Rural unemployment : a survey of concepts and estimates for India

Rural Unemployment - A Survey of SWP- 234 Concepts and Estimates for India World Bank Staff Working Paper No. 234 April 1976 This paper is preparqd for staff use. The views expressed are those of the author and not necessarily those of the World Bank.   This paper is prepared for staff use. The views are those of the author and not necessarily those of the Bank. THE WORLD BANK Bank Staff Working Paper No. 234 April 1976 RURAL UNEMPLOYMENT - A SURVEY OF CONCEPTS AND ESTIMATES FOR INDIA This paper presents a conceptual framework for measuring rural unemployment and underemployment in developing countries. A disaggregated model of the labor market is outlined. Also, five distinct and significant rural unemployment rates are calculated for India and their implications examined. Estimates of surplus labor for India are reviewed. Prepared by: Raj Krishna (Consultant) Office of the Director Development Economics Department Development Policy Staff   RURAL UNEMPLOYMENT - A SURVEY OF CONCEPTS AND ESTIMATES FOR INDIA Table of Contents Page No. SUMMARY i-v 1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . 2. QUANTITATIVE CONCEPTS FOR THE MEASUREMENT OF UNEMPLOYMENT RATES ................... 3. CRITERIA FOR DEFINING THE STATE OF UNEMPLOYMENT . . . . . . . 12 4. UNEMPLOYMENT IN 27TH AND 25TH ROUNDS OF N.S.S. AND EARLIER ESTIMATES ... . . . . . . . . . . . . . . . . 22 5. THE CORRELATES OF UNEMPLOYMENT . . . . . . . . . . . . . . . . 38 6. SURPLUS LABOR ................. 55 7. TOWARD A MODEL OF THE LABOR MARKET . . . . . . . . . . . . . . 63 REFERENCES ......................... . 66 LIST OF TABLES Summary Table 1 - Five Rural Unemployment Rates and Magnitudes Selected Years, India . . . . . . . . . . . . . ii Table 1: Illustrative Measurement of the Flow Unemployment Rate . 5 2: Assumed Cut-off Limits for Counting Unemployed/ Underemployed Persons. 7 3: Illustrative Measurement of the Stock Unemployment Underemployment Rates. 8 4: Summary of All Illustrative Flow and Stock Rates Computed in Tables 1 and 3 . . . . . . . . . . . . . . 11 5: Unemployment Rates by Daily, Weekly and Annual Status Rural India . . . . . . . . . . . . . . . 23 6: Unemployment Rates of Small Farmers and Landless Workers, 25th Round (1970-71 . . . . . . . . . . . . . 27  LIST OF TABLES (cont.) Page No. Table 7: Weekly Status Rural Unemployment and Underemployment Rates, NSS Rounds 14th to 19th . . . . . . . . . . . . . 28 8: Usual Status Unemployment Rates, Rural India, Census 1961 and 1971 . . . . . . . . . . . . . . . . . . 29 9: Number of Usual Status Unemployed Persons, Rural India . . 31 10: Number of Unemployed Persons Estimated From NSS Data, Rural India ... . . . . . . . . . . . . . . . . 33 11: End-of-the Week Unemployment, 27th Round 1972-73 . . . . . 34 12: Numbers of Unemployed, Severely Underemployed, and Moderately Underemployed Persons, Rural India . . . . . 35 13: Unemployment in Person-Days per Week 27th Round (1972-73). 37 14: Proportion of Persons Willing to Move Outside the Village for Full-time Employment, 25th Round (1970-71) . 49 15: Proportion of Persons in the Age-Group 15-44 in Willing Migrants ... . . . . . . . . . . . . . . . . . 50 16: Proportion of Illiterates in Willing Migrants . . . . . . 50 17: Expected Annual Wage Interval with the Highest Frequency of Willing Migrants . . . . . . . . . . . . . 51 18. Expected Income Range and Proportion of Migrants . . . . . 52 19: Annual Wage Expected and Proportions of Potential Migrants in Small Farms and Farm Labor Households NSS 25th Round (1970-71) . . . . . . . . . . . . . . . . 54 20: Surplus Labor Estimates, India . . . . . . . . . . . . . . 56  RURAL UNEMPLOYMENT - A SURVEY OF CONCEPTS AND ESTIMATES FOR INDIA SUMMARY l. A large mass of data on rural unemployment has been collected in India by the Census authorities in 1961 and 1971 and the National Sample Survey (NSS) organization in 8 Rounds during 1958-59/1972-73. Much of this data still remains unanalyzed. Expert committees who have analyzed some of it have abstained from presenting usable estimates of rural unemployment and discussing trends because of controversies about, and frequent changes in, the concepts used by the Census/NSS agencies. But it is shown in this paper that, with a careful examination of various concepts, and a few critical distinctions, five distinct and significant rural unemployment rates and magnitudes can be identified in the data. All the available estimates of these five rates and magnitudes are brought together in the paper. The estimates of each type of unemployment for different years are broadly comparable. They reveal striking trends in the decade of the sixties and early seventies. 2. The crucial distinctions required for the interpretation of Indian data are those between (a) time-rates and person-rates, (b) current status and usual status rates and (c) rates excluding and including underemployment. Other distinctions discussed in the paper are those between micro and macro rates, and rates based on the voluntary and the normatively determined supply of labor. (a) The first distinction relates to the fact that the unemployment rate may be measured either in units of labor time per period, i.e., unemployed person-days per week as a proportion of total voluntary labor supply in the same units, or in numbers of unemployed persons at a given point of time as a proportion of the number of persons in the-labor force. They may be called time rates and person rates, respectively. (b) The second distinction refers to the fact that some unemployment rates have been measured on the basis of employment status over a short reference period, usually a week preceding the Census/Survey, others have been measured on the basis of employment status over a long reference period, usually a year preceding the Census/Survey. They are called current status rates and usual status rates, respectively. (c) Third, some estimates of unemployment include open unemployment only, while others include open unemployment as well as underem- ployment. Underemployment may either (i) include persons available for more work, or (ii) be a normative estimate computed with a pre-determined time-criterion (42 hours/week) only.  - ii - 3. Given these distinctions we have the following five roughly comparable rates in the Indian data: (1) Current Status Time Rate (CST). NSS data. 2 years 1970-71, 1972-73. (2) Current Status Person Rate - Unemployment Only (CSPU). NSS data. 7 years between 1958-59/1972-73. (3) Current Status Person Rate - Unemployment plus Available Under- employment (CSPU + UUA). NSS data. 5 years between 1958-59/ 1964-65. (4) Current Status Person Rate - Unemployment plus Normative Under- employment (CSPU + UUN). NSS data. 5 years between 1958-59/ 1964-65. (5) Usual Status Person Rate (USP). Census data 2 years 1961 and 1971; and NSS data one year 1972-73. 4. The ranges of these rates, all recomputed as percentages of the relevant estimates of the labor force are summarized in Table 1. SUMMARY TABLE 1 FIVE RURAL UNEMPLOYMENT RATES AND MAGNITUDES, SELECTED YEARS, INDIA Rates (Percent of Labor Force) Rate 1958-59 1960-61 1964-65 1966-67 1970-71 1972-73 1. CST - - - - 6.4 - 7.6/a 6.8 2. CSPU 5.6 - - 2.7 - 3.4 3. CSPU + UUA 16.1 - 9.2 - - 4. CSPU + UUN 47.0 - 31.8 - - 5. USP - 0.4 - - 1.1 1.0 Magnitudes (Millions) Magnitude 1958-59 1960-61 1964-65 1966-67 1970-71 1972-73 1. CST - - - - 5.6 17.4 2. CSPU 8.1 - - 4.5 - 10.0 3. CSPU + UUA 23.4 - 14.0 - - 4. CSPU + UUN 68.4 - 48.1 -- - 5. USP - 0.6 - - 2.1 2.0 la A sample of the bottom decile of farmer households, and a sample of landless farm households only.  - iii - 5. Striking trends are indicated by these magnitudes. The CST indicates that in an average week in 1972-73, 17.4 million personweeks of labor-time was idle and available for work (in a rural labor force of about 239 million persons). About one-third of this unemployment was concentrated in landless households and the smallest ten percent of farmer households. 6. The current status open unemployment was going down from 8 to 4.5 million persons during 1958-59/1966-67 and has risen back to 10 million in 1972-73. 7. The usual status 'chronic' unemployment has also increased from about half a million in 1961 to about 2 million in 1971. 8. Thus the roughly comparable rates available both for the 60s and early 70s show that unemployment was declining in the 60s but increased rapidly in the early 70s presumably due to a series of bad harvest and low rates of growth. 9. The figures of underemployment, computable only for the 60s, are truly staggering--averaging 19 million on the basis of reported availability for more work and 58 million on the basis of a cut-off point of 42 hours of work per week. 10. The data on distributions of the unemployed by various charac teristics--age, sex, education, region, expected wages, current income, tenurial status, work and location preferences, willingness to migrate and job-search behavior--are also summarized in the paper. Some facts deserve special notice. (a) About 10 percent of regular farm and non-farm workers and 56 percent of casual workers are willing to migrate for full-time wage employment. (b) An overwhelming majority of these willing migrants are illiterate prime-age workers. (c) They expect an annual wage in the range Rs. 800-1200. The average expected wage, weighted by the proportion of willing migrants in each wage interval is about 1.6 times the wage actually earned. (d) The relationship between the expected wage and the proportion of willing migrants in each expected wage group suggests three backward bending supply curves at three different wage levels, presumably depending on educational and other characteristics. (e) The relationship of unemployment levels and income is expectedly negative. And, of course, the incidence of unemployment is higher in landless than in small farm households.  - iv - (f) Rural unemployment is more evenly spread across age-groups than urban unemployment. And there are much fewer first- job-seekers among the rural unemployed than among the urban unemployed. (g) The incidence of current status unemployment and under- employment is higher and the incidence of usual status or chronic unemployment is lower among females than among males. (h) The unemployment rate hits its seasonal peak in the summer quarter (April-June) and its rough in the monsoon quarter (July-September). 11. Surplus labor estimates for India are also reviewed in the paper, along with the limitations of the methods used to generate the estimates. For the whole agricultural sector Mehra's estimate (27% of the work force in 1961) is the only one available. If at least one worker per holding is allowed the rate drops to 6.4 percent which is comparable with the minimum (CSPU + UUA) rate for the 60s (6.5 percent). 12. The paper also outlines a 5-sector model of labor allocation and wage determination which theoretically distinguishes five types of unemploy- ment according to the dominant causes: (a) income-leisure preference; (b) scarcity of non-labor factors, land and capital; (c) technological and product-mix choices; (d) institutional wage rigidity; and (e) job-search behavior. 13. A quantitative decomposition of measured unemployment according to these categories may not be possible in the immediate future because of data limitations but the model suggests that they must be separated. At low income levels the need for income keeps the labor supply at a high level and as income rises leisure preference begins to reduce the effective supply. Thus minimum income-needs and leisure preferences affect unemployment through their influence on the labor supply. The capacity to finance job search also influences the part of labor supply which remains voluntarily unemployed. On the demand side, shortage of land and capital relative to the labor supply, and choices of inappropriate techniques and products dte to institutional factors or price distortions are well-known causes which create segments of unemployment in the affected sectors.l/ 14. Policy makers in mixed economies have little leverage on labor supply behavior. But land distribution and tenurial arrangements, the rate and allocation of investment, and the nature of institutional and price 1/ A decomposition of employment growth in India and Zambia into the growth effect, the technology effect and the demand-mix effect has been attempted in two forthcoming papers.  regimes can all be influenced by public policy so as to increase greater labor absorption at reasonable levels of earnings. 15. One section of the paper reviews and synthesizes the recent literature on concepts for measuring sets of the unemployed by the time, income, willingness, and misemployment criteria, and the unions and intersections of these sets. The relationship of these sets with actually measured stock rates and flow rates is also clarified. April 1976.   1. INTRODUCTION 1. In this paper I present a conceptual framework for the measurement of rural unemployment and underemployment in a developing country. The framework is used to review and interpret all the available estimates of rural unemployment/underemployment developed in India on the basis of the Censuses of 1961 and 1971 and eight Rounds of the National Sample Surveyl! between 1958-59 and 1972-73. 2. The discussion of the conceptual problem of measuring unemployment has been advanced by numerous recent contributions (Hauser 1973, Krishna 1973, Sen 1973, Dantwala and Visaria 1974, ILO 1974, Sinha 1974, Ahuja 1975, Sabot 1975, Sen 1975 and Thormann 1975). The National Sample Survey Organization of India has also issued some of the tables resulting from its latest 25th Round (1970-71) and 27th Round (1972-73) surveys. In these Rounds, a new set of concepts based on the advice of the Dantwala Committee (1970) was adopted. The availability of these tables provides a good oppor- tunity to evaluate the new estimates, compare them with earlier estimates, and see how far they approximate the theoretical concepts suggested in the literature. 3. The task of conceptualising unemployment for the operational purpose of measurement and policy-making has two aspects. First, the quantitative concepts involved in the measurement of unemployment rates have to be clarified. Second, the merits of different sets of qualitative char- acteristics used to identify various kinds of unemployment must be critically examined. For it is these characteristics which the numbers collected in the surveys are supposed to reflect. 4. The clarification of quantitative relationships is attempted first in the next section, because much of the confusion in the unemployment literature is simply due to lack of clarity about these relationships. In particular, we discuss the units in which labor might be measured; the dis- tinction and the relation between stock-rates and flow-rates; and the aggregation of micro-rates into macro-rates and short-period rates into long-period rates. 5. Criteria for defining different kinds of unemployment are discussed in section 3 in the light of recent contributions. 6. The unemployment rates and the correlates of unemployment revealed by recent and earlier survey data for India are interpreted in sections 4 and 5 and exercises to measure surplus labor (Mehra 1966, Rudra 1973, Ahuja 1975) are reviewed in section 6. The relationship of surplus labor estimates and other,estimates of unemployment is also clarified. 1/ Denoted by N.S.S. in the text.  - 2 - 7. Throughout the paper, since the aim is to clarify the concepts involved in the collection and interpretation of survey data, labor supply and labor demand are treated as scalars observed at a point of time or over the reference period of a survey. But an analytical explanation of cross-section or time-series variations in unemployment.and wages clearly requires the specification of a model of the labor mark,et in which labor supply and labor absorption appear not as scalars but as functions of earnings and other variables, and are simultaneously determined. Such a model has been developed elsewhere.l/ But in the last section, the main features of the model are outlined. The model establishes the conditions under which different types of labor market disequilibrium (unemployment) and low-wage equilibrium can exist. It identifies the main variables which can produce each type of unemployment, and suggests policies required to increase labor absorption. 1/"A Five-Sector Model of Labor Allocation and Wage Determination in a Developing Economy."  - 3 - 2. OUANTITATIVE CONCEPTS FOR THE MEASUREMENT OF UNEMPLOYMENT RATES 8. In trying to measure the 'unemployment' of a working person (at the micro level) or of a working population (at the macro level) we are fundamentally in search of the difference between labor supply (Ls) and labor demand (Ld). If these magnitudes are known the unemployment rate (u) can be simply defined as the difference divided by the labor supply: U = U = _ (1) LS Ls 9. The dimensions in which Ls and Ld should be measured require careful consideration. Time is obviously the primary dimension for measuring the flows of labor. The units of time can be either natural hours, days or weeks, or some 'efficiency units' computed as equivalent to the natural time units of different types of labor. An efficiency unit can be defined as the amount of time required to yield either a constant physical output or a constant income to the worker or the economy. In the following argument when we discuss the measurement of labor in time- units, we shall be concerned only with natural time-units. For the productivity of natural labor-time varies widely as between persons, professions, enterprises and industries, and over different time periods, depending not only on the quality of the workers' subjective effort, but also the quantity and quality of the cooperating factors of production, and other circumstances. and therefore any effort to measure all labor flows in efficiency units must be regarded as a hopeless enterprise for the time being. The data required are simply not available and the continuous variations in the productivity of natural labor-time would render any standardization ratios obsolete before they are chosen. 10. But even though time units of labor cannot be standardized, it is essential that we try to measure the flows of undifferentiated labor- time as well as labor-time classified by skills and other characteristics. 11. When flows of labor are measured as flows of labor time it should be remembered that they are not only measured in time-units but they are also flows per unit-,of time. Thus we may have person-hours offered/demanded per day, or person-days per week, or person-days or person-weeks per year or even per life-time. 12. Alternatively, we may measure labor not as a flow of labor-time per period of time, but a stock of persons at particular points of time.1/ 1/ The third possibility of measuring unemployed labor in rupees suggested by Dandekar and Rath (1971) will be discussed below.  - 4 - 13. Two unemployment rates can then be defined: a flow-of-time rate FTUR and a stock-of-persons rate SPUR. Using superscripts t (time) and p (persons) to distinguish flow and stock quantities, we have stdt t FTUR = Ut = L Ldt _ U (2) Lst Lst and SPUR = u- = LSD sp U (3) ut and uP are flow and stock ratios and U and UP denote the total amount of unemployed time and the total number of unemployed persons, respectively. 14. A great deal of unnecessary controversy which accompanies the interpretation and comparisons of different unemployment rates can be avoided if the flow rates and stock rates are clearly distinguished. 15. The flow rates of unemployment of individuals can be easily aggregated into a macro flow rate. If ui is the unemployment rate of person i, and u the macro unemployment rate of a population of n persons, we can compute St- dt u = ( Li L. n (4) E LSt i=l I n -E ai ut i=l 1 where a. = L. / L . The macro unemployment rate is the weighted sum of micro unemployment rates, the weights being the individual shares in the total labour time supplied. 16. Consider, for example, a society of 6 persons whose labour supply and working time per week are given in the left panel of Table 1.  - 5 - Table 1: ILLUSTRATIVE MEASUREMENT OF THE FLOW UNEMPLOYMENT RATE Persons Days per week Ratio Days per week Ratio Lst Ldt ut ut Lst/Lst Est jt U-t A 6 5 1 0.17 0.20 7 2 0.29 B 4 2 2 0.50 0.13 2 0 0.00 C 5 5 0 0.00 0.17 6 1 0.17 D 5 0 5 1.00 0.17 6 6 1.00 E 6 1 5 0.86 0.20 7 6 0.86 F 4 4 0 0.00 0.13 7 3 0.43 Total 30 17 13 0.43 1.00 35 18 0.51 The macro unemployment rate (43 percent) is the weighted average of individual unemployment rates weighted by the proportions of labor time supplied by each person.l/ 17. Two characteristics of this formulation should be noted. First, the computation of the micro and macro unemployment rates is completely consistent. Since the macro u is an aggregation of micro u's it fully reflects variations in the labour time offered as well as the work obtained by individuals. 18. Second, 'full employment,' 'full unemployment' and 'underemployment' are measured along the same continuous axis so that Ui or u = 0 denotes full employment, Ui = 1 denotes full or open unemployment, and 0<(ui or u)<l denotes underemployment. All degrees of underemployment can be measured by ui or u assuming all possible values between 0 and 1. Underemployment can be seen as incomplete employment (unemployment), and full employment/unemployment as the limits of decreasing/increasing underemployment. 1/ .43 = (.17) (.20) + (.50) (.13) + (.00) (.17) + (1.0) (.17) + (.86) (.20) + (.00) (.13)  19. The relationship between the stock-rate and the flow rate is extremely important. The stock-status of a person at a point of time (when the Census/survey takes place) is determined by his/her flow-status over some period of time just preceding the Census/Survey--the so-called reference period. The survey agencies must arbitrarily fix cut-off levels (norms) of labour time supplied and the flow rate of.unemployment (Lst and ut* ) so that a person whose supply Lst exceeds Lst* in the reference period would be counted as a member of the labor forcel/ LSP and a worker whose ut exceeds ut* in the reference period would be counted as underemployed/unemployed. The most non-arbitrary procedure would be to include everybody whose labor supply is positive (LSt> 0) in the labor force, and everybody whose underemployment rate is positive (ut 0) as unemployed or underemployed. Persons whose ut = 1 and ut = o would obviously be counted as fully (openly) unemployed and fully employed. 20. If several levels of ut* are chosen as cut-off points to distinguish different degrees of underemployment each person can be placed in one class of the underemployed. When every person's stock-status is determined, the stock-of-persons underemployment/unemployment rate(s) can be computed. 21. The upper panel of Table 3 illustrates the measurement of stock- rates with the data of Table 1, on the assumption that the survey agency fixes the cut-off limits shown in Table 2. The cut-off limits can be specified either in terms of the ratio of labor demand to labor supply or ranges of the flow rate of unemployment. Both specifications are illus- trated in Table 2. 22. Since all members of the labor force are being placed in one of the five categories depending on the fraction of their supply of labor-time for which they obtained work or their time rate of unemployment in the reference period, there will be five stock rates. Let UPJ be the number of persons in category j. Then upj = Usp (j=l . . . . 5) (5) l/'Labor force' is a convenient term to denote the stock of workin_and idlepe_esons. 'Labor supply' on the other hand can be used to represent the flow of labor time supplied per unit of time.  - 7 - Table 2: ASSUMED CUT-OFF LIMITS FOR COUNTING UNEMPLOYED/UNDEREMPLOYED PERSONS For inclusion of a Positive labor supply in the reference person in the labor week. force. (Lst > 0) For a person being Labor demand equals labor supply in the counted as fully reference week. employed. (Ld = LS) or (ut = 0) For a person being Labor demand exceeds 66 percent of labor counted as marginally supply in the reference week. underemployed. (.66LS < Ld < Ls) or (0 < ut < .33) For a person being Labor demand lies between 33 and 66 percent counted as moderately of labor supply in the reference week. underemployed. (.33Ls < Ld < .66L)or (.33 < ut < .66) For a person being Labor demand is less than 33 percent of counted as severely labor supply in the reference week. underemployed. (O < Ld < .33Ls) or (.66 < ut<1). For a person being Labor demand = 0 in the reference week. counted as fully d t unemployed. (L = 0) or (u = 1).  Table 3:- ILLUSTRATIVE MEASUREMENT OF THE STOCK IUNEMPLOYMENT/UNDEREMPLOYMENT RATES Members of Fully Marginally Moderately Severely Fully All Unemployed Categories the Labor Employed Underemployed Underemployed Underemployed Unemployed or Underem- Force Persons Persons Persons Persons Persons ployed Persons Criteria Lst > 0 ut = O 0 < ut < .33 .33 < ut < .66 .66 < ut < 1 ut= ut > 0 Persons 6 2 1 1 1 1 4 uPl = Stock rate of Full Employment = .33 uP4 = Stock Rate of Severe Underemployment = .17 uP2 = Stock Rate of Marginal Underemployment = .17 uP5 = Stock Rate of Full Unemployment = .17 uP3 = Stock Rate of Moderate Underemployment = .17 4 5 E uP3 = .50 E UP] = .66 l j=2 j=2