Document(s) 5 de 16

Introduction

This chapter attempts to examine and understand the determinants of key labor-market indicators by looking at the experience across Indian states and regions by analyzing the NSS thick round data from 1983–2000.² It does so in two parts. First, it shows how states and regions display considerable variation in labor-market outcomes: some states and regions have been able to provide significantly more employment opportunities, and thus show higher employment and participation rates.³ Others states, not all in the same group, have lower unemployment rates.

Wage rates also vary significantly across regions leading to large differences in earnings. This leads to the second part of the chapter, which attempts to understand these patterns by trying to answer three key questions, two from the demand and the other from the supply side of labor markets. First, what are the factors that have enabled some states to create more jobs, and specifically what role did differences in economic activity and economic growth play in creating these jobs? Second, a related issue is, the role of differences in economic activity in affecting the quality of jobs as measured by earnings of workers. On the supply side the key issue that we address is the determinants of the variation in participation rates and especially in female participation rates across states and regions.

It may be useful to highlight the four main findings of this chapter. First, regional level differences in employment indicators are significant across regions and display a geographical clustering: 34 out of 78 NSS regions have statistically significantly different – better or worse – employment and unemployment outcomes than all India averages. Further, not only are these differences significant, but they have persisted over time. A related finding is that employment outcomes are clustered within certain states and regions: regions in the North Eastern states, Bihar and parts of UP, Jammu and Kashmir and Arunachal Pradesh, the coastal belts of Orissa and states such as Goa and Pondicherry have significantly worse employment outcomes. On the other hand Western and Southern States show better employment outcomes (see the third section below).

Second, wage and earning trends across regions present a complex picture with two countervailing aspects. On the one hand, significant differences in wages across regions are being balanced to some extent by evidence of convergence in wages across regions and between rural and urban areas. This finding is consistent with observations that not only are migration rates in India surprisingly low, but they have shown little signs of increasing in the 1990s. Further urbanization rates have also been much lower than predicted (Mohan and Dasgupta, 2004) and lower than in many other comparable countries. If wage rates are converging across regions and between rural and urban areas, then, other things remaining the same, the incentives to migrate to other states and to urban areas will decrease. On the other hand, dualism in wages between the formal, salaried sector and the informal, casual sector persists in that there remains a substantial premium for salaried workers (around 16 percent of the labor force) over casual labor (about 36 percent of the labor force) even after controlling for human-capital characteristics across all regions.

Third, we take up the issue of job content of growth by relating employment and earnings to GSDP levels across states as well as to changes in GSDP across four rounds from 1983 to 2000. Although GSDP growth is significant in explaining the growth of employment it cannot, per se, explain much of the variation in employment growth.⁴ However, we find that income differences across states and growth of income within states exert substantial positive effect on female employment levels while it also reduces unemployment rates for females in rural areas; but they are not significant in explaining employment-level differences for men. However, GSDP level changes across states do have a significant impact on raising rural earnings for males.

Fourth, regional variations in employment outcomes can be explained on the supply side, by differences in female employment and participation rates. Although female employment rates are uniformly lower than those of males, the variance in regional participation rates is also much higher for female workers. In addition, female participation rates have declined in the 1990s, a trend difficult to reconcile with the declining fertility and increasing education rates of female workers, factors that have contributed to rising female participation in other regions.⁵ In this chapter, we test two competing hypotheses to explain differences across regions and time in female participation: The first is that changes in participation are driven by income effects – increasing spouse earnings are driving female workers out of the labor force (see Sundaram and Tendulkar 2005b). The second hypothesis is that women are withdrawing from the labor market due to lack of opportunities (substitution effect). We find that while both forces are at work, the lack of opportunities, as indicated by unemployment rates and low expected earnings, have a greater role in explaining this trend.

The rest of this chapter is organized in the following manner: the second section briefly discusses the methodology used here and how the Chapter advances the literature. The third section documents the remarkable variance in labor-market outcomes across Indian states and regions. The following section analyzes these differences focusing on the role played by GSDP in affecting employment and earnings and the differences in participation rates. The final section concludes.

Methodology and how this chapter advances the literature

The evident disparities in economic conditions, growth and human development across Indian states have attracted considerable attention over the last few years. A sizeable literature has developed attempting to explain differences in growth and poverty-reduction performance across states. However, the cross-state and, particularly, the regional dimensions of employment remain relatively unexplored.

The literature on cross-state growth has highlighted the importance of differences in the investment climate in explaining differences in total factor productivity across states (Dollar et al. 2004); other studies have emphasized the differences in infrastructure and regulations (Lall and Mengistae 2005), the decline and variations in plan expenditures, the greater use of private capital flows and the wide variations in credit utilization (Dev 2002); the degree of urbanization (Sach et al. 2002); differences in land reforms, access to credit, education and labor-market related regulatory policies (Besley and Burgess 2004) as factors that have led to divergences in growth across Indian states. While most papers have confined their analysis to the state level, an important paper by Palmer-Jones and Sen (2004) has extended the analysis to the regional level, highlighting agro-ecological factors, irrigation and the interaction between these factors to explain divergence in agricultural growth rates. Overall the literature appears to have converged on a consensus that there is growing divergence in economic performance across states.

As noted, the literature on labor-market differences across states is relatively less developed. Some of the papers have presented at some length the differences in employment and unemployment rates across regions and have emphasized the role of divergent labor-market outcomes being a driving factor behind regional inequality (Bhattacharya and Sakkhivel 2004). A paper by Kijima and Lanjwou (2004) has estimated differences in agricultural wages across different regions. Another by Besley and Burgess (2004) has analyzed the effects of labor regulations in explaining state-wise variations in manufacturing output and employment growth. Hasan et al. (2003) estimated the effect of trade on labor-demand elasticity in industry and showed it to be positive. Extending further, they have shown how regulatory policies in states affect these demand elasticity adversely. At a broader level, the literature on labor-market differences has stopped short at two important points. First, the literature has focused more on describing differences in employment indicators across states and less on analyzing these differences, with Besley and Burgess (2004) and Hasan et al. (2004) being important exceptions. Second, the discussion has stayed focused at the state level except for Palmer-Jones and Sen (2004).

This chapter contributes to the literature on regional labor-market analysis in India in four ways. First, we construct synthetic panel data set on labor-market indicators at the regional and state levels to identify the extent of state and regional differences in Indian labor-market outcomes. The panel-based research on employment in India has so far mainly focused on using the annual survey of industries or smaller sample based data such as the NCAER surveys (e.g., Foster and Rozenweig 2004). By definition such research has excluded more than 80 percent of the Indian labor force or has been based on small nationally unrepresentative surveys. In this chapter we use the nationally representative panel to present the regional differences in employment, wages and participation, and analyze their determinants.

Second, this chapter takes the analysis beyond the state level to the (NSS-sample-based) regional level. Hitherto, analysis of regional level has been confined to a few studies on poverty and agricultural wage rate estimates. This is important since significant differences in employment indicators lie at the regional level. An examination of the data of the 55th round, as many as ten out of 32 states display a higher within state variation (as measured by the coefficient of variation) than the variation across states. Similarly, eight regions show higher within state variation in rural casual wages than the all-India variation.

Third, this chapter extends the discussion to the smaller states of the Northeast such as Arunachal Pradesh, Mizoram, Nagaland and Himachal Pradesh, where much of the variation in employment is found.

Fourth and last, this chapter uses a variety of panel data-estimation techniques using region-level data to examine the drivers of regional variations. These include estimating equations of varying degrees of complexity, fixed-effects models, and using instruments to account for endogeneity of employment, participation, wages and GSDP.

How different are labor-market conditions across states? Some stylized facts

Labor-market outcomes are significantly different across India in a number of respects. In what follows we focus on regional differences in employment, participation and earnings.

Stylized fact 1: striking regional clustering of employment

The first stylized fact is the clustering of low employment rates in the northeastern states of Arunachal Pradesh, Assam, Nagaland, Tripura and Manipur. In addition, the lagging states of UP and Bihar have particularly lower employment rates (Figure 5.1). This trend is largely mirrored in the low participation rates as shown in Figure 5.2. In general one finds a high correlation between employment rates and participation rates (0.95). This correlation is stronger for females (at 0.99 percent) than for men (0.95 percent). The relationship between these variables can be in both directions. Participation can lead to employment. Most workers seeking work in a poor developing-country labor market such as India can find work even if it is in a low-productivity job or a self-employed job. That is supply of labor creates its own demand. On the other direction, low employment rates in a region can lead to a discouragement of workers and lower participation. We examine this issue at some length later in the section 'Understanding differences across states and regions', where we try to analyze some key regional differences and trends in labor-market outcomes.

Participation rates for males and females are shown for all states in Figure 5.2. The relatively lower participation and employment rates in states such as Delhi, Kerala and West Bengal are puzzling (employment rates are presented in Figure 5.1). Given their higher income levels, the low employment rates in these states need more explanation. On the other hand, and more predictably, the prosperous states of the South and West, Tamil Nadu, Karnataka, Gujarat and Andhra Pradesh show significantly higher employment and participation rates. An interesting exception is Rajasthan where employment and participation rates are high.

Figure 5.1 Employment rates for males and female, 55th round (source: Authors estimates. Erm and Erf refer to employment rate of males and females respectively (measured in 0 to 1 scale; i.e. 0.8 refers to 80 percent employment rate). Employment rates for males and female are defined as workers in 15–59 age group as a share of the population in their age group. Derived from NSS 55th Round data).

Figure 5.2 Participation rates for males and female, 55th round (source: Authors estimates). Participation rates for men (prm) and women (prf) are defined as workers and unemployed in 15–59 age group as a share of the population in their age group (units are scaled to 0 to 1). Derived from NSS 55th Round data.

This could be one factor explaining why poverty rates in Rajasthan are low despite its relatively lower level of income.

Stylized fact 2: cross-regional differences in employment rates are much larger for women than for men

The second interesting stylized fact emerging from state and regional analysis is that the variation in female employment rates is significantly higher than for men: the coefficient of variation of female employment and participation rates is nearly four times as large as for men. However, the regional patterns in these variations are not as clear as the ones for males. In addition to low employment and participation rates in the North-eastern regions (including Tripura), and the low rates in UP and Bihar, female employment rates are also very low in West Bengal, and perhaps not so surprisingly in prosperous Punjab. But once again female employment and participation rates are much higher in the prosperous Southern and Western states of Gujarat, Maharashtra, Karnataka, AP and Tamil Nadu. One implication of this grouping is that it will be difficult to attribute low participation rates in West Bengal to schooling rates, as these are also high in the Western and Southern regions.

Stylized fact 3: regional differences in employment rates have persisted

The third stylized fact is that the divergence in employment rates across regions is persistent. We use the threshold of one standard deviation from the Indian mean to classify regions as being significantly different from all-India averages. For most variables, the number of regions significantly different from the mean have either stayed the same or increased in the 55th round (1999–2000) compared with the 50th round (1993–1994). Particularly noteworthy is the increasing divergence in rural employment rates in the 55th round compared with the 50th round. This is also confirmed when we see that employment and participation rates are highly correlated across consecutive rounds: employment and participation rates in the 55th round are closely correlated to those in the 50th round and so on. That is employment and participation rates tend to show very high persistence across regions, in sharp contrast to the case for real wages, which show a small (and negative) correlation across rounds (Table 5.1).

Another and perhaps better way to look for convergence would be to run unconditional convergence regressions (as done in the section 'Understanding differences across states and regions').

Stylized fact 4: there are some signs of convergence in wages for casual workers

As Table 5.1 has hinted, in contrast to employment indicators, there are signs of convergence in wages across regions at a time when wage growth has taken place in most regions. First, wage inequality is falling across regions for all categories of casual wages (see Table 5.2). There is a drop in the coefficient of variation and Gini coefficients in all casual wage categories – rural and urban non-agricultural between 1993 and 2000. However, there was an increase in regional inequality in salaried wages, although it was low to begin with.

Second, convergence in wages is also suggested by econometric tests of convergence that show wages in all categories to have unconditional convergence across regions between 1983 and 2000 (see Table 5.3). Growth rates of real wages are robustly negatively related to initial real wages in all categories. Significantly, the convergence is least for wages in agricultural operations. Given that agricultural productivity will vary widely depending on agro-ecological conditions, a slower degree of convergence is not unexpected.

Third, there is evidence that dualism between rural and urban areas has either mildly declined or, at least, it has not increased. The ratio of real casual wages in urban and rural (non-agricultural sector) shows a decline in all states except Nagaland, Manipur and Orissa, all states in the Eastern and North-eastern part of India. Similarly there are declines in the ratio of salaried to casual wage differentials between 1993 and 2000 within each region in both rural and urban areas. Most states show these trends, excluding West Bengal, UP, Meghalaya, Nagaland, Orissa and Tripura, where the ratio of salaried to casual wages have increased in the 55th round compared with the 50th.

However, the raw differential between rural and urban wages can be misleading as one needs to take into account human-capital characteristics in analyzing wage differentials. Urban–rural premium for both casual and salaried workers fell in the 1990s and largely disappears once human-capital characteristics are taken into account. In terms of regions, the number with significantly higher premium (say more than 20 percent) has fallen from 28 to five regions in the case of casual workers, and from 12 to six in the case of salaried workers. However, as far as the more difficult issue of salaried to casual workers is concerned, marked dualism still remains. Even after accounting for education, age and gender, there is no evidence of a narrowing premium, which remain high at about 30 percent.

To summarize, there is a large heterogeneity in employment and in earnings both across and within states. The dispersion in employment outcomes is higher for women than for men. And while there are few signs of employment convergence across regions, regional divergences in wages, as well as in urban–rural wage gaps, are declining. This leads to the next important theme in Indian labor markets: low migration and urbanization rates.

Stylized fact 5: economic migration between regions and urbanization rates are very low

Given the significant differences in labor-market conditions across the different regions, India's unusually low economic migration rates present somewhat of a puzzle. Overall while about 1.8 percent of India's population migrated on average each year between 1997 and 2000, only about 0.3 percent points of this were due to economic factors. Also a similarly small share, 0.3 percent points, migrated outside of their districts or states. In comparison some 5.5 percent of the US population migrated across county or state lines in a similar period.⁶

A look at the pattern of migration from and to different regions in Map 5.1 confirms that migration rates are low across many regions. In three years from 1998 to 2000, most regions show less than 1 percent net in- or out-migration. Chandigarh, Goa, Daman and Diu, Haryana, Punjab, Delhi, Mumbai and the Kolkata areas show the maximum inflow, exceeding 1 percent of the labor force in the three years from 1998 to 2000. Overall though Maharashtra and Gujarat show in-migration to be around 0.5 percent and 0.2 percent respectively, Northern Tamil Nadu, AP and parts of MP, and, less expectedly, Mizoram and Naga-land, also show in-migration. The main out-migration regions are Bihar, western Rajasthan and J&K. Kerala, Karnataka and Southern Tamil Nadu are also regions from where out-migration takes place.

Map 5.1 Economic migration across states and regions, 1997–2000 (source: Estimated from NSS data, 55th Round).

The convergent trend of wages across regions and growing unemployment rates in the major urban areas can help to explain why migration rates have not picked up. While wage differences are high, they are converging and do not appear to significantly affect migration, though urban casual wages – the best proxy of spot-market wages – are positively related with in-migration. On the other hand, unemployment rates are significantly inversely related to net economic migration rates.

Another issue related to labor markets is India's low urbanization rates. Even in the larger metropolitan areas of Mumbai, Delhi, Kolkata and Chennai that attract the highest rates of migrants, the in-migration rates, about 1.5 percent of the population per annum, are low. Further, the share of economic migration to urban areas has been stagnant from 66 percent in the mid-1990s to 62 percent. Compared with Asian countries such as China, Indonesia, Vietnam, Pakistan and Bangladesh, India has the lowest rate of urban population growth. China provides a dramatic contrast: urban population grew by some 190 million over 1990 to 2003. In India the corresponding number was 80 million or less than half.

Urbanization slowed down in India in the 1980s and 1990s as casual wages in rural and urban areas converged. Demographic projections in 1981 estimated that India's urban population would be about 31 percent in 2001. In reality, it turned out to be 27 percent of the population, i.e., lower by about 40 million persons.⁷ Part of the answer behind low urbanization rates would appear to lie in the converging trend in rural–urban wages. As the gap between rural and urban wages narrow and urban unemployment rates are growing, the expected earnings from migrating are falling. It follows that the incentives to migrate to the cities are declining accordingly.⁸ An important implication could be that urban infrastructure and service development may not be proceeding fast enough to create jobs that are better paying than rural areas. This may have further implications: because economic growth shows up in the growth of cities and towns, this slow urbanization has the potential to slow down growth.

Not only has urbanization slowed down, there is also evidence that job and population growth has shifted away from the large metropolitan cities and rural areas to mid-size towns. Decomposing urban growth by size of cities (Table 5.4) we see that there is significant shift of jobs from the rural centre and large cities to secondary towns and to a lesser degree in peri-metro areas. The implication of these developments has to be interpreted carefully. The growth of the large cities (100,000 or more) is not fast enough to accommodate the movement of labor and population out of rural areas to secondary cities with population between 20,000 to 50,000 persons. Given that these town sizes are probably too small to take advantage of economies of scale, there is a particular need to develop peri-metro areas.

Understanding differences across states and regions

In this section, we attempt to answer three key questions regarding differences in labor market performance across regions. We first study how economic growth has affected job creation and address the question of whether growth has been jobless and driven mainly by productivity growth. For this we estimate the impact of GSDP growth and GSDP levels on employment and unemployment across regions and across four time periods corresponding to the last four thick rounds (1983, 1987, 1993–1994, 1999–2000). Second, we use the state and regional variation to estimate the effect of GSDP and economic activity on earnings. Third, we analyze the determinants of regional differences in female participation rates to understand the variation in participation rates and its declining trend.

Explaining differences in employment performance across regions

In the broadest terms the relationship between GSDP growth and employment growth while significant in urban areas but is not by itself able to explain much of the variation across regions. We used estimates of the correlation between changes in GSDP and changes in employment across regions for rural and urban areas and for the total population (Table 5.5).⁹ To filter away changes in employment that can result from secular changes in schooling and marriage decisions by females, we take the labor force for persons age 25 and above. Finally, in some specifications we account for unobservable differences across states and rounds by including state and round fixed effects.

Our results presented in Table 5.5 indicate that growth of GSDP is significantly correlated to employment growth, but the effect is confined to urban areas. Overall 1 percent point increase in GSDP growth is associated with a 0.28 to 0.42 percentage point increase in employment growth rates. Two points are worth stressing. The employment effects of GSDP takes place mainly in urban areas. Second, however, growth, by itself, can explain very little in the variation in employment growth; only about 2 percent overall and about 9 percent in urban areas (columns 3 and 2 in Table 5.5). Our results indicate that employment growth fell significantly in the 1990s as the round dummies for 1999–2000 in columns 6, 8 and 9 have a negative and statistically significant sign.

Although these estimates suggest a strong correlation of employment with economic growth they can be misleading as these do not account for changes in wages or other factors. They do not account either for the endogeneity of economic growth and wages to changes in economic growth. We, therefore, make additional estimates of the relationship between output and employment by means of estimating labor-demand functions, for male and female workers, and for rural and urban areas.¹⁰ These relate employment levels in different states to output, wages and other factors after trying to account for endogeneity – i.e., by attempting to account for possibility that wages and output can be related to each other in both directions or be related through the impact of a third factor, e.g., investment.¹¹

Once again we find a strong relationship between output and employment across the difference regions and periods. In the Appendix, Table 5A.1 presents our estimates for males and females separately. We find that the elasticity of employment to output, i.e., the effect of a percentage increase in GSDP on male employment levels, across all states and periods and after accounting for wage changes, is estimated to be about 0.4 percent on average, 0.2 percent in rural areas and 0.8 percent in urban areas (columns 1–3 in Table 5A.1). Thus, across India, richer states employ more workers because GSDP is positively and statistically related to employment. As we saw previously in Table 5.5, employment effects are here also stronger in urban areas than in rural areas.

We then make the same estimates for males including state dummies (columns 4 to 6 in Table 5A.1). These variables absorb all the unobserved heterogeneity across states. This implies that the estimated relationship between output and employment would now main measure the effect of employment within states across and across time. Once we do this it turns out the relationship is much weaker. After including state dummies the elasticity of output-employment is positive and sizeable – though smaller than the one estimated without state dummies – but not statistically significant. These results imply that while there is a sizeable relationship between income and employment across regions, within states such relationship is less clear. Increases in state income are not necessarily related to an increase in employment in that state. While this may be evidence of jobless growth in recent periods, it may also reflect the low cyclical variation of male employment rates: i.e., most male workers have to find work of some kind.

Hence we next turn to see the effect of output change on females, who may have more flexibility. Making the same estimates for female employment (Table 5A.1, columns 7 through 12), we find that the elasticity of female employment with respect to GSDP levels is significant and higher (0.7 on average, 0.5 in rural areas and 0.8 in urban areas) than that for male employment. Thus, across states, there is an unambiguously higher impact of GSDP on female employment. This suggest that female employment responds more significantly to changes in the levels of GSDP across states partly accounting for the variations in participation we see across the regions. The estimates also suggest that within regions, increases in output are associated with larger increases in female than in male employment in the rural areas. Instead, we don't find much of a relationship between time changes in employment and output within urban areas. As in the case of male workers, despite a sizeable and robust regional correlation between income and employment across regions, there is no evidence that within states increases in output lead to increasing employment for females in urban areas. This latter result may be the result of a weakening relationship between income and employment in urban areas. It could also be driven by the fact that an expanding output may increase household income and allow women to buy more leisure (and therefore not increasing their labor participation and employment rates). We examine this point in more detail below.

It is also worth noting that as for males, female employment is substantially lower in urban areas than in rural ones and that the difference between urban and rural employment rates is much larger for women. Our results also suggest that the decline in employment registered in the latest round (1999–2000), and also shown in Table 5.5 is mostly due to a decline in the employment rates of women in rural areas (columns 7 and 10 of Table 5A.1).

Finally, our results also provide some estimates for wage-elasticities – i.e., how sensitive is employment to wages. Our results (in columns (1)–(3) of Table 5A.1) suggest that states with higher urban wages for males tend to register a lower demand for male urban salaried employment. There is, however, no evidence of such negative relationship between the price of labor and employment across states for male casual rural workers or for women in general.

Explaining differences in earnings

We next assess the impact of economic activity on real weekly earnings of males. Earnings are defined as the product of actual employment in a week and wages received.¹² The results, presented in Table 5A.2, show an interesting contrast to the previous result on employment. There is little evidence, when we do not take into account state-specific effects, that weekly earnings in rural or urban areas are higher in richer states (Table 5A.2 columns (1–4)). Once we control for overall state differences, we find that within states, earnings increase with output in rural but not in urban areas. This may reflect the much more significant presence of the formal and public sector (which provides for two-thirds of formal-sector jobs) in urban areas, which are less sensitive to cyclical changes in GSDP. It may also indicate that labor supply in rural areas is more elastic than in urban areas: an increase in economic activity in rural areas may require a higher increase in wages to pull people into the labor market. Combining these findings with those related to employment, we find that while an increase in economic activity increases employment and earnings for males in rural areas, not much change is registered in the urban areas.

We also find interesting results about the effects of caste and education on earnings. We find that regions with higher shares of scheduled tribe and caste people in population experience lower casual agriculture earnings and higher wages for salaried workers. Finally, we find the share of labor force with primary education to be positively correlated with higher earnings in rural areas, while the share of workers with post-primary education is positively correlated with earnings in urban areas.

Understanding regional variation and trends in participation rates

One key issue in determining employment outcomes is the variation in female participation rates across the different regions and time. As we have seen earlier in Figures 5.2 to 5.3, the main variation across regions takes place in female rates of participation. This is also evident in Map 5.2.

Map 5.2 shows that participation rates are particularly low in Bihar and UP, the Northern parts of Madhya Pradesh, parts of Punjab, and coastal Orissa and Goa. Interestingly, except for one region in Assam, participation rates are not particularly low in the North-east. On the other hand, parts of Tamil Nadu and Kerala show relatively lower participation rates. In general, as Table 5.7 shows, while participation rates for women are markedly lower than for men in both rural and urban areas in all regions, the variation in participation for females across regions is 15 to 20 times higher than for male.

In addition, not only are female participation rates significantly lower than for men but there has been a decline in the female participation rates in the 1990s.

Map 5.2 Participation rates for females, 1999–2000, NSS 55th round (source: Based on author's estimates from the NSS rounds).

Several authors (Vaidiyanathan 2001; Mazumdar 2005, Sundaram and Tendulkar 2005a) have suggested that much of the decline in participation is explained by the rise in school attendance by females. Further, marriage at the age of 15 to 24 may also account for women dropping out of the labor force. These points are generally confirmed in Table 5.6 which shows that while participation rates markedly declined in the 1990s for females in the prime age group (15 to 59 years), this decline was not significant for females in the 25 to 59 age group.

However, it is clear from the coefficients of variation presented in Table 5.7 that the regional variation among the 25–59 age group is also high and not very different from the variation for females. Understanding the regional variation among females in the 25–59 age group would help us better understand the determinants of female labor-force participation. We now turn to estimates of the determinants of participation of this group.

The key issue that we take up is what role do income and substitution effects play in explaining differences in participation rates for females? Income effect refers to the effect of rise in income of the household from increasing earnings of the spouse or due to other sources of household income due to which female workers can opt out of the labor force to do housework or enjoy leisure. Substitution effect refers to the greater incentives for females to work due to higher wages or better employment opportunities for women. Conversely, substitution effects will lead to lower female participation if opportunities for gainful work decline. If substitution effects are present, then the scope for bringing more women to the labor force increases by providing them with greater opportunities.

We approach this issue from two different sides. In the first approach, presented in Table 5A.3, we estimate the determinants of participation rates for females 25 years or older by using both female wages and spouses' wages to capture substitution and income effects. The unemployment rate is taken to measure the absence of opportunities in the labor market. Our results suggest that urban unemployment and overall high unemployment rates for females tend to discourage participation. Higher wages encourage participation in rural casual work for females, denoting the presence of substitution effects. Men's wages appear to have little impact indicating the weak role of income effects in this approach.

In our second approach to estimating income and substitution effects, we construct variables to represent expected earnings by female and male workers by multiplying wages by the probability of employment (or 1 minus the unemployment rate). Female expected earnings represent substitution effects and also capture opportunities available. Men's expected earnings capture income effects. The results shown in Table 5A.4 consistently indicate that higher expected female earnings in rural areas robustly increase female participation. The same effects are found in urban areas, but the coefficients are not statistically reliable. Still, the overall indication is that raising opportunities for female employment increases female participation across regions, particularly in rural areas. This association is also observed within regions across time (see first row of Table 5A.4 columns 4 and 6, and 7 and 9). Thus, in periods when women enjoy higher work opportunities (measured by higher expected earnings), female participation increases. Conversely, increase in male casual wages in rural areas and salaried wages in urban areas reduce female participation, a sign of income effects working. We estimate that substitution effects would have led to a 25 percent increase in female participation, while income effects would have reduced participation by 16 percent between the early and the late nineties. The net result, assuming no other effects were at play and that expected earnings by male and female would increase by the same proportion, would have been an increase in female participation of 9 percent.¹³

Summing up

In this chapter we have characterized labor-market outcomes across Indian states and regions over a period spanning the last four thick rounds, from 1983 to 2000. We have shown how regional differences in labor-market outcomes are striking in India, and have persisted over the last two decades. The exception is wages which show signs of converging across regions and across rural and urban areas. The latter fact combined with unemployment in states may help to explain why economic migration rates and urbanization rates are unusually low in India. Some interesting implications can be drawn.

Foremost among these is economic growth and activity levels have been important in causing good labor-market outcomes, though in a somewhat nuanced way. When regional differences are taken into account, growth has not been jobless. In the short run though, growth has a muted effect on employment. Increasing labor productivity, which has led to growth, is associated with lower employment growth as an immediate effect. But in the medium term, increasing productivity does not adversely affect employment growth. Over the longer term, however, the relationship with growth and employment is clearer. States with higher levels of GSDP are also states which have created more urban employment and rural earnings in the case of males. Given that male-unemployment rates are negligible in rural areas this result is understandable. Significantly, the effect of differences in GSDP levels is more striking for female employment, which tend to vary much more than male employment. Higher GSDP levels lead to higher female employment in rural and urban areas.

Our analysis also suggests that increasing employment opportunities for females will also help to arrest the decline in female-participation rates. Although there is some evidence of income effects that lead females to drop out of the labor force, economic opportunities are the strongest factor affecting female participation.

The analysis in this chapter also highlights the importance of urbanization and domestic migration. The narrowing of the wage gap between rural and urban areas in each region and higher unemployment rates has lowered urbanization rates. Seen from the opposite direction, impediments to urbanization lower the growth of employment and higher wages. At present slow urban development also slows down manufacturing growth – with about half of new manufacturing jobs being created in rural areas. A complementary approach would also be to facilitate economic migration both to regions that are more dynamic and also to urban areas. Policies that can mitigate obstacles to domestic migration, through better safety nets and insurance for migrants, will also improve labor-market outcomes by allowing workers to work in areas where there are more opportunities and higher return.

Given that poor employment outcomes are persistently clustered in Northern, North-eastern and some coastal regions a regional focus on growth and employment is called for. Investment in infrastructure - power, road, and irrigation and credit facilities, which are found to affect GSDP positively, can lead to higher employment prospects. Related to this is the need to improve investment climate in these regions - a key aspect of which are labor-market-related regulatory reforms.

Appendix

Table 5A.1 Instrumental variable estimates of the effect of GSDP on employment levels for male and female workers

Table 5A.2 Estimates of the effect of GSDP on earnings for male workers

Table 5A.3 Estimates of determinants of female participation rates: female and male wages, household earnings and unemployment rates

Table 5A.4 Determinants of female participation rates: expected earnings of male and females

Document(s) 5 de 16