You are on page 1of 27

Labor Force Participation Rates, Wages and their Determinants in India 1

Ashish Narain

(Preliminary Draft: do not cite)

July 16, 2006


Background Paper for the India Labor and Employment Study

1
I thank Ahmad Ahsan, Carmen Pages, Dr. Sundaram and Dr. Tendulkar for their useful comments and
guidance. I am also grateful to Prakarsh Singh and Sergiy Biletsky for their excellent research assistance..
I. Introduction
Indian policy makers have stated as their goal, the provision of “gainful and high quality
employment to at least the additions to the labor force.” 2 In a country where the working age
population grew by 2.5 percent per annum between 1993 and 2000, and is expected to continue
growing by more than 1.5% till 2025, this constitutes a major challenge. The challenge would
have become even greater in the 1990s, when evidence suggests that employment growth slowed
down, but for the fact that labor force growth also slowed down to 1.3 percent according to
Planning Commission estimates. This was attributable to low and declining labor force
participation rates (Table 1), especially among rural women.

Table 1: Labor Force Participation Rates for Men and Women (UPSS)
Year Rural Urban
Male Female Male Female
1983 54.8 34.2 53.6 15.5
1987-88 54.9 33.1 53.4 16.2
1993-94 56.1 33.0 54.2 16.4
1999-2000 54.0 30.2 54.2 14.7
Source: Planning Commission (2001)

The big question then is what was driving participation rates down? If participation rates
were declining because individuals valued leisure or home activities more, or because they chose
to invest more in acquiring human capital, the slowing down of labor supply and, in part,
employment growth may be rationalized as the consequence of rapid development. A more
pessimistic alternate, however, is that some sections of the working age population may have
chosen not to seek work because they became discouraged by lack of employment opportunities
in the market. A stressed employment situation in the 1990s may then be the reason for slower
labor force growth. A better understanding of these dynamics is necessary to gauge whether the
slowing down of employment in the late 1990s was a supply-led or demand-driven story.
Looking forward, such an analysis is also useful in understanding whether the declining trend is
likely to continue, and for making projections. This is of immense relevance not only to Indian
policy makers in devising appropriate interventions, but also to the world as India accounts for
almost 16 percent of the world’s labor force.

2
Planning Commission of India, Tenth five-Year Plan
The importance of In the Indian context, another important element of the story has been
the widespread and robust growth in wages. Real casual wages, the closest measure of returns to
labor in the unorganized sector, grew at a faster pace in the 1990s than in the 1980s both in the
urban and rural areas (Mazumdar and Sarkar, 2004). A similar picture also seems to be true of the
regular wage/salaried workers category, where more than two-thirds of occupational groups
witnessed higher real earnings growth in the 1990s relative to the 1980s (Sundaram and
Tendulkar, 2005). There was also a reduction in the male-female average earnings differentials
(Sundaram and Tendulkar, 2005). Assuming that wages are determined largely competitively,
these observations imply higher productivity growth and are indicative of either increased labor
demand, reduced supply or a combination of both. In this paper, we also try and examine the
determinants of wage setting and how these have evolved over time. Thus for example, have
wages risen because the human capital of the average worker has increased, or because the
premium that this can command in the market has risen? Has wage growth been rapid in all
regions, after controlling for worker and industry characteristics, or has there been a regional bias
to wage growth? Since wages are observed only for that section of the population who are in the
labor force, but are neither unemployed nor self-employed, the picture which comes out is
necessarily blinkered. However to the extent that observed wages represent the opportunity cost
of labor, understanding their determinants is of some merit.

This paper utilizes individual level data from four quinquennial NSS “thick” rounds to answer
some of these questions. These rounds are the 38th (1983), 43rd (1987-88), 50th (1993-94) and the
55th round (1999-2000). The paper adds to current literature in many different ways. Firstly by
subjecting individual decisions to rigorous econometric analysis, we provide evidence of the
effect of individual and household characteristics and socio-economic factors on participation.
Secondly, we examine the effects of educational attainment on participation. We present evidence
for the existence of a U-shaped relationship between education and participation, but show that
there are strong gender differences in this relationship. Thirdly and most importantly, the paper
tries to estimate whether tough labor market conditions discourage people from participating in
labor markets, something which to the best of the authors’ knowledge has not been looked at in
the Indian context.

Not surprisingly, we find that the single most important factor determining the probability that an
individual participates in the labor force is their sex. Being female reduces the probability of
participation by 65 percent, holding all else equal. The regression results suggest that household
responsibilities, caste and religion all have a strong impact on women’s participation decision.
Another interesting finding of the paper is that the effect of education on participation varies with
gender. We show that for men, an inverse relationship between educational attainment and
participation exists only in the case of the young (up to 29 years). For women, however, there is a
U-shaped relationship which holds across rural and urban areas, and surprisingly for all age
groups though the “depth” varies by rural or urban location and age. We also find evidence that
labor market conditions have an effect on labor supply. Holding all else constant, higher
unemployment deters working age individuals from participating. The effect is strongest in the
case of rural women. For them, a one percent increase in regional unemployment reduces their
probability of participating by 2.8 percent. In the case of men, local unemployment conditions
have a smaller impact on their participation decisions. The strongest effect in on youth below the
age of 25, and is almost negligible for men in the age group 25-59.

The rest of the paper is organized as follows. Section II discusses the data and methodology used
in the paper. Section III presents the results, while section IV discusses their policy relevance and
conclusion

2. Labor Supply in India in the 1980s and 1990s

Between 1983 and 1993-94, the total labor force in India grew at close to 1.8 percent per annum
in terms of all the usual definitions. This was slightly lower than the growth rate of population. In
the subsequent period 1993-94 to 1999-2000, however, trends vary depending on how one
defines the labor force. In terms of Principal Status, labor force growth accelerated slightly. 3
This was especially true for rural females and urban males. However when subsidiary workers are
included as part of the labor force, the trend reverses. There was a substantial slowdown in
growth by this definition, from 2.07 percent in the first period to 1.63 percent in the second
(Table 2). The largest drops were for females in general, and for rural females in particular
(Sundaram and Tendulkar, 2005a). In a similar vein, Mazumdar and Sarkar (2004) point out that
participation rate declines over the 1990s were more muted for all groups on the principal status,
but larger on principal plus subsidiary status. They estimate that while between 1983 and 1993,
subsidiary workers accounted for less than 10 percent of the total decline in participation, in the

3
On Principal Status, a person is counted as being part of the labor force is he/she was either working or
reported seeking/being available for work for the major part of the last 365 days. A Subsidiary worker is
one who was engaged in economic activity on a more or less regular basis, but was classified as
unemployed on the Principal status.
second period, this increased to 30 percent for males, and 85 percent for females. They also
highlight an interesting aspect by splitting the sample by age. They estimate that 90 percent of
withdrawals from the labor market in the first period were accounted for by workers less than 24
years of age. In the subsequent period however, withdrawals came significantly from the
working age group 25-59. These observations suggest that different dynamics were at play in the
two periods. They also highlight the role of two groups in explaining the trends in the mid 1990s:
subsidiary workers, especially the young and female subsidiary workers, and urban females.

Table 2: Labor force Growth Rates (percent per annum)


1983-1993/94 1993/94-1999/2000
UPS UPSS UPS UPSS
Rural Males 1.96 1.96 1.59 1.49
Rural Females 1.14 1.49 2.16 0.77
Rural Persons 1.72 1.79 1.75 1.24
-Of which 15-29 age group 1.45 1.54 1.06 0.41
Urban Males 2.96 2.93 3.2 3.15
Urban Females 3.43 3.61 2.79 1.66
Urban Persons 3.04 3.07 3.13 2.85
-Of which 15-29 age group 2.21 2.21 2.23 1.89
Males 2.23 2.21 2.05 1.96
-Of which 15-29 age group 1.81 1.80 1.48 1.28
Females 1.48 1.78 2.27 0.91
-Of which 15-29 age group 1.12 1.47 0.96 -0.41
Source : compiled from Sundaram and Tendulkar, 2005a

What could explain declining female participation? Female participation rates in India are already
low compared to the rest of the world (World Bank, 2004). In part, this may reflect a preferred
division of labor within the family, with women responsible for within household activities, and
men for outside work. 4 This would imply that demographic factors like changes in the child-
dependency ratios may have an important role to play. Poverty may also be an issue. The poor

4
Social and cultural norms in India restrict especially the high status households from allowing women to
work outside the home. Keeping women at home is often considered a sign of honor (Das and Desai, 2003).
Since household activities do not classify as economic activities, this would depress female participation
rates.
often have no alternative but to work to support their families (Das and Desai 2003). Rising
income levels could then be expected to lower female participation rates (Sundaram and
Tendulkar 2005a, Mazumdar and Sarkar, 2004). Education is also often posited as another
important factor. The argument is that the low educational levels of women impairs their ability
to get the kind of employment they desire, and deters them from entering the labor force. If this
were the case, since higher educational attainments imply higher potential wages, participation
should go up with education. However empirical studies in the region have often found a negative
relationship between women’s education and labor force participation (Kingdon and Unni (1997);
Sathar and Desai (2000), Dubey, Pala and Thomas (2004)). 5 Another factor could be that female
workers find it hard to get the kind of jobs they need. In India social norms often restrict the range
of jobs considered acceptable for them. This restricts their flexibility in the market place, and may
make them particularly vulnerable to tough employment conditions. Women could arguably be
more prone to becoming discouraged, and keeping away from the labor force.

Another interesting aspect has been the decline in participation by the young. As noted above, the
declines have been larger on the subsidiary status, and especially for rural females. Interestingly
this group has also seen significant increases in the student-population ratio, leading many to
believe that their declining LFPRs are the result of increasing education levels among the
population. 6 (see Figure 1) Education can depress participation through increased retention in the
education system. However there may also be a less positive explanation behind this curve. In the
Indian context, the preferred occupation of the educated is a non-manual job in the organized
sector (Planning Commission 2001). This is true even for workers who have received only middle
level education. When these aspirations cannot be met, some young workers may become
discouraged and withdraw from the labor force. The current low participation rates, especially
among the youth, could possibly in part be due to this phenomenon (Ghose 1999).

5
It is possible though that this negative relationship is a reflection of the close association between
education and family income: Women from richer families tend to study more, while their need to
contribute to household income is lower, reducing their incentive to seek work.
6
. Researchers have often posited an elongated U-shaped relation where participation rates first fall as the
level of education attained by a population group rises, but then rise for people with more than 12 years of
education (see Ghose 1999, Chadha and Sahu (2002)).
Figure 1: Scholl Enrolment at Different Levels

School Enrolment at Different Levels

1200

1000
Enrolment inlakhs

800

600

400

200

0
1980-81 1990-91 1999-00

Primary level Secondary level

II. Data and Methodology

This paper utilizes individual level data from four quinquennial NSS “thick” rounds: the 38th
(1983), 43rd (1987-88), 50th (1993-94) and the 55th round (1999-2000). NSS data is chosen as it is
widely believed to be the most reliable and comprehensive source of labor related information in
India, and has been the basis of most survey-based labor work. In line with the interests of this
paper, we restrict the sample size to individuals in the prime working age group 15-59.

In analyzing labor force participation decisions, the chosen dependent variable is a dummy
variable which takes the value 1 if a person is in the labor force, and 0 if he does not participate.
Two different definitions are used to define participation: Usual Principal Status, and
participation on Subsidiary Status. 7 Unlike previous work, we do not add the two together under
the assumption that there may be systematic differences in the drivers of participation on this
basis. Women, for example, are more likely to take up part-time or seasonal employment or take
up work during times of crisis to supplement family incomes (Das and Desai 2003). In
accordance with the NSS, under the usual status definition, a person is defined as being in the
labor force if his principal status is that he is either working or seeking work. People who are

7
Under this definition, persons who work on the family farm or in business are counted as being in the
labor force even if they are unpaid. However since attending to domestic duties is not considered an
economic activity, persons engaged in unpaid household work (who mostly tend to be women), are not
counted as working. Thus there is an underreporting of women in the labor force (Beneria (1981))
attending educational institutions, or who attend to domestic duties are thus not considered to be
part of the labor force. The NSS definition is also used to define subsidiary workers.

The right hand side variables we control for are a host of individual level characteristics (age, sex,
marital status, religion and caste), dummies for educational attainment (up to primary, educated
up to secondary level, educated till graduate level or higher), location (rural/urban), the number of
dependents (children, adults above working age), indicators of labor market conditions and of real
per capital expenditure in the family. We also control for regional fixed effects. Definitions for
the control variables are provided in Appendix 1.

The proxy we use for the labor market conditions is the regional unemployment in the rural or
urban area where the individual under consideration resides. Given the large number of
observations which make up one cell even at this level of disaggregation, we are reasonably
assured that this does not create any endogenity problems. A bigger theoretical issue however is
the extent to which open unemployment captures the difficulty workers’ face or expect to face in
getting job. It has often been pointed out that open unemployment rates in developing countries
may understate the actual problem since very few individuals, especially among the poor can
afford to be unemployed (World Bank, 2004). Not surprisingly, it is the youth in the age group
15-29, who can still depend on family support, who account for over 85 percent of the
unemployed on the usual status (Sundaram and Tendulkar, 2005a). Ideally underemployment
would be a better measure of labor market conditions in the Indian context. Unfortunately this is
notoriously hard to measure, and we do not attempt it here. Instead we argue that while
unemployment may be understated in level terms, the variation in this level between different
regions may still be a good proxy for variations in labor market conditions between them.

To check the robustness of our results, we run the same regressions using another proxy for labor
market conditions: the proportion of casual workers in the region and location under
consideration who report having worked more of less regularly the past year. A number of
studies show that casual workers are at the bottom of the income scale, and are among the most
vulnerable segment of both the rural and the urban labor market (Glinskaya and Jalan, 2005).
Employment in this section is also believed to be largely demand determined (Sundaram and
Tendulkar, 2005b). Any tightening of labor markets would therefore be most directly reflected in
the ease with which casual workers can find work. There is the danger though that if different
segments of the labor market are not well integrated with each other, this may not reflect overall
labor market conditions very well.

There is a possibility that both of our proxies differ systematically between rural and urban areas.
We correct for this by running regressions separately for urban and rural areas, and for males and
females separately.

We estimate a probit model, which predicts the probability of participating in the labor force by
relating it to the control variables above. The results are presented as increases in the probability
of participation in response to a small change in explanatory variable. A discussion of these, and
of our main results follows in the section below.

Another issue of concern is the issue of wage determination, and what explains rising wages in
India. To examine these, we estimate a wage equation. However since we only observe wages for
only the wage employed section of the working age population, this creates a sample selection
bias. 8 To correct for this we use the maximum likelihood procedure suggested by Heckman
(1979). This involves estimating both the participation and wage equation equations in a
simultaneous equation framework. While the estimation of the participation equation is explained
above, in the wage equation the log of wages is regressed against control variables like worker
characteristics (age, sex, marital status, religion and caste), educational attainment, location,
whether a person is employed on regular or casual basis, dummies for industry, occupational
category and regional dummies. Wages are estimated as those accruing from the main activity in
the previous week, adjusted for prices and the number of working days reported. All observations
are appropriately weighted using sample weights provided by the NSS.

III. Results

In trying to understand the drivers of declining participation rates in India, it is important to first
examine the changes that occurred in the work age population. 9 Looking at the NSS sample
suggests that the most major change has been the increasing levels of educational attainment
among the population, most dramatically at the secondary school level, but also at the graduate

8
A sample selection bias may arise because unobservable factors determining whether an individual is in
wage employment may be correlated with unobservables determining wages.
9
Working age population here is defined to include all individuals in the prime working age 15-59.
level (see appendix 2). The effect was particularly significant in rural areas. Consider the case of
rural women: In the 38th round (1983-84), only 7 percent were educated till the secondary level,
while the percentage of graduates among them was only 0.3 percent. However in the 55th round
sample (1999-2000), the same ratios were 19 percent and 1 percent. Per capita expenditures have
also risen, both in rural and urban areas. The other major change has been the growing
unemployment. In terms of demographics, there was a slight increase in the average age of the
sample, reflecting a slowing down of the growth of the working age population. There was also a
reduction in the average number of children per household, and in the rural areas a reduction in
the percentage of married individuals mostly between the 38th and 50th rounds.

What impact did these changes have on the participation rate? To understand this we estimate the
Principal status participation rates, the results of which are presented in Appendix 2. We run
regressions for four categories: males and females, in urban and rural areas. To estimate the
effects on various age-groups, the same model is then also run for males and females, but cut by
different age groups in the prime working age 15-59.

Not surprisingly, the single most important factor determining whether a person participates in
the labor force is their sex. Being female reduces the probability of participation by 65 percent
compared to males. Women’s participation is higher for castes considered to be low in the social
hierarchy (SC/ST) in both rural and urban areas, and across all age groups. This is true even after
controlling for income (in terms of the per capita expenditure), and is consistent with research
which shows the prevalence of social norms, especially among the higher castes, of keeping
women within the home. Muslim women are also less likely to participate, especially in rural
areas.

There is a sharp gender divergence in the effect of some household characteristics on


participation probability. For a woman, being married, or belonging to a family with more
children below the age of 5, reduces the chances that she enters the labor force. Interestingly, the
negative effect of marriage on participation is larger in urban areas than in rural areas. The effect
is most significant for women in the age-group 30-45, which is the main child bearing and rearing
age. On the other hand, for men, the same factors raise the probability that they participate. This
clearly points towards a division of labor where women are primarily responsible for domestic
household responsibilities, and the primary responsibility of men is working and earning outside
the home.
How does educational attainment affect participation? We approach this question by looking at
the effect this has on the likelihood that an individual participates in the labor force. For men, the
relationship is downward sloping, that is, higher educational attainments are correlated with lower
participation probability. However cutting the sample by age shows that this relationship holds
only for the young (ages 15-29). The reduced likelihood of participation most likely reflects the
retention of this cohort in the educational system.

The effect of education on women is however harder to understand. One, the relationship is a
well-defined U shape. While women with up to secondary education are 10-18 percent less likely
to participate than women with no education, graduates are 3-10 percent more likely to be in the
labor force than their illiterate counterparts depending on urban or rural location. Two, unlike
men, this U-shape holds for all age groups. It is not clear what explains this difference in the
response of men and women. One possibility is that it is mostly affluent, upper caste families
which can afford to educate women beyond the secondary school level. However as noted before,
these are also the families whose social norms restrict women’s entry into the labor force, and
where the need for women to work is the least. Another possible explanation could be that only
the returns from graduate education are large enough to induce women to join the workforce. This
is however an area which needs more analysis.

0.1

0.05

0
Prim ary Secondary Graduate

-0.05

-0.1

-0.15

-0.2

Urban Men Rural Men Urban women Rural women


Another important factor which could have an impact on participation decisions is income. The
need for individuals in higher income households to participate in the labor force is lesser, and
especially for women one may expect that this would depress their participation rates. We use
household per capital expenditures as a proxy for household income. Since this is expected to be
endogenous to the participation decision, we instrument for this variable. 10 For all the four
groups, the effect of per capita expenditures on participation is found to be weakly positive,
though there is some evidence of a negative relationship for individuals in the top quintile of the
distribution.

A third interesting line of enquiry relates to the extent to which labor market conditions impact
individual’s decision to enter the labor force. Using unemployment as a proxy for the difficulty
individuals expect to face in getting jobs, we relate participation probability to regional
unemployment in the urban or rural location of the respondent. Our results show that labor market
conditions do have an effect on participation, with the effect being generally larger for women
than for male workers. The largest effect is in the case of rural women (4 percent decline in
probability for every one percent increase in unemployment rate) and the smallest for rural men
(0.2 percent decrease). The effect is also larger for the young in both sexes. For the older age
groups, unemployment does not have a significant effect on male participation, but it continues to
have an effect for women in all age groups. This suggests the vulnerability of women and the
young to tight labor market conditions.

Looking together at the results of the participation regressions, and the changes in the labor force
allows us to understand the importance of different drivers of participation rates. In the case of
rural women, the model suggests a large decline in participation on the principal status in the 80s,
arising primarily out of a worsening of the unemployment situation. Rising educational levels and
income levels in the top quintile also had a role to play, though the effect was much smaller.
Between the 50th and 55th rounds, participation declined further, though by a smaller percentage,
larger on account of a further worsening of the rural unemployment rate. For urban women the
model predicts an increase in participation probability between the 38th and 50th rounds largely on
account of rising per capita expenditures, which was moderated by rising unemployment. By the
later period though, both these effects become more muted. For rural men, the main factors in

10
The instruments used are the per capita household land ownership, log of household size, the ratio of
children below age 5 and elderly above age 60 to the number of adults in the working age, and for caste
(SC/ST).
order of importance responsible for declining participation rates in the 80s indicated are
increasing education levels, and increasing rural unemployment respectively. For urban males,
rising urban unemployment and increasing levels of educational attainment were both equally
responsible for a decline in the participation rate, though this was balanced by a gradual increase
in the average age of the population which tended to raise participation rates. All these effects had
largely played out in the later period 1993-2000.

Turning now to look at subsidiary workers, the category responsible for most of the decline in
labor force participation in the mid-90s, a very interesting result stands out. While most other
factors affect subsidiary workers in a similar manner as labor force participants on the principal
status (coefficients have the same sign), when it comes to the effects of per capita incomes, there
is a strong negative income effect. This is true for both rural and urban, males and females. The
income effect by far dominates all other factors in explaining changes in the probability of
participating on the subsidiary status. It is larger for females than for male workers, and the
effects were more pronounced in the 80s and early 90s than in the later period. Our results
suggest that subsidiary employment is strongly need driven employment, driven by the need to
provide supplemental income when per capita expenditure levels are low.

Thus while rising unemployment, but also increasing education levels among the population
appear to be important in explaining declining participation on the principal status, it is the rising
income levels which explain the largest part of the declines on the subsidiary status.

Given that a large part of the decline in labor supply seems to be coming from the worsening
unemployment position, an important question is to what extent wages have been adjusting to
take care of the slack? What explains robust wage growth along with a worsening unemployment
situation? The other questions relate to the extent to which labor markets have been to deal with
the wage differentials which exist between the two sexes, different locations and by employment
status (regular vs. casual status). To answer these questions, we estimate a wage equation,
correcting for the sample selection bias by using Heckman’s technique.

One axis of discrimination has been the wage differential which exists between male and female
workers even after controlling for other worker characteristics. This gap declined sharply in the
80s and early 90s, but thereafter showed a slight increase in the 1999-2000 survey (Figure 2). The
reasons for this result are not immediately clear. It may be that we are inadequately controlling
for occupation or casual/regular wage status. Indeed as Sundaram and Tendulkar (2005b) point
out, in terms of both casual wages, growth was higher for female workers than for males workers
in both rural and urban areas, though the picture is more complex in the case of regular workers.

Figure 2: Male-Female Wage Differential

0.7

0.6

0.5

0.4
38 43 50 55

Another very interesting result is that after controlling for worker characteristics, industry and
occupation, there has been a secular decline in the urban-rural wage gap (Figure 3). This result is
robust, and holds even when the regression is run separately for men and for women. The finding
is counter-intuitive in that it indicates the increasing integration of rural and urban factor markets.
The question however is how is this happening? One possibility is that this is happening through
increased labor migration to higher wage areas. However, the magnitude of migration is too small
and declining (Ahsan and Pages, 2005) to explain all the change. Others believe that this has
occurred through a movement of capital away from high wage urban areas and into lower wage
rural, peri-urban areas (Foster and. Rosenzweig, 2004). The mechanism through which this is
occurring remains a matter for further enquiry.
Figure3: Urban-Rural Wage Differentials

0.4

0.3
log wage

0.2

0.1

0
38 43 50 55

Male Female All

Understanding differentials between different employment status is important because as argued


by Sundaram and Tendulkar (2005b), regular wage status can provide a good approximation to
organized employment, whereas casual wages provide some indication of the status of workers in
the informal sector. The trends in these are presented in figure 4. These show that other factors
remaining the same, workers in regular wages jobs earn significantly higher wages than casual
workers, and that the differential between the two has been increasing. Further, most of this trend
seems to be driven by trends among male workers.

Figure 4: Wages Differentials between Regular and Casual Workers

0.35

0.3
logwages

0.25

0.2

0.15
38 43 50 55

All Male Female


References

Aggarwal, Suresh C., 2004. Labor Quality in Indian Manufacturing. Economic and Political
Weekly of India, December 11, 2004.

Biradar, R. R., 2004. Human Capital Base of Labor Force among Social Groups in India:
Emerging Issues and Challenges. Indian Journal of Labor Economics, Volume 47 (4), October-
December 2004

Chadha G. K. and P. P. Sahu, 2002. Post-Reform Setbacks in Rural Employmen: Issues That
Need Further Scrutiny. Economic and Political Weekly of India, Mat 25, 2002.

Ghose, Ajit K., 1999. Current Issues of Employment Policy in India. Economic and Political
Weekly of India, September 4, 1999.

Mathur, Ashok, 2002. Skill Acquisition and the Indian Labor Force Employment and Labor
Market Reforms in India. Paper presented in Consultative Workshop on Employment and Labor
Market Reforms in India, mimeo. Institute of Human Development, New Delhi

Planning Commission of India.(2001). Report of the Task Force on Employment Opportunities.


July 2001

Planning Commission of India, 2002. Report of the Special Group on Targeting Ten Million
Employment Opportunities per Year over the 10th Plan Period. May, 2002

Singh, C. S. K., 2003. Skill, Education and Employment: A Dissenting Essay. Economic and
Political Weekly of India, August, 2002.

Sundaram, K., 2001. Employment and Poverty in 1990s: Further Results from NSS 55th Round
Employment-Unemployment Survey, 1999-2000. Economic and Political Weekly of India, August
11, 2001.

Unni, Jeemol and Uma Rani, 2004. Technical Change and Workforce in India: Skill Biased
Growth? Indian Journal of Labor Economics, Volume 47 (4), October-December 2004
Appendix 1

Independent Variables and their Description

Variable Coding
Age, age square In years, as per survey

Sex 0 = male, 1 = female

Marital status Married =1 if currently married


=0 otherwise

Muslim Muslim = 1 if religion is muslim


= 0 otherwise

Other religion Other religion = 1 if not hindu or muslim


= 0 otherwise

Scheduled Caste SC = 1 if caste is Scheduled caste


=0 if scheduled tribe or general

Scheduled Tribe ST = 1 if caste is Scheduled tribe


=0 if scheduled caste or general

Education Primary = 1 if literate through primary

Secondary =1 if literate through higher


secondary

Graduate = 1 if graduate in various disciplies

Reference: Uneducated
Region North = 1 if Himachal Pradesh, Punjab,
Haryana, Rajasthan, Chandigarh, Delhi
(Reference)

East =1 if West Bengal, Orissa, Andaman and


Nicobar

West =1 if Gujarat, Maharashtra, Goa, Dadar


and Nagar Haveli, Daman and Diu

South=1 if Tamil Nadu, Karnataka, Kerala,


Andhra Pradesh, Lakshadweep and
Pondicherry

Central = 1 if Bihar, Uttar Pradesh, Madhya


Pradesh

North-east =1 if Manipur, Tripura, Mizoram,


Nagaland, Assam, Meghalaya, Sikkim,
Arunachal Pradesh

Household size Medium = 1 if size is 4-6 members


Large =1 if size>6
Reference small, size <4

Extended family Dummy =1 if respondent’s relationship to head


of family is married child, unmarried child,
grandchild, father, mother, father-in-law,
mother-in-law, brother, sister, brother-in-law,
sister-in-law, other relative

Dummy=0 if self, spouse, servant, employee or


other non-relative
Sector Urban =0, rural =1

Unemployment rate Calculated as number of unemployed divided


by number of people in labor force (weighted
by sample weights) in the NSS region and
sector of the respondent

Proportion of Self-employed Calculated as number of self-employed divided


by number of employed people (weighted by
sample weights) in the NSS region and sector
of the respondent
Appendix 2
Regression Results

Model 2
(All persons
with alternate
measure of
Model 1 Labor market Model 3 Model 4
Variables (All persons) conditions) (Only Males) (only Females)
Age 0.08 0.08 0.03 0.04
Sex -0.68 -0.68
Age square 0.00 0.00 0.00 0.00
marital status # (1 =married) 0.01 0.02 0.06 -0.17
Dummy for muslim # -0.07 -0.08 0.00 -0.11
Dummy for other religion # -0.04 -0.04 -0.02 -0.02
SC # 0.04 0.04 0.01 0.05
ST # 0.18 0.18 0.00* 0.25
Literate till primary level # -0.07 -0.07 -0.01 -0.08
Literate till secondary level # -0.17 -0.18 -0.06 -0.14
Literate till graduate level or above # 0.01 0.00* -0.06 0.08
Medium sized family # -0.10 -0.10 -0.01 -0.07
Large sized family # -0.13 -0.14 0.00* -0.10
Extended family # 0.0* 0.01 -0.01 -0.03
Dummy for sector # (urban =0, rural =1) 0.0* 0.13 -0.01 0.02
Number of children 0.01 0.01 0.00 -0.01
Unemployment rate -0.81 -0.12 -0.89
prop. Of self employed ----------- -0.587421 ------------ -----------

observed P 0.5746185 0.574619 0.875276 0.286844


predicted P 0.6397585 0.639465 0.958053 0.26106

Figures show change in probability of being in the labor force for a small change in the independent variable, except for
# probability change for a discrete change of the dummy variable from 0 to 1
* indicates not significant at the 5% level
Model 5 Model 6 Model 7 Model 8
(Urban males) (Rural Males) (Urban Females) (Rural Females)
Variables
Age 0.03 0.03 0.04 0.04
Age square 0.00 0.00 0.00 0.00
marital status # (1 =married) 0.09 0.05 -0.23 -0.13
Dummy for muslim # 0.01 0.00* -0.05 -0.14
Dummy for other religion # -0.01 -0.02 0.02 -0.04
SC # 0.01 0.01 0.06 0.04
ST # -0.01 0.00* 0.16 0.26
Literate till primary level # 0.01 -0.02 -0.06 -0.08
Literate till secondary level # -0.05 -0.07 -0.09 -0.15
Literate till graduate level or above # -0.05 -0.08 0.08 0.05
Medium sized family # -0.01 -0.01 -0.06 -0.07
Large sized family # 0.00* -0.01 -0.08 -0.11
Extended family # -0.03 0.00 -0.02 -0.05
Number of children 0.01 0.00 -0.01 -0.01
Unemployment rate -0.11 -0.14 -0.18 -1.54

observed P 0.860629 0.88359 0.188397 0.341236


predicted P 0.951165 0.961655 0.166424 0.320875
Figures show change in probability of being in the labor force for a small change in the independent variable, except for
# probability change for a discrete change of the dummy variable from 0 to 1
* indicates insignificant at the 5% level
Model 9 Model 10
(For Males with (For Females
interaction with interaction
term) term)
Variables
Age 0.03 0.04
Age square 0.00 0.00
marital status* (1 =married) 0.07 -0.17
Dummy for muslim # 0.00 -0.12
Dummy for other religion # -0.02 -0.04
SC # 0.01 0.05
ST # 0.00* 0.24
Literacy till primary level # 0.00* 0.00*
Literate till secondary level # -0.04 -0.12
Literate till graduate level or above* -0.06 -0.01*
East* 0.00 -0.02
West* 0.01 -0.05
south* 0.00 0.07
northe~t* 0.00* 0.05
central* 0.00* 0.09
Medium sized family # -0.01 -0.07
Large sized family # 0.00 -0.11
Extended family* -0.01 -0.03
Dummy for sector* (urban =0, rural =1) 0.00* 0.11
Number of children 0.00 -0.01
Interaction between unemployment rate
primary literacy -0.066 -0.48
Interaction between unemployment rate and
Secondary literacyl -0.10 -0.14
Interaction between unemployment rate and
literacy graduate level or above -0.02* 0.27
observed P 0.875276 0.286844
predicted P 0.957656 0.262949

Figures show change in probability of being in the labor force for a small change in the
independent variable, except for
# probability change for a discrete change of the dummy variable from 0 to 1
* indicates insignificant at the 5% level
Model 11 Model 12 Model 13
(Males in the (Males in the (Males in the
age group age group age group 46-
15-24) 25-45) 59)
Variables
Age 0.26 0.00 0.05
Age square 0.00 0.00 0.00
marital status* (1 =married) 0.16 0.04 0.06
Dummy for muslim # 0.06 -0.01 -0.02
Dummy for other religion # -0.05 -0.01 -0.02
SC # 0.06 0.00* 0.00*
ST # 0.02 0.00 0.00*
Literacy till primary level # -0.19 0.00 0.01*
Literate till secondary level # -0.48 0.01 0.01
Literate till graduate level or above # -0.60 0.00 0.02
Medium sized family # -0.05 0.00 0.01
Large sized family # -0.05 0.01 0.03
Extended family* -0.15 -0.01 -0.07
Dummy for sector* (urban =0, rural =1) -0.08 0.00* 0.05
Number of children 0.02 0.00* -0.01
Unemployment rate -1.08 -0.02 0.12

observed P 0.591456 0.979209 0.939012


predicted P 0.659497 0.984402 0.954519

Figures show change in probability of being in the labor force for a small change in the
independent variable, except for
# probability change for a discrete change of the dummy variable from 0 to 1
* indicates insignificant at the 5% level
Model 14 Model 15 Model 16
(Females in (Females in (Females in
the age the age the age group
group 15-24) group 25-45) 46-59)
Variables
Age 0.12 0.03 -0.05
Age square 0.00 0.00 0.00*
marital status* (1 =married) -0.11 -0.31 -0.16
Dummy for muslim # -0.05 -0.13 -0.12
Dummy for other religion # -0.03 -0.02 -0.02
SC # 0.00* 0.06 0.08
ST # 0.20 0.27 0.24
Literacy till primary level # -0.07 -0.08 -0.11
Literate till secondary level # -0.16 -0.12 -0.10
Literate till graduate level or above* -0.05 0.11 0.24
Medium sized family # -0.01* -0.05 -0.07
Large sized family # -0.04 -0.08 -0.10
Extended family # -0.01* -0.04 -0.14
Dummy for sector* (urban =0, rural =1) # -0.02 0.03 0.07
Number of children -0.01 -0.01 -0.01*
Unemployment rate -0.91 -0.96 -0.50

observed P 0.201287 0.31968 0.308116


predicted P 0.169576 0.298305 0.287993

Figures show change in probability of being in the labor force for a small change in the
independent variable, except for
# probability change for a discrete change of the dummy variable from 0 to 1
* indicates insignificant at the 5% level

You might also like