American Economic Association

Does Aid Affect Governance?

Author(s): Raghuram Rajan and Arvind Subramanian
Source: The American Economic Review, Vol. 97, No. 2 (May, 2007), pp. 322-327
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/30034469
Accessed: 30-11-2015 16:34 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The American Economic
Review.

http://www.jstor.org

This content downloaded from 128.135.215.125 on Mon, 30 Nov 2015 16:34:29 UTC
All use subject to JSTOR Terms and Conditions
 Does AidAffectGovernance?
By RAGHURAMRAJAN AND ARVIND SUBRAMANIAN*
I. ChannelsfromAid to Growth
Why is there little robust evidence that for-
eign aid significantly enhances the economic
growth of poor countries?1The search for an
explanation is becoming immensely impor-
tant as industrial countries are being exhorted
to increase their aid budgets in order to help
developing countries achieve the Millennium
Development Goals.
Perhapsone should not expect an impact on
growthfromthe mere infusion of additionalcap-
ital into a country. But, perhaps,any beneficial
effects are offset by adverse spillover effects,
and academic focus should shift to determining
what these are and how to mitigatethem. In this
regard, Figure 1 is suggestive. We plot the log
of the manufacturingto gross domestic product
(GDP) ratio in a country against the log of the
ratio of aid received to GDP for that country
for two separate time periods (the late 1990s
and the early 1980s), conditional on a number
of variables. As the figure suggests, the more
aid a country has received, the smaller its share
of manufacturing.The coefficient estimate sug-
gests that a 1 percentage point increase in the
ratio of aid to GDP is associated with a reduced
share of manufacturingin total GDP of about
0.2 to 0.3 percentagepoints.2
*Rajan: Graduate School of Business, University of
Chicago, 5807 S. Woodlawn, Chicago IL 60637 (e-mail:
rajan@chicagogsb.edu); Subramanian:InternationalMone-
tary Fund, 700 19thStreet, Washington,DC 20431 (e-mail:
asubramanian@imf.org).
1 See, for example, William Easterly (2003), Easterly,
Ross Levine, and David Roodman (2004), Henrik Hansen
and Finn Tarp (2001), Roodman (2004), and Rajan and
Subramanian(2005a).
2 The
conditioning variables include a country's per
capita purchasingpower parity (PPP) GDP, per capita PPP
GDP squared,and country and time fixed effects. Because
we include fixed effects, the association between aid and
manufacturing depicted in the chart is a temporal one,
that is, within countries over time ratherthan one between
countries.
322
This content downloaded from 128.135.215.125 on Mon, 30 Nov 2015 16:34:29 UTC
All use subject to JSTOR Terms and Conditions
Without further analysis, we do not know if
this relationshipis causal. If causal, this figure
offers a proximateexplanationfor why aid may
not have led to substantialgrowth. As pointed
out by Simon Johnson, Jonathan Ostry, and
Subramanian(2007); and BenjaminF.Jonesand
Benjamin A. Olken (2005), virtually all coun-
tries that have had a sustainedperiod of growth
in the postwarperiod have seen a large increase
in their share of manufacturingproductionand
manufacturingexports.
But what is the deep underlying cause?
Manufacturingis more complex than agricul-
ture or mining in that it requires many more
transactions between unrelated parties. Thus,
manufacturingis likely to be moredependenton
a good-governanceenvironmentthat can foster
multiple transactions.Here is where aid could
hurt. By expanding a government's resource
envelope, aid reduces the need for govern-
ments to explain their actions to citizens, which
may reduce its need to govern well (Deborah
A. Brautigam and Stephen Knack 2004). In
particular, the government could falter in its
responsibilityto maintain rule of law, ensure a
predictablejudiciary and contractenforcement,
and limit corruption. In a companion paper
(Rajan and Subramanian2005b), we examine
anotherchannelthroughwhich poor governance
could affect manufacturing-the mismanage-
ment of the real exchangerate.
We use the methodology in Rajanand Luigi
Zingales (1998) to test the hypothesis that aid
might reduce the quality of governance.They
suggest thatone way to check whethera channel
is at workis to see whetherindustriesthatmight
be most affected by a channel grow differen-
tially (faster or slower depending on the nature
of the effect) in countries where that channel is
likely to be more operative.The industrychar-
acteristic we are interested in is the degree to
which an industry depends on governance
(hereafterreferredto as "governance"sensitiv-
ity or dependence).The channelis the quality of
governance,and countriesthat get more aid are
likely to be the ones where the channel is most
VOL.97 NO. 2 DOES AID AFFECT GOVERNANCE?
adversely affected. The estimation strategy is
then to run regressionsof the form
(1) Growthij= Constant
+ Indicators
(*1.....m*Country
+ ,m+1....n*IndustryIndicators
+ yn+,* (Industryi's share of
manufacturingin countryj
in the initial period)
+ a (Aid to countryj* Sensitivityof
industryi to governance)
+ Eij,
where Growthijis the annual average rate of
growth of value addedof industryi in countryj
over the ten-yearperiod 1981-1990, obtainedby
normalizingthe growthin nominal value added
by the GDP deflator; 1.....m are the coefficients
of the country fixed effects; are the coef-
ficients of the industry fixedm+m1....n
effects;3 and yn+1
is the coefficient of the initial period share
of industry i in total value added in countryj
(which controls for convergence-type effects).
Aid to countryj is the averageaid-to-GDPratio
for that country over the sample period. The
coefficient of interestfor us is a. It capturesan
interactionbetween a country-specificaid vari-
able and an industry'ssensitivityto governance.
We posit countries that receive more aid should
see a more negative impact in industrialsectors
that are more governance sensitive. The chief
advantageof this strategyis that by controlling
for country and industryfixed effects, the prob-
lem of omitted variablesbias or model specifi-
cation, which seriously afflicts cross-country
regressions,is diminished.4
3 We focus on the 1980s because of the large sample
size. For the 1990s, data are available for only 15 develop-
ing countries.
4 Essentially, we are
making predictions about within-
country differences between industries based on an inter-
action between a country and industry characteristic.
Moreover,because we focus on differences between manu-
facturing industries (ratherthan between, say, manufactur-
ing and services industries), we can rule out factors that
would keep manufacturingunderdevelopedas explanations
This content downloaded from 128.135.215.125 on Mon, 30 Nov 2015 16:34:29 UTC
All use subject to JSTOR Terms and Conditions
323
II. Sensitivityto Governance
Details on data sources are availablefrom the
authors,while summary statistics and detailed
tables with robustness checks are available
in a longer version of this paper available at
Rajan'sWeb site (http://faculty.chicagogsb.edu/
raghuram.rajan/research/).The sample com-
prises all developing countries in the United
Nations Industrial Development Organization
(UNIDO) database(see the Web version of this
paperfor a list of countriesin the sample).
Clearly, the success of our method depends
on whetherwe have a plausiblemeasurefor the
dependence of an industry on the governance
environment. Olivier Blanchard and Michael
Kremer(1997) develop such a measure.Broadly
speaking, the fewer other industries an indus-
try buys from-that is, the more concentrated
its purchases-the more possibility it has of
regulating transactionsvia long-term,repeated
relationships or vertical integration, and less
possibility of its need to rely on explicit gover-
nance by the courts or regulatory authorities.
If the country has little governance capacity,
industries that depend on contracts or arm's-
length transactions(that is, score lower on the
concentrationindex) are likely to either have to
distorttheir organizationalstructureor transact
less, both of which will affect their growth.
The governance-dependencemeasure (from
Andre A. Levchenko 2006), which is based on
Blanchardand Kremer (1997), is calculated as
follows from US input-outputdata. For each
manufacturingindustry i, we obtained its inter-
mediate goods purchases from other industries
k. The Herfindahlindex for each sector i is cal-
culated as the sum of the squaresof the sharesof
the purchasesof each input k in total input pur-
chases of i. We multiplythe Herfindahlindex by
-1 to get the HERF measure,where a value of
zero indicates an industrythat is highly depen-
dent on public institutionsto govern its transac-
tions, while a value of -1 denotes an industry
that transacts fairly narrowly,and thus is less
dependenton public institutions.
Ideally, we would like to measure the num-
ber of outside transactionswith different firms
of our results-for these factors should not affect the differ-
ences between manufacturingindustries.
324 AEA PAPERSAND PROCEEDINGS MAY2007

TABLE 1--THE GOVERNANCE-DEPENDENCE INDEX, THE IMPACT OF AID ON INSTITUTIONS, AND ROBUSTNESS CHECKS
(Dependentvariable is the annual averagerate of growthof value addedof industry(i) in country(j)for the 1980s)

1 2 3 4 5 6
Estimation method OLS OLS IV IV IV IV
Initial industry share (ij) -0.370*** -0.452*** -0.509*** -0.494*** -0.416*** -0.366***
[0.069] [0.068] [0.074] [0.070] [0.068] [0.080]
Investmentprotection (j)* 1.194***
Governance-dependence(i) (Herfindahl)
[0.336]
Aid/GDP(j)* -2.126* -9.902*** -9.684** -8.583**
Governance-dependence(i) (Herfindahl) [1.146] [3.778] [4.066] [3.832]
Technical aid/GDP (j)* -41.831**
Governance-dependence(i) (Herfindahl) [16.791]
Initial income (j)* -0.231**
Governance-dependence(i) (Herfindahl) [0.092]
Aid/GDP (j) -1.395**
[0.592]
Initial income 1980s -0.007
[0.006]
SW trade policy index 0.059***
[0.007]
Observations 811 809 809 809 758 758
R-squared 0.46 0.46 0.43 0.44 0.45 0.13
Notes: All standarderrors reportedbelow the coefficient estimates are robust. All regressions include country and indus-
try effects, not reported above for presentationalsimplicity. Governance-dependence(Herfindahl)(i) is the negative of the
Herfindahlindex of concentrationof outside purchases for industry (i), calculated as the sum of the squareof the shares of
purchases of each input on the total input purchases. It ranges between 0 (institutions-intensive)and -1 (not institutions-
intensive). Initial industry share (ij) is the initial shareof the value added of industry(i) in the total value added of all manu-
facturing sectors in country (j). Investmentprotection(j) is the ICRGmeasureof investmentprofile in country(j) thatrelies
on risk assessments regardingcontractviability/expropriation,profitrepatriation,and paymentdelays. We use a normalized
0 to 1 version of the original 1 to 12 ICRGmeasure (averageover the decade for countryj). Aid to GDP in country(j) is the
average over the decade of the ratio of aid to GDP for that country. Technical aid to GDP is the respective measurefor the
technical component of aid to country (j). Initial income is measuredby the log of the initial level of per capitaPPP GDP,as
reportedin the Penn WorldTablesdatabase.SW tradepolicy index is the Sachs and Warnermeasure; see Sachs and Warner
(1995). In regressions (2) to (6), the dependentand independentvariables are winsorized at 5 percent and 95 percentof their
sample distributions.In the IV estimations, we use fitted aid to GDP, or fitted technical aid to GDP, as an instrumentfor the
respective variables, or its interactions,as in Rajan and Subramanian(2005a).
*** Significant at or below 1 percent.
** Significant at or below 5 percent.
* Significant at or below 10 percent.

at the firm level, whereas our measure is at the
industry level. Also, ours is a measure of con-
centrationof outside purchases,which does not
account for the fact that the fraction of outside
purchasesin total output can vary. Finally, our
measure is derived from US data which raises
questions about its applicability to developing
countries, where, for example, patternsof verti-
cal integrationcould be very differentfrom that
in the United States (though,clearly, that adap-
tation is part of the channel throughwhich poor
governanceinstitutionswill affect growth).
One way of validating our measure is the
following. If good governanceis necessary for
the growth of governance-dependentindus-
tries, then it must be the case that in countries
that have better governance,industriesthat are
more governance-dependentshould grow faster
relative to sectors that are less governance-
dependent.This is a hypothesisthatwe can test.
In Table 1, column 1, we regress the growth
rate of industry i in country j on a country-
specific measure of governance quality and
the industry-specific measure of governance-
dependence, after controlling for country and
industry fixed effects. The country measure
of governance quality is investment protec-
tion from the InternationalCountryRisk Guide

This content downloaded from 128.135.215.125 on Mon, 30 Nov 2015 16:34:29 UTC
All use subject to JSTOR Terms and Conditions
VOL.97 NO. 2 DOESAID AFFECTGOVERNANCE?
(ICRG),which is a good proxy for our notion of
governance quality as representingthe rule of
law and contract enforcement.The coefficient
on the interactiontermis positive and significant
at the 1 percentconfidencelevel, suggestingthat
in countries that have better governance,indus-
tries that are more governance-dependentgrow
faster.Specifically, the annualgrowth rate of an
industrythat is one standarddeviationhigherin
governance-dependencein a country that has
one standard deviation better governance is
0.8 percent higher, which is a sizeable magni-
tude, given that the averageannual growth rate
of industries is 2.1 percent in our sample. This
result survives various robustness checks (see
Web version, Table2).
It appears then that our governance-depen-
dence variable, postulated on theoretical con-
siderations by Blanchard and Kremer (1997),
picks out industriesthat thrive (relatively)in an
environment with good governance. A finding
in our core difference-in-difference estima-
tion, that aid affects the relative growth rate of
these industries,shouldthen give us more confi-
dence (thanin a simple cross-countryregression
between aid and institutions)that the channel
through which the effect is transmittedis gov-
ernance quality.
III. Endogeneityof Aid
Because we examine growth differentials
between industries within countries, the results
are less sensitive to the rationalefor why aid is
given. For example, even if aid is given only
to countries that display poor growth, interin-
dustry growth differentials should not be seri-
ously affected. Suppose, however, low growth
is primarily because countries have poor qual-
ity institutions, and aid is given systematically
to countries that have low growth. In this case,
we might be attributingto aid what is driven
directly by governance quality. The way to
address this is through instrumentation,which
allows us to disentanglethe direction of causal-
ity. We instrument for aid based on strategic,
historic, and cultural links between donor and
recipient(see Rajanand Subramanian2005a for
details of the instrumentationprocess).
One possible concern is that our instruments,
based on strategic and colonial links, could
be correlated with governance quality, thus
This content downloaded from 128.135.215.125 on Mon, 30 Nov 2015 16:34:29 UTC
All use subject to JSTOR Terms and Conditions
325
violating the exclusion restriction underlying
the instrumentation.In our sample, the direct
correlationbetween our instrumentfor aid and
investmentprotectionis low (0.07 in magnitude)
and not statistically significant, suggesting this
should not be a strongconcern.
IV. CoreResult
In column 2 of Table 1, we present the OLS
estimate for model 1. We present the IV esti-
mate in column 3. The interactioncoefficient is
negative and significant at standardlevels, sug-
gesting that in a countrythat receives more aid,
governance-dependentindustriesgrowrelatively
more slowly. Specifically,using the IV estimate,
the annual growth rate of an industry that is
one standarddeviation higher in governance-
dependence (HERF) in a country that receives
one standarddeviation more aid is 2.8 percent-
age points lower,which is a sizeable magnitude,
given that the average annual growth rate of
industriesis 2.1 in our sample. The IV estimate
remains significantwhen we clusterby country
or by industry.
One could argue that not all forms of aid are
equal. A number of economists try to identify
certain forms of aid that are likely to be "good"
or developmentoriented. Specifically, technical
assistance is geared toward building capacity
and institutions,and if the logic bears out, we
should expect to see a less negative interaction
with this form of aid. This is what we check in
the estimate reportedin column 4. It turns out
that the estimate is again negative and signifi-
cant. Eitherit could be that all forms of aid are
correlatedwith one anotherso that "good' aid is
hard to tell apartfrom other aid, or aid inflows
are fungible, so it does not matter whether the
use of certain inflows are carefully specified or
not, aid frees up resourceselsewhere.
In column5, we includeaninteractionbetween
initial per capita GDP and governance-depen-
dence to check whetheraid is just a proxyfor the
country's per capita income. The basic interac-
tion coefficient continuesto be significant,even
with this inclusion. In the core specification of
column 3, we can includean interactionbetween
governancequalityandgovernance-dependence.
If aid inflows "work" by reducing the qual-
ity of institutions, then introducing this direct
interactionshould weaken the aid governance-
326 AEA PAPERSAND PROCEEDINGS MAY2007

IAGO
.4

IGHA SYC
GRD OZMB
TGO 0 IDN
.2
* LSO ........ FA MRT. . . .
GDP
/ ILK, ~Aer~ CiG
BLZ
DCGV IUGA
BL
eL~w,~ CTFJI 1 GyP

0
SNGA
SNI ka-1,
SyDG
*JOR1 (R 1*Jt JOR
J ,'A N C~~~~CCCC~~~~CCCC~~
I
l
NGA
WE
' 1'BEN Bai
Manufacturing SUGA 0BLZ '\CfA YQuY
)FTJ LKA
Log
-.2
SCIV
IIDN
MRT WR TGO
LSO
COM
ZMB IGRD
SYC GHA

.4
AGO
-1 -0.5 0 0.5 1
Log Aid/GDP
coef = -0.15726536, (robust) se = 0.05571313, t = -2.82

FIGURE1. MANUFACTURING
ANDAID BETWEEN1980 AND 2000

Notes: This plot represents the conditional relationship between the change in the size of
the manufacturing sector and the change in aid in a country between 1980 and 2000. It
is based on running a panel regression where the dependent variable is log of the ratio of
the share of value added in manufacturingto GDP for a country in two time periods (the
late 1990s and the early 1980s). The explanatoryvariables are the country's per capita PPP
GDP, per capita PPP GDP squared, and fixed effects for the country and the time period.
All variables are averagesfor the period 1980-1985 and 1995-2000, respectively. To facili-
tate comparabilitywith the core results in the paper,the sample was chosen according to the
same criteria as in the core sample of the paper,namely to include countries that had an aid-
to-GDP ratio greaterthan 1 percent or are low-income countries. Data on manufacturingare
from the WorldBank's WorldDevelopmentIndicators.

dependence interaction.It does: the magnitude V. Conclusions

on the coefficient declines substantiallyand it is
also no longer statistically significant (see Web
version of this paperfor estimates).
Finally, when we include country indicators,
we cannot estimate the effects of individual
country-specificvariables.In column 6, we drop
the country indicators and include standard
explanatory variables in growth regressions.
The direct correlation of aid inflows (which is
instrumented)is negative, suggesting industry
growth rates are lower in countries that receive
more aid.' This is consistentwith the figurewith
which we startedthe paper.
5 More robustness checks are reported in the Web ver-
sion of this paper.
One of the ways aid might have affected
growth adverselyis by constrainingthe growth
of the manufacturingsector. A possible chan-
nel for which we provideevidence in this paper
is that aid might be particularlyassociatedwith
weak governance,possibly because aid inflows
reduce the need for governmentsto tax the gov-
erned or enlist theircooperation.The paperalso
offers a methodological approachto checking
whether governanceactually matters,based on
the growth of governance-dependentindustries.
Broadlyspeaking,the papersuggeststhateven
if the paucityof capitalis the missing ingredient
in the process of setting countries on the path
to prosperity (and the jury is still out on this),
the form in which the capital is received could

This content downloaded from 128.135.215.125 on Mon, 30 Nov 2015 16:34:29 UTC
All use subject to JSTOR Terms and Conditions
VOL.97 NO. 2 DOESAID AFFECTGOVERNANCE?
have adversespillovereffects thatlimit its value.
Indeed, if foreign aid reduces competitiveness
by inflating the exchange rate, or if foreign aid
reduces the efficiency of manufacturinginvest-
ment by adversely affecting governance, and
thus contracting, then aid inflows may reduce
the profitabilityof investmentand limit growth,
especially in export sectors that have tradition-
ally provedto be the engine of growth.
None of these effects is inevitable, though
they are not easily resolvedeither.An important
step for both aid givers and aid recipients is to
recognize thataid inflowsarenot an unmitigated
blessing, and that the worst thing that can hap-
pen with a greaterquantityof aid is not only that
it will be wasted. More aid, if ineffectively used,
can actually set back a country in its path to
development.This is why it is extremelyimpor-
tantto examine ways of improvingaid effective-
ness, recognizing that the state of knowledge
here is limited. Open-mindedness,allowing for
various approachesto aid delivery and utiliza-
tion, is essential here.
Finally, it is importantto recognize thatwhile
aid is essential in situations of humanitarian
need, it is probablylikely to play a much smaller
part in the process of economic growth. The
sooner countries recognize that aid is no pana-
cea, the less likely they are to postponedevelop-
ment indefinitely.
REFERENCES
Blanchard,Olivier, and Michael Kremer. 1997.
"Disorganization."QuarterlyJournal of Eco-
nomics, 112(4):1091-1126.
Brautigam,DeborahA., and StephenKnack.2004.
"ForeignAid, Institutions,and Governancein
This content downloaded from 128.135.215.125 on Mon, 30 Nov 2015 16:34:29 UTC
All use subject to JSTOR Terms and Conditions
327
Sub-SaharanAfrica."EconomicDevelopment
and CulturalChange,52(2):255-85.
Easterly,William. 2003. "CanForeignAid Buy
Growth?"Journal of Economic Perspectives,
17(3):23-48.
Easterly,William,Ross Levine,and David Rood-
man. 2004. "Aid,Policies, and Growth:Com-
ment." American Economic Review, 94(3):
774-80.
Hansen,Henrik,and Finn Tarp. 2001. "Aid and
GrowthRegressions."JournalofDevelopment
Economics,64(2): 547-70.
Johnson, Simon, Jonathan Ostry, and Arvind
Subramanian.2007. "Africa'sGrowth Pros-
pects:Benchmarkingthe Constraints."Interna-
tionalMonetaryFundWorkingPaper07/52.
Jones, Benjamin F., and Benjamin A. Olken.
2005. "The Anatomy of Stop-Start Growth."
NationalBureauof EconomicResearchWork-
ing Paper11528.
Levchenko,Andrei A. "InstitutionalQuality and
InternationalTrade."Unpublished.
Rajan, RaghuramG., and Luigi Zingales. 1998.
"FinancialDependence and Growth."Ameri-
can EconomicReview,88(3):559-86.
Rajan,RaghuramG., and ArvindSubramanian.
2005a. "AidandGrowth:WhatDoes the Cross-
Country Evidence Really Show?" National
Bureauof Economic ResearchWorkingPaper
11513.
Rajan,RaghuramG., and ArvindSubramanian.
2005b. "What Undermines Aid's Impact
on Growth?"National Bureau of Economic
ResearchWorkingPaper11657.
Roodman,David. 2004. "The Anarchyof Num-
bers: Aid, Development, and Cross-Country
Empirics." Center for Global Development
WorkingPaper32.