Document(s) 9 of 16

1 Introduction: an overview of globalization, reforms and macro-economic developments in India

¹ There is a large literature on this subject. Here it is sufficient to note that, while institutional factors might exacerbate the wage differential between the formal and the informal sectors, the original cause of the emergence and persistence of the differential are likely to be economic (see Mazumdar 1989).

2 Poverty, growth and inequality in the pre- and post-reform periods and the patterns of urbanization in India: an analysis for all-India and the major states

¹ We are thankful to Niranjan Sarangi for decomposition exercises done from NSS Unit Level Data.

² The quinquennial rounds are known as thick surveys constituting hundreds of thousands of households that allows analysis to be taken down to major state level as well.

³ We adopted the procedure followed by Sundaram and Tendulkar (2003a). We could reproduce the APCE figures by fractile group for URP for both rural and urban areas as presented in their revised calculations (Sundarm and Tendulkar 2003a). The changes in APCE by fractile group due to changes in the reporting period of five items from 30-day to 365-day (MRP) differs marginally. The details of the difference are given in Appendix 2. In this appendix we also discuss briefly the revised estimates of head count ratio (HCR) as obtained by Sen and Himanshu (2004).

⁴ This point was made by Professor Angus Deaton in correspondence with us.

⁵ Unlike the official price index, the Tornqvist price index is calculated using the information from the consumer expenditure surveys itself, and, hence, it allows for substitution behaviour as households adapt to relative price changes over time. It is also considered superior to Laspeyre's or Paasche's index as it satisfies both time-reversal and factor-reversal tests. These indexes, however, differ from official indexes in a number of ways. In particular, they rise somewhat more slowly over time than do the official price indexes, especially in the rural sector. This implies that poverty estimates will remain below that of the official estimates in subsequent periods.

⁶ Equation (Figure 2.5): y = 2.72 + 0.21*** x; R² = 0.386; Equation (Figure 2.6): y = 3.41
+ 0.06 x, R² = 0.015.                  (2.97)
(0.47)

Equation (Figure 2.7): y = 3.57 + 0.37*** x; R² = 0.859; Equation (Figure 2.8): y = 0.22
+ 0.29** x, R² = 0.253.              (9.26)

(2.17)

Figures in parentheses () are t-values and *** and ** denote significance at 1 and 5 percent levels respectively.

⁷ Datt and Ravallion (1992) had criticized an earlier formulation by Kakwani and Subbarao (1990) in which the change in poverty was decomposed into the growth and inequality components by taking the former as resulting from a change in mean income between the two years of comparison. The difference between this and the actual change in poverty was assumed to be that due to change in inequality. This procedure is equivalent to assuming that the Lorenz curve of the initial period is constant in calculating the growth component. Datt and Ravallion correctly point out that in this formulation the decomposition is not an exact one, but contains a residual term – which has been arbitrarily allocated to the redistribution component. It is shown that the residual vanishes if the Lorenz curve remains unchanged.

3 Trends in employment and earnings 1983–2000

¹ In some official statements (e.g., the Economic Survey 2004) employment is calculated on the basis of numbers estimated by the NSS on the basis of 'current daily status' (CDS). This gives a much sharper decline in the growth rate of employment. But this procedure is inappropriate because the CDS measures person-days not persons (see the Appendix 1 to this chapter).

² Gordon and Gupta (2003) also reached similar conclusions. They found that compared to the general trend in developing countries, India's service sector absorbed far less labor (relative to other similarly placed countries in Asia). In terms of services' share in GDP, India moved slightly above the general trend during the nineties. This was accompanied by a fall in the share of services in gross capital formation (Chandrasekhar and Ghosh 1999). Labor productivity increase without any increase in capital intensity in services might have occurred because of a faster rise in sub-sectors that depend on skilled labor (Gordon and Gupta 2003). Hansda (2001) found (from input – output tables) that the proportion of intermediate use of services output to be declining slightly between 1978–1979 and 1993–1994. There also exists a widespread view (Virmani 2002) that the increase in the salaries of public administration and defense contributed substantially in increasing contribution of services in GDP in India. In other words, it is largely spurious. But this interpretation has been challenged by Nagaraj (1999, 2000). He did not find that these components made any positive contribution to the acceleration in service-sector growth during the nineties.

4 Accounting for the decline in labor supply in the nineties

¹ The age-specific LFPR are explained in detail in Appendix 2.

² In West Bengal, there is clear evidence of increased absorption of labor in the 1980s as evident from workforce data of both NSS and Census. NSS also reports substantial contraction of employment in agriculture in the 1990s in this state. For Madhya Pradesh, under oilseed development program in the 1980s, there was huge expansion of oilseed cultivation. In the 1990s it faltered and downturn in oilseed was further compounded by allowance of import of cheap palm oil.

³ There might be an 'endogeneity problem' in this analysis in that a fall in PR itself reduces the household welfare level as measured. In this case, in the classification by household expenditure levels for the 55th round, there would be relatively more households in the lower welfare groups with less participation of females. But in this case we would expect to see a relative increase in the higher welfare groups for the participant households – not a fall as we in fact find.

5 Some implications of regional differences in labor-market outcomes in India

¹ Ahmad Ahasan and Carmen Pages are staff members of the World Bank and the Inter-American Development Bank respectively, both working at their headquarters in Washington DC. The findings, interpretations and conclusions of this chapter are those of the authors and should not be attributed to the World Bank, its Executive Directors, or the countries they represent. All errors are ours. The authors thank Sergiy Biletsky, Zhaoyang Hou and Mikhail B. Oslmolovski for excellent research assistance and the World Bank's Human Development team for providing the NSS region map coordinates, and members of the India Labor and Employment study team for their valuable comments in discussions held in September 2005.

² The latest thick round data is from the 2004–2005 survey.

³ Employment rates are defined as the share of workers (i.e. with status codes of 11 to 51 in Question 3, Block 5.1 in Schedule 10 questionnaire of the NSS) in the age group of 15 to 59. Participation rates are defined as the share of worker and population looking for jobs (i.e. with status codes from 11 to 81) in the 15 to 59 age cohort. Finally, unemployment rates are defined as the ratio of those unemployed but searching for jobs (Code 81) as a proportion of the population participating in the labor market, i.e. those working and unemployed but searching for jobs.

⁴ As the fourth section explains, most of these estimates have taken into account endogeneity of labor-market outcomes, wages, earnings and GSDP.

⁵ For instance, Duryea, Edwards and Ureta (2004) find that increases in education explain 30 percent of the increase in female participation rates in Latin America.

⁶ U.S. Bureau of the Census, Geographic Mobility: March 2000 to March 2001

⁷ Mohan and Dasgupta (2004).

⁸ It is true though that part of the answer may lie in the lack of updating of rural areas due to which many actually urban areas may be classified as a rural area – such as environs of Delhi. Thus higher "rural" wages may be partly reflecting this misclassification and the migration may be being under-reported. Still, it would be hard to ascribe all the changes to this fact.

⁹ We account for differences between urban and rural areas by including an urban dummy. In addition, we account for the fact that output only varies at the state and not at the regional level by computing robust standard errors clustered at the state level.

¹⁰ We estimate specifications of the following kind:

Where E_ijt denotes log of employment in state i, region j, round t, Y_jt is log of state GDP, w_ijt denotes log of wages and X_ijt is a vector of additional controls. We also control for aggregate, round effects by means of time (round) dummies, denoted as τ_t.

¹¹ As in the first differences estimates, we account for the fact that GDSP only varies at the state level by computing robust standard errors clustered at the state level. We also account for a likely endogeneity of GSDP and wages by instrumenting wages and GSDP with the following set of variables: share of industry in the state economy, log of state credit to industry per capita, log of state credit to agriculture per capita, log of power (measured as percentage of villages that are electrified) and log of km of roads per 100 square km per state. These variables capture differences in state development and output that are uncorrelated with contemporaneous labor supply.

¹² We estimate specifications parallel to specification (1) but substituting employment at the regional level by average regional weekly earnings. As usual, we compute robust standard errors clustered at the state level to account for the lack of within state variation of GSDP.

¹³ The difference between the predicted increase and the actual decline indicate that a part of the participation decline is not accounted for by substitution effects and that other causes, beyond lack of opportunities or increased husband incomes, are behind the participation decline. Understanding which additional factors are behind should be an important priority for future research.

6 Trends in the regional disparities in poverty incidence: an analysis based on NSS regions

¹ For 1972–1973, Jain and Tendulkar have obtained the poverty line a) by using a poverty line of Rs.15 per capital per month at 1960–1961 prices and b) by converting this poverty line at 1972–1973 prices by using the consumer price index for the agricultural laborers (CPIAL). For 1999–2000 we have used poverty line as obtained by Deaton (2002). He has taken poverty line of each state separately obtained from 'expert group on poverty line' by planning commission for the year 1987–1988. He updated it to the current (1999–2000) year from the higher prices that rural consumers pay in 1999–2000 for various items (obtained from unit level data) compared with 1987–1988.

² For the triennium ending 1982 and 1992 value of output and net sown area figures at district level (1961 classification) was obtained from Bhalla and Singh, 2001. From district level we aggregated to NSS region level and further to broad-region level. For the late nineties we managed to get district-level crop-wise season-wise data for two years 1997–1998 and 1998–1999 from Govt. of India, Ministry of Agriculture, District-wise Area and Production of Crops in India, 2001. These production data of different crops were converted to value by using crop wise price data (for the year 1993–1994) used by Bhalla and Singh (2001) to maintain comparability. These were aggregated to all crop NSS region level and further to broad region level. For two crops, rubber and coffee, we collected value of output and net-sown area from the Rubber Board of India and Coffee Board of India. For net-sown area we collected data from several issues of IFFCO, Agricultural and Fertilizer Statistics. The two years data were aggregated to get data for 1997–1999.

7 Agricultural productivity, off-farm employment and rural poverty: the problem of labor absorption in agriculture

¹ The non-specified activities could be irrigation, spraying of fertilizer and pesticides, post harvesting operations, etc.

² First we distinguished rural households dependent on agriculture in two categories cultivators and agricultural-labor households, on the basis of main source of earnings – labor income and self-employment in agriculture. Households whose main income comes from self-employment in agriculture (i.e. cultivators) are further divided on the basis of operational landholding size. These are divided into marginal (0–1 hectare), small (1–2 hectare), medium (2–4 hectares) and large (more that four hectares). Vakulbrahmanam's (2005) land classification is based on landownership whereas our land classification is based on operational landholding.

8 Employment elasticity in organized manufacturing in India

¹ See, e.g., Eichner (1973). The author notes the affinity of the theoretical tradition to the empirical literature which has found from surveys of business-pricing decisions that firms set prices on the basis of the 'cost plus' concept. The 'plus' margin is in its turn set by the necessity to generate the profit share which finances the investment ratio. The affinity of these ideas to the Kalecki models of the firm has been stressed by Asimakopoulos (1971) among others.

² See Mazumdar (2003b) for a more extended discussion of the labor-market theories relevant to this set of decisions.

³ We are grateful to Ahmad Ahsan of the World Bank for providing the ideas and calculations for this sub-section.

⁴ The data period of this section ends at 1994–1995.

⁵ Our data period for this section ends at 1994–1995.

9 Dualism in Indian manufacturing: causes and consequences

¹ The term "directory" and "non-directory" establishments presumably refer to the 'supposed' registration practice of the government that never materialized.

² Data on average earnings for the two countries can be found in the same national sources as are cited for Table 9.1.

³ It is apparent that the reason why the difference in value added per worker between the largest and the smallest size groups in Taiwan is larger than that in wages per worker is because Taiwan has a fair presence of large conglomerates, with a large share of capital, along with the small- and medium-scale firms. Such conglomerates play a smaller role in Hong Kong's manufacturing economy.

⁴ It should be emphasized once again that the sets of data considered here exclude the very large household and other parts of the informal sector in establishments employing less than five workers.

⁵ But under WTO agreement India was under obligation to remove quantitative restriction by 1 April, 2001. So by 2001–2002, all remaining 799 items reserved in the SSI list has been put under OGL. However, through a series of de-reservation over the years, the number of reserved items in small scale has been pruned to 239.

⁶ For more detailed examination of the origins and consequences of the Indian industrial policy affecting the small-scale sector see Mazumdar (1991) and Little et al. (1987)

⁷ For a poor country spent and still spends much more on tertiary education. In 2000 India spent 86 percent of per capita GDP per student on tertiary education, and only 14 percent on primary education. The corresponding percentages for China were 10.7 and 12.1 respectively. The Indian level of expenditure on tertiary education was even higher than Korea or Indonesia.

(IMF, WP 06/22 2006, pp. 6–7)

⁸ The contents and impact of this set of policies have been discussed in detail in Little et al. (1987). See also Mazumdar 2003b.

10 Growth of employment and earnings in the tertiary sector

¹ The APCE (Average Per Capita Consumption Expenditure) that we have used for the 55th round comes from employment schedule. It is an abridged questionnaire compared to the consumption-expenditure schedule. Consequently APCE in abridged employment schedule is underestimated. When APCE of both these questionnaires was compared across percentile APCE distribution, APCE from employment schedule was uniformly lower by 9–11 percentages compared to consumption-expenditure schedule. The distribution of APCE in both these series would be similar.

11 Legislation, enforcement and adjudication in Indian labor markets: origins, consequences and the way forward

¹ Ahmad Ahasan and Carmen Pages are staff members of the World Bank and the Inter-American Development Bank respectively, both working at their headquarters in Washington, DC. Tirthankar Roy is a Professor at the Gohkale School of Economics in Poona. The views expressed in this chapter are the personal views of the authors, and in no way represent the position of their respective organizations.
   The authors would like to thank Gautam Mehta for valuable research assistance, Farah Zahir for her indispensable help in gathering the data used in this study, and K.V. Ramaswamy for very helpful discussions. The authors also thank the Labor Bureau and the Central Statistical Organization for kindly sharing their data.

² There was a surge in the share of factories inspected in the year 2001. Given the existent data is impossible to determine whether this is a short-term reversion or a more permanent effect.

³ If a firm did not report a change in behavior, the reduction in the number of visits is zero. To compute the adjusted number of inspections the reported number of inspections is divided by (1 percent reduction in visit due to unofficial payments/100).

⁴ Notice, however, that a positive relation between the number of inspections and the perceived stringency of labor laws is not sufficient to conclude that inspections increase compliance. This is because inspections could concentrate in firms for which the law is more binding.

⁵ Conversations with Labour Commissioners in a number of states suggest that the grip of Chapter Vb is still very high. In the year 2004 there were six applications for permission to retrench in Karnataka, out of which only one was granted. In Tamil Nadu, there were three. One was granted, one was denied and the other one was still pending. In West Bengal, only one case was granted out of 20 applications.

⁶ Systematic data on the use of contract labor is only available from the Annual Survey of Industries (ASI) which only covers the manufacturing sector.

⁷ At the time of this study, the Investment Climate Survey data (ICS) provided firm level information for 25,582 firms from 55 emerging countries, but data for additional countries are periodically added to the data set. Surveys are administered in each country to a sample of firms stratified by size, sector and location, following a common framework. The surveys contain a number of questions regarding firms' characteristics, outcomes and perceptions about the business environment. The sample size for each country ranges from less than 100 establishments, for countries with a small private sector like Bhutan and Eritrea, to more than 1,500 for China, India, and Brazil. Information on ICS data can be obtained at http://rru.worldbank.org/external/cicic/portal.htm.

⁸ Firms' perceptions about labor-market legislation may be colored by the overall degree of optimism or pessimism about the economic outlook in each country. However, accounting for this factor by measuring each employer's response in relation to her average response to all other obstacles does not change the results. India still remains among the countries in which labor legislation is perceived to be more binding.

⁹ See Ahsan and Pagés (2006) for further details.

¹⁰ For example, in 1988, Tamil Nadu introduced an amendment increasing the power of conciliation officers in terms of enforcing attendance, compelling the production of documents and issuing commissions for the examination of witnesses. Other examples of laws that reduce the cost of disputes are (i) eliminating the need for the parties to refer labor disputes to the government prior to referring them to a Tribunal. or (ii) granting the government the power to transfer any industrial dispute pending before a tribunal to any other tribunal constituted by the state government for adjudication.

¹¹ They relate labor regulations to economic outcomes, by estimating the following specification:

where Y_it is an economic outcome such as manufacturing output, employment or wages in state i in period t, X_it is a vector of state controls, and τ_i and τ_t denote a state and time dummy, respectively. L _it-1 is a vector of legislation measures, which refer to the accumulated sum of amendments in a given type of laws over time in a state up to period t. In some specifications, outcome variables vary at the industry – state level. They lag regulatory variables one period to account for the average lag between enactment and implementation of the law and to reduce the possible endogeneity of labor laws.

¹² For instance, there is a large degree of uncertainty and ambiguity regarding whether business processing operations, calling centers or software companies are included in the definition of industry.

¹³ The term market infrastructure is adopted from Rajan and Zingales (2003).

¹⁴ It should be noted that a level of compensation of 45–60 days of compensation per year of service in case of a restructuring sick industries or profit-making companies, respectively, are above the norm in both developing and developed countries.

¹⁵ For more information on the operation of these schemes see Heckman and Pagés (2004), IADB (2004), Kugler (2004, 2005), Saavedra and Torero (2004) and Paes de Barros and Corseuil (2004).

¹⁶ See for example Brown and Medoff (1989).

¹⁷ An example is useful to clarify this issue. Assume a certain firm hires a contractor to run a cafeteria. Workers employed in the cafeteria are the employees of the contracted firm, which would then be responsible for abiding by all labor regulations with its employees. Nonetheless, the contracting firm could have a subsidiary responsibility if it hires a contractor that is not abiding by the law.

12 Strengthening employment and social security for unorganized-sector workers in India

¹ Philip O'Keefe and Robert Palacios are currently working at World Bank office in New Delhi. The views expressed in this chapter are the personal views of the authors, and in no way represent the position of their respective organizations.

We are grateful to Ihsan Ajwad, Corinne Siaens and Sangeeta Goyal for analysis of NCAER, NSS and ADB/MOF data respectively, and to Rinku Murgai and Martin Ravallion for discussions of their work on NREG. Thanks also to Dipak Mazumdar for comments on an earlier version.

² WDR 2006; Munshi and Rosenzweig (2005) for empirical evidence of credit and insurance-market failures driving limited mobility in rural areas in India.

³ For summaries of international experience, see Ravallion (1991); Devereux (2002); Subbarao (2003).

⁴ South Africa's Special Public Works Programme (SPWP) provides a good example of the latter, with participants required to undertake two days training for every 22 days worked, and 2 percent of each sub-project budget allocated to worker training (Devereux and Solomon 2005).

⁵ This did not happen in the lead-up to the 1999 national election, though spending may have been unusually constrained in the aftermath of the 5th Pay Commission.

⁶ This is to change in the 62nd round, which has revised questions on participation in works.

⁷ Based on estimates of around 65 million BPL households in 2000 (Indiastat.com).

⁸ Based on a usual status (PS+SS), workforce of just over 300 million. Sundaram, 2001.

⁹ This is consistent with international evidence of targeting of public works. For example, in Argentina's Trabajar program, 80 percent of beneficiaries were from the poorest quintile, while in Chile the share was close to 100 percent. In Bangladesh, around 70 percent of beneficiaries of the Food-for-Work Program were in the lowest income bracket, while in Indonesia, the post-crisis public works program was more likely to reach household who had suffered large shocks (Subbarao 2003).

¹⁰ CAG 1997. The CAG report on EAS from 2000 estimated that 10.9 million workdays were lost due to contractor margins, assuming a 10 percent margin.

¹¹ AFC SGRY program assessment for Orissa, 2005.

¹² Deshingkar and Johnson (2003). The statewide figure for AP for 2002–2003 in the MRD evaluation was 25.3 percent of works with contractor involvement.

¹³ In the last year of the period, the village coverage rate for EAS was 32 percent.

¹⁴ Acharya (2005) for seasonal MEGS employment data up to 2002–2003.

¹⁵ IDS for Planning Commission (2002) which found 80 percent of works occurring in this period.

¹⁶ CMD evaluation of SGRY on behalf of MRD, 2005.

¹⁷ The two shares are not necessarily inconsistent, but one would need to assume that female participants worked more than twice as many days.

¹⁸ General rate based on Sundaram revised estimates from 2001. Swamy (2003) on gender and workfare schemes internationally.

¹⁹ CMD study for MRD on SGRY (2005); PEO EAS evaluation (2000); and for Rajasthan an IDS evaluation of employment programs (2002).

²⁰ CMD for MRD 2005.

²¹ This is a rate of Rs.50 in 1999–2000 prices, scaled up for subsequent movements in the CPI for agricultural and rural laborers (RBI series).

²² Townsend (1994); Munshi and Rosenzweig (2005); Ravallion and Chauduri (1997). Gertler and Gruber (1997) re Indonesia.

²³ Peters et al. (2002); Dev et al. (2001); World Bank, op. cit.; and Duflo (2005).

²⁴ Though note the collinearity between society ageing and country income levels.

²⁵ Gertler (2001) re Korea, Taiwan, Singapore and Japan. O'Keefe and Palacios (2006).

²⁶ Dev et al. (2001); Jhabvala and Subrahmanya (2000) and World Bank, op. cit. This section focuses on contributory schemes. Dev et al. demonstrate the value of a wide-ranging consideration of social security.

²⁷ The central funds are: Mica Mines Welfare Fund (1946); Limestone and Dolomite Mines Labor Welfare Fund (1972); Beedi Workers' Welfare Fund (1976); Iron Ore, Manganese Ore and Chrome Ore Labor Welfare Fund (1978); Cine Workers' Welfare Fund (1981); Building and Other Construction Workers' Welfare Fund (1996). ILO, op. cit., and Rajan (2001).

²⁸ Among funds based on contributions, 12 had tripartite financing, ten were funded from employer and worker contributions, and two from employer contributions only. See ILO (2004).

²⁹ In contrast, for the Karnataka and AP Labor Welfare Funds, education has been the dominant expenditure – in Karnataka, accounting for around 98 percent of non-administration spending in 1997–2000, and in AP for 74 percent in 1998–2001. Rajan 2001.

³⁰ In addition to central initiatives, many states and specific national ministries have schemes of social insurance. They are not addressed here.

³¹ Berman and Ahuja 2005; Gupta and Trivedi (2005).

³² Devadasan et al. 2004, gives a typology; Ahuja and Narang (2005).

³³ Kuruvilla et al. (2005) re Yeshasvini, and ILO (2005).

³⁴ There are also revolving funds, often for drug costs, but these are not covered here.

³⁵ Another limitation is the regulatory obstacles for deposit taking. SEWA did not face this problem because it is also licensed as a banking institution.

³⁶ For a review of international experience with social pensions, Palacios and Sluchynsky (2006).

³⁷ Other factors such as savings capacity, risk preferences, and other demands on disposable income are also crucial (Palacios and Goyal forthcoming).

³⁸ Gertler and Hammer (1997), which finds price elasticities of demand for health services in developing countries of –0.5 to –1.0, again an average of around –0.2 from developed countries.

13 Epilogue

¹ The published report for the 61st round does not give expenditure distribution for the MRP at the state level. Therefore the authors

estimated the MRP poverty ratios using the Lorenz curve method based on URP distribution of persons and MRP per capita expenditures. This assumes that MRP expenditure levels are monotonic with URP levels for all observations. Since this is unlikely to hold strictly, it may marginally affect the poverty estimates.

(Ibid., p. 509)

Document(s) 9 of 16