management 16(1): 6488. Black, Bernard, Hasung Jang, and Woochan Kim. 2006. Does Corporate Governance Affect Firm Value? Evidence from Korea. Journal of Law, Economics, and Organization 22(2): 366413.

Chong, Alberto, and Florencio Lopez-de-Silanes. 2006. Corporate Governance and Firm Value in Mexico. IADB Working Paper No. 564. Washington, DC: Inter-American Development Bank. Chung, Kee, and Stephen Pruitt. 1994. A Simple Approximation of Tobins Q. Financial Management 23(3): 7074. Claessens Stijn, Simeon Djankov, and Larry Lang. 2000. The Separation of Ownership and Control in East Asian Corporations. Journal of Financial Economics 58(12): 81112. della Porta, Donatella, and Alberto Vannucci. 1997. The Perverse Effects of Political Corruption. Political Studies 45(3): 51638. Faccio, Mara. 2006. Politically Connected Firms. American Economic Review 96(1):369386. Faccio, Mara, Ronald Masulis, and John McConnell. 2006. Political Connections and Corporate Bailouts. Journal of Finance 61(6): 25972635. Fan, Joseph, T. J. Wong, and Tianyu Zhang. 2007. Politically Connected CEOs, Corporate Governance, and Post-IPO Performance of Chinas Newly Partially Privatized Firms. Journal of Financial Economics 84(2): 33057. Ferguson, Thomas, and Hans-Joachim Voth. 2008. Betting on HitlerThe Value of Political Connections in Nazi Germany. Quarterly Journal of Economics 123(1): 10137. Fisman, Raymond. 2001. Estimating the Value of Political Connections. American Economic Review 91(4): 10921102. Friedman, Thomas. 1999. The Lexus and the Olive Tree. London: Harper Collins. Goldman, Eitan, Jorg Rocholl, and Jongil So. 2009. Do Politically Connected Boards Affect Firm Value? Review of Financial Studies 22(6): 233160. Hellman, Joel, Geraint Jones, and Daniel Kaufmann. 2000. Seize the State, Seize the Day. State Capture, Corruption and Inuence in Transition. Policy Research Working Paper No. 2444. Washington, DC: World Bank. Imai, Masami. 2006. Mixing Family Business with Politics. Asian Economic Journal 20(3): 24156. Johnson, Simon, Kalpana Kochhar, Todd Mitton, and Natalia Tamirisa. 2006. Malaysian Capital Controls: Macroeconomics and Institutions. IMF Working Paper No. WP/06/51. Washington, DC: International Monetary Fund. Khwaja, Asim Ijaz, and Atif Mian. 2004. Do Lenders Favor Politically Connected Firms? Rent Provision in and Emerging Financial Market. Quarterly Journal of Economics 120(4): 13711411. Leuz, Christian, and Felix Oberholzer-Gee. 2006. Political Relationships, Global Financing, and Corporate Transparency: Evidence from Indonesia. Journal of Financial Economics 81(2): 41139. Li Donghui, Fariborz Moshirian, Pascal Nguyen, and Li-wen Tan. 2007. Corporate Governance and Globalization: An Analysis of CEO Compensation in China. Research in International Business and Finance 21(1): 3249. Morck, Randall, Andrei Shleifer, and Robert Vishny. 1998. Management Ownership and Market Valuation: An Empirical Analysis. Journal of Financial Economics 20(12): 293316. Phan, Phillip, and Toru Yoshikawa. 2005. Corporate Governance in Singapore: Developments and Prognoses. Working paper. Singapore Management University, Singapore. Prowse, Stephen. 1998. Corporate Governance in East Asia: A Framework for Analysis. Working paper. Federal Reserve Bank of Dallas, Texas. Ritter, Jay. 1991. The Long-run Performance of Initial Public Offerings. Journal of Finance 46(1): 327.
166 ASIAN DEVELOPMENT REVIEW Ritter, Jay, and Ivo Welch. 2002. A Review of IPO Activity, Pricing, and Allocations. Journal of Finance 57(4): 17951828. Sapienza, Paola. 2004. The Effects of Government Ownership on Bank Lending. Journal of Financial Economics 72(2): 35784. Shleifer, Andrei, and Robert Vishny. 1994. Politicians and Firms. Quarterly Journal of Economics 109(4): 9951025. . 1997. A Survey of Corporate Governance. Journal of Finance 52(2): 73783. Xu, Lixin, Tian Zhu, and Yi-min Lin. 2002. Politician Control, Agency Problems and Ownership Reform: Evidence from China. Working paper. World Bank, Washington, DC. Yermack, David. 1996. Higher Market Valuations for Firms with a Small Board of Directors. Journal of Financial Economics 40(2): 185211.
2013 LIST OF REFEREES

Prema-chandra Athukorala Australian National University Pranab Bardhan University of California Berkeley Carlos Bautista University of the Philippines Loren Brandt University of Toronto Natalie Chun Asian Development Bank Emmanuel de Dios University of the Philippines Alissa di Caprio Asian Development Bank Pascaline Dupas Stanford University Avraham Yehuda Ebenstein Hebrew University of Jerusalem Shinichi Fukuda University of Tokyo Stephen Haggard University of California San Diego Akiko Terada Hagiwara Asian Development Bank Robert Hammond North Carolina State University Xuehui Han Asian Development Bank Vu Quoc Huy Vietnam National University Soyoung Kim Seoul National University Akira Kohsaka Osaka University Nimfa Mendoza University of the Philippines Suresh Narayanan Universiti Sains Malaysia Marcus Noland Petersen Institute Takaaki Nomoto Asian Development Bank John V.C. Nye George Mason University Nina Pavcnik Dartmouth College Sun Qian Fudan University Agnes Quisumbing International Food Policy Research Institute Eric D. Ramstetter International Centre for the Study of East Asian Development Eli Remolona Bank for International Settlements Hong Kong Eric Sidgwick Asian Development Bank Steven Sheffrin Tulane University Hyun H. Son Asian Development Bank Tan Eu Chye University of Malaya Doo Yong Yang Kyung Hee University Wu Yanrui University of Seattle Kar Yiu Wong Washington University in Seattle Lixin Colin Xu World Bank
167
CALL FOR PAPERS
Submissions are welcome on the following topics relevant to Asias development: growth (including inclusive growth, escaping the middle-income trap, new or alternative growth and development paradigms) institutions and development inequality and income distribution urbanization and rural development nancial sector development macroeconomic management regional integration and cooperation hard and soft infrastructure To submit manuscripts, please write to: Co-Managing Editors Asian Development Review Economics and Research Department, Asian Development Bank 6 ADB Avenue, Mandaluyong City 1550 Metro Manila Philippines E-mail: asiandevreview@adb.org
168
INSTRUCTIONS FOR AUTHORS

Editorial Ofce of the Asian Development Review Co-Managing Editors Economics and Research Department, Asian Development Bank 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines E-mail: asiandevreview@adb.org
Authors should submit the nal version of the text le, along with any separate art les, in electronic form. Microsoft Word is preferred. Graphs, line drawings, and diagrams (art les) should be separate TIFF or EPS les, if possible, or Excel les, and in black and white only. Address them to the Co-Managing Editors at asiandevreview@adb.org All articles submitted for publication must represent original work not previously published and should not be under current review by another publisher. 1. Manuscripts should be single spaced throughout and single sided. Text should be unjustied, with 1.25-inch margins on all sides. 2. All pages should be numbered consecutively. Order of the manuscript: title page that includes the name of authors, institutional afliations, acknowledgments, JEL classication codes, keywords, abstract (approximately 100 words), followed by the main text of paper including footnotes, appendix (if applicable), references, tables, and gures. 3. Use 11-pt Times New Roman throughout the main text (10-pt is acceptable for gures and tables). Observe a maximum of three heading levels. Only when absolutely necessary may a fourth heading level be used. 4. Acronyms should be spelled out the rst time they are used in the text. 5. Use italics for scalar variables, use boldface to specify vectors and matrices, and use script for sets. All other letter symbols will appear as italics in print (except log, cov, max, exp, lim, In, var, etc.). 6. Equations should be numbered consecutively, and the equation numbers should be in parentheses ush right at the end of the line. Numbers between 1 and +1 should have 0 before the decimal point. 7. Present all notes as footnotes. Keep footnotes to a minimum, ensuring that they carry substantive related material. Do not place reference details in the footnotes, rather present all bibliographic details in a Reference List. Number footnotes consecutively throughout the text with Arabic numerals. Use a size 9 point for footnotes. 8. Figure titles and captions should not be drawn on the gurethey will be typeset at the time of printing and appear with the gure, ush left. Figures must be able to t on page size (12 17 cm). Tables should contain only essential data and should not have more than 8 columns. 9. Citations in the text or footnotes should give only the name of the author(s), year of publication, and possibly a page number or chapter number, as follows: Hayek (1960, p. 200), or (Cornwall 1965a, 1965b; Watts 1999).
169
10. References to authors in the text must exactly match those in the reference section. Citations should take the place of footnotes whenever possible. The section for references, which should follow the main text and begin on a new page, should have complete, veried information including the rst names (not just the initials) and surnames of the authors and journal names in full (no abbreviations). No item should appear in the references if it has not been cited in the text. The reference list should be organized alphabetically by authors last name and then by year, in ascending order. References should conform to the following examples: Frenkel, Jacob A. and Richard M. Levich. 1975. Covered Interest Arbitrage: Unexploited Prots? Journal of Political Economy 83:325338. Bhagwati, Jagdish, and Robert Hudec, eds. 1996. Fair Trade and Harmonization: Prerequisites for Free Trade? Vols. 1 and 2. Cambridge, Massachusetts: MIT Press. Hufbauer, Gary Clyde, and Jeffrey J. Scott. 1992. North American Free Trade. Washington, DC: Institute for International Economics. Otsuka, Keijiro. 1993. Land Tenure and Rural Poverty. In M. G. Quibria, ed. Rural Poverty in Asia: Priority Issues and Policy Options. Hong Kong: Oxford University Press for the Asian Development Bank. World Resources Institute. 1996. World Resources 1996-97 (On-line). Available: www.wri. org 11. Authors will receive page proofs by e-mail for correction, which must be returned within 72 hours of receipt. Authors publishing in the journal will be asked to sign an Agreement to Publish and Transfer Copyright. In signing the form it is assumed that authors have obtained permission to use any copyrighted or previously published material. All authors must read and agree to the conditions outlined in the form, and must sign the form or agree that the corresponding author can sign on their behalf.
170
Asian Development Review (ISSN 0116-1105, E-ISSN 1996-7241) is published twice a year (March, September) by the MIT Press, 55 Hayward Street, Cambridge MA 02142-1315, USA. A free electronic full-text version of Asian Development Review is available from the MIT Press. Vol. 30 No. 2 2013, Asian Development Bank and Asian Development Bank Institute Publication Stock No. JRN 135934-3 ADB and ADBI encourage printing or copying information exclusively for personal and noncommercial use with proper acknowledgment. Users are restricted from reselling, redistributing, or creating derivative works for commercial purposes without the express, written consent of MIT Press Journals. Please address inquiries to the Permissions Department at journals-rights@mit.edu.
Subscription Information Subscription and address changes should be addressed to MIT Press, Journals Customer Service, 55 Hayward Street, Cambridge MA 02142-1315; phone 617-253-2889, US/Canada 800-207-8354; fax 617-577-1545; e-mail: journals-cs@mit.edu. Subscriptions begin with the first issue of the volume. Subscription rates are: International/DMCs* $20.00, International/non-DMCs* $35.00, United States $35.00. Canadians add 5% GST. Outside the US and Canada add $17.00 for postage and handling. Single Issues: International/DMCs* $12.00, International/non-DMCs* $20.00, United States $20.00. Canadians add 5% GST. Outside the US and Canada add $9.00 per issue for postage and handling. *Purchases of the journal from within the Developing Member Countries (DMCs) of the Asian Development Bank (ADB) can be made at a reduced rate. Claims for missing issues will be honored free of charge if made within three months of the publication date of the issue. Claims may be submitted to: journals-cs@mit.edu. Prices are subject to change without notice. Send address changes to Asian Development Review, MIT Press Journals, 55 Hayward Street, Cambridge, MA 02142-1315.
Advertising and Mailing List Rentals To advertise within the pages of Asian Development Review, or on the ADR homepage of the MIT Press Journals website, or to rent the subscriber mailing list, contact the Marketing Department, MIT Press Journals, 55 Hayward Street, Cambridge, MA 02142-1315; phone: 617-253-2866; fax: 617-253-1709; e-mail: journals-ads@mitpress.mit.edu.
Asian Development Review The Asian Development Review is a professional journal for disseminating the results of economic and development research relevant to Asia. The journal seeks high-quality papers done in an empirically rigorous way. Articles are intended for readership among economists and social scientists in government, private sector, academia, and international organizations.

Asian Development Review: Volume 30, Number 2

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Asian Development Review: Volume 30, Number 2

Uploaded by

Copyright:

Available Formats

Asian Development Review

Asian Development Review

167 168 169

Demographic Dividends Revisited

I. Looking Backwards: The Demographic Dividend Convergence Model

Asian Development Review, vol. 30, no. 2, pp. 125

C 2013 Asian Development Bank and Asian Development Bank Institute

2 ASIAN DEVELOPMENT REVIEW

DEMOGRAPHIC DIVIDENDS REVISITED 3

Birth rate Death rate

Population growth rate

Birth rate Death rate Time

Population Growth and the Age Structure

Birth rate minus death rate

4 ASIAN DEVELOPMENT REVIEW

Table 1. Contribution of Demographic Change to Economic Growth, 19651990

Average Growth: Population (%)

3.33 6.11 3.80 1.71 0.97 2.83 0.85 1.61 1.97

DEMOGRAPHIC DIVIDENDS REVISITED 5

Source: Bloom and Williamson (1998), Table 6.

6 ASIAN DEVELOPMENT REVIEW

Table 2. Contribution of Demographic Change to Future Economic Growth, 19902025

1.36 0.43 1.29 1.65 2.40 0.17 1.50 1.28 1.08

Source: Bloom and Williamson (1998), Table 7.

DEMOGRAPHIC DIVIDENDS REVISITED 7

Growth rate of real GDP per capita

Other transitional forces Demographic gift

Youth demographic burden

Source: Bloom and Williamson (1998), Figure 6.

8 ASIAN DEVELOPMENT REVIEW

DEMOGRAPHIC DIVIDENDS REVISITED 9

10 ASIAN DEVELOPMENT REVIEW

DEMOGRAPHIC DIVIDENDS REVISITED 11

12 ASIAN DEVELOPMENT REVIEW

DEMOGRAPHIC DIVIDENDS REVISITED 13

V. Channels of Dividend Impact: Human Capital Accumulation and Vintage Effects

14 ASIAN DEVELOPMENT REVIEW

Endogenous Schooling from the Demand Side: Quality versus Quantity

DEMOGRAPHIC DIVIDENDS REVISITED 15

16 ASIAN DEVELOPMENT REVIEW

DEMOGRAPHIC DIVIDENDS REVISITED 17

18 ASIAN DEVELOPMENT REVIEW

VII. Beneath the Macro: RuralUrban Migration and Income Gaps

DEMOGRAPHIC DIVIDENDS REVISITED 19

VIII. The Demographic Transition and Inequality Connection

20 ASIAN DEVELOPMENT REVIEW

DEMOGRAPHIC DIVIDENDS REVISITED 21

DEMOGRAPHIC DIVIDENDS REVISITED 23

DEMOGRAPHIC DIVIDENDS REVISITED 25

I. Policy Distortions, Connections, and Inequality in Transition Economies

Asian Development Review, vol. 30, no. 2, pp. 2648

PRINCELINGS AND PAUPERS? 27

Kernel Density Estimate .00006

kernel = epanechnikov, bandwidth = 1501.76

PRINCELINGS AND PAUPERS? 29

30 ASIAN DEVELOPMENT REVIEW

PRINCELINGS AND PAUPERS? 31

II. A New Dimension of Inequality: State and Nonstate Households A. Data

Table 2. Average Years of Education of Workers in State and Nonstate Sectors

PRINCELINGS AND PAUPERS? 33

34 ASIAN DEVELOPMENT REVIEW

PRINCELINGS AND PAUPERS? 35

III. Determinants of Household Educational Investments A. Empirical Strategy

36 ASIAN DEVELOPMENT REVIEW Figure 3a. Net Enrollment Rate