blSCUSSION PAPER Report No. DRD103 TRENDS IN REAL WAGES IN RURAL INDIA 1880-1980 by Deepak Lal June 1984 Development Research Department Economics and Research Staff World Bank The World Bank does not accept responsibility for the views expressed herein which are those of the author(s) and should not be attributed to the World Bank or to its affiliated organizations. The findings, interpretations, and conclusions are the results of research supported by the Bank; they do not necessarily represent official policy of the Bank. The designations employed, the presentation of material, and any maps used in this document are solely for the convenience of the reader and do not imply the expression of any opinion whatsoever on the part of the World Bank or its affiliates concerning the legal status of any country, territory, city, area, or of its authorities, or concerning the delimitations of its boundaries, or national affiliation. Revised, June, 1984 This paper is for a volume on 'Poverty in India' edited by Pranab Bardhan and T. N. Srinivasan. TRENDS IN REAL WAGES IN RURAL INDIA 1880-1980 by Deepak Lal University College, London and - Development Research Department, World Bank Part of the research for this paper was supported-by the World Bank under its RPO 671-84.- Comments by T. N. Srinivasan and statistical assistance by Masocd Ahmed are gratefully acknowledged. The World Bank does not accept responsibility for the views expressed herein which are those of the authors and should not be attributed to the World Bank or to its affiliated organizations. The findings, interpretations, and conclusions are the results of research supported-by the Bank; they do not necessarily represent official policy of the Bank. The designations employed, the presentation of material, and any maps used in this document are solely for the convenience of the reader and do not imply the expression of any opinion whatsoever on the part of the World Bank or its affiliates concerning the legal status of any country, territory, city, area, or of its authorities, or concerning the delimitation of its boundaries, or national affiliation. Abstract This paper charts long-term trends (over a century at the All India level) in rural real wages, which can be derived from available administrative statistics on rural wages, to see if the conclusions of an earlier study (Lal, 1976) based solely on National Sample Survey data (for the post 1950 period) concerning a positive association between trends in rural real wages and in the changing balance between the demand and supply of rural labour, are upheld. A real rural wage series for a century is derived, and used with time series data on labour supply and agricultural output to estimate a simple neo- classical model of wage determination for the whole 1880-1980 period and for three sub-periods. It is found that some mild support is provided, particularly for the pre 1914 and post 1950's period for a --classical model of rural wage determination. The paper also summarises the results from other studies concerning the post Independence performance of rural real wages, which also support the "neo-classical" model of wage determination as a working hypothesis. The paper lends support to the view that, given the likely future growth in the rural labour force, raising the demand for labour remains the single most important means for alleViating rural poverty. The paper's estimate of the increase in the rate of growth of agricultural output in the post Independence period to 2.4% p.a. over its long-term trend of just under 1% p.a., has not been spectacular in relation to the growth in rural labour supply (of about 2% p.a.). Hence, there has been no acceleration of the rate of growth of rural real wages in the post Independence period over the long-run trend rate of under half a percent per annum, and hence it would be unrealistic to expett any significant reduction in the extent of Indian rural poverty. But to the extent that the supply of rural labour is also influenced by the absorption of labour in industry, it is the failure of modern Indian industry to markedly increase its demand for labour despite high growth rates of investment which must also be assigned a major role in the limited poverty redressal in India. Introduction Delineating trends in rural real wages in India has been highly controversial, in part because of the different time periods over which the trends are charted. 1/ The passion the subject still arouses is similar in many ways to that surrounding the statistical debate concerning the trends in the terms of trade of developing and developed countries. In an earlier paper (Lal, 1976) I had questioned the empirical basis of the view that within the existing agrarian st-ucture, agricultural growth which raises the demand for labour will be in sufficient to raise rural real wages and/or alleviate rural poverty. 2/ On the basis of the data available from the National Sample Survey for 1956-57 and 1970-71 (both agriculturally 'normal' years), for changes in the real wages of male agEicultural labourers, and changes in the household consumption of the 'weaker sections' in the rural areas of the states in India, I had argued that earlier studies which had sought to show, not merely that, the agricultural growth flowing from the Green Revolution had not led to real wage rises but had led to the immesirisation of the rural poor, were flawed, as they were based on the unreliable data provided by the administrative statistics on money wages and prices collected and published by the Ministry of Agriculture in Agricultural Wages in India (AWI). As Rao (1972) had shown, in his comparison of the agricultural wage data from the NSS and AWI that, the latter were not scientifically collected as compared with the former, I had argued for relying solely on the NSS data in charting real wage trends. This conclusion has been found to be too extreme by some (see Jose (1978)) who have cited Rao in their defense, as not excluding "the possibility of (the AWI data) being used for analysing spatial variations and long-term trends in the behaviour of wage DLD/AR-078/7/23/84 -2- rates." (Jose, ibid, p.A16". This paper therefore is concerned with charting long term trends in .rural real wages, from the available administrative statistics on rural wages, to see if the conclusions of the earlier study based solely on NSS data, concerning a positive association between trends in rural real wages and in the changing balance between the demand and supply of ruraly labour, would still apply. But whereas past studies of real wage trends (including our own) have sought to chart these trends largely in terms of end point comparisons of the level of real or money wages, at different dates, in the post Independence period, in this paper I attempt to present a time series analysis of the trends in a 'composite All India rural real wage rate" (derived in the first part of the paper), over the century 1880-1980. This analysis is undertaken in the second part of the paper. Needless to say, given my published qualms about using these administrative statistics, with their measurement biases of unknown direction and extent, as well as those unknown errors involved in deriving an All India aggregate figure, I would not claim any definitiveness whatsoever for the reported trends, and hence all the results in this paper must be treated with an even greater degree of caution than is usual. On the other hand economic historians, who do not have the luxury of ever hoping to find definitive quantitative evidence for many aspects of the past, may not be too worried about such scruples, and may find our derived trends over a century of some interest. DLD/AR-078/7/23/84 -3- The third part of the paper seeks to explain these trends in terms of a simple supply and demand model of wage determination. In this context it should be noted that the surplus labour theory of Lewis, whilst cQo.qpatible with a demand and supply determination of rural wage rates, asserts that given the assumed perfectly elastic supply curve of rural labour, raising the demand for rural labour would have no influence on th' rural real wage until the labour surplus has been removed. 3/ The empirics in part three should therefore also be useful for determining the likely validity of this view for rural India. in this third part we also summarise the results of recent studies which are relevant in assessing the disputes concerning (a) the existence of surplus labour, (b) the changing incidence of rural poverty 4/ and (c) the process of rural wage determination for which the reading of the entrails from real rural wage trends has been considered to be important. II Derivation of Money and Real Rural Wage Series - 1880-1980 Till 1873 there are only scattered literary accounts which provide some indication of likely wage trends. Radhakamal Mukherjee has attempted to examine real wages from the time of Akbar (1600) to 1938. He finds that over this whole period there is a 50 percent decline in real wages, but for the 1850-1940 and 1900-40 period he shows substantial increases (see Col (1), Appendix Table 1). For the 1850-73 period, scattered evidence from a number of dubious sources is summarised by M. Mukherjee (1969, pp. 87 & 88), which show a fairly constant real wage over this period. For the period from 1873 till the early part of the 20th century, a relatively more systematic source of wage data is available in the serial official publication Prices and Wages DLD/AR-078/7/23/84 -4-- (PAW), the precursor of the AWI. Its faults are similar to those of the post Independence AWI data. It was rejected in 1914 by K.L. Datta (1914), as agricultural wages were collected by untrained reporting agencies, who commonly reported cash wages for labourers in towns and, their neightbourhood, and these were unlikely to be typical of the rates in the relevant: rural areas; a critique which echoes that of Rao (1972)'s of the post Independence AWI data. This PAW data is given in col (2) Table A.1. For the 1900-46 period, the best data available according to Sivasubramonian (1977) is from the Labour Gazette of the Government of India, for Bombay province. The workers are classified by urban and rural sectors, and as field labourers, ordinary labor and skilled labor. For our purposes it is the 'field labor' series which is relevant. This "includes all workers who are engaged in agriculture pursuits and are actively employed in wages in occupations such as ploughing, F,)wing, transplanting, weeding and reaping" (Sivasubramonian, 1977, p. 480). This series is shown in column (3) Table A.1. For the post Independence period, the AWI data by state has been assembled by Jose for the period 1956-57 to 1971-72. He uses wage rates for the category of 'male field labour', or for 'ploughman' where the former is not available, by district for one month in the peak agricultural season in each state. The district figures are weighted by the male agricultural laborers in the district to obtain the statewise average. His data, and the price deflators (the agricultural consumer price indices) are summarised in Table A.2. DLD/AR-078/7/23/84 -5- The data from the NSS for 1950-51, 1956-57, 1964-65, 1970-71, and 1974-75 is summarised in Table A.3.. This relates to average wages of males aged between 15-59 of non-cultivating wage earner households in each state. Finally, Table A.4 provides the statewise data on the increase in money and real wage rates of male agricultural labourers in India between 1970-71 and 1979, based on the AWI data partially published in the Ministry of Agriculture's Agricultural Sittation in India, Sept. 1979. This is the only currently available data to form any judgment on real rural wages in the late 1970's. We have combined the PAW data from 1873-1900, with the rural series of Sivasubramonian for the period 1919-46, and the AWI series from 1956-57 to 1970-71, from Jose's data with the 1979 (partial) AWI data, and interpolations for 1950-51 and 1974-75 from the NSS to derive the All-India rural money and- real wage series given in columns (1) and (2) of Table 1. it should be noted that for the post Independence period the statewise data shown in the Appendix Table has been aggregated by weighting it with the statewise data on the male agricultural labour force from the 1961 and 1971 Censuses. Moreover, the price index used to derive the real wage till 1950, is that given (and used) by M. Mukherjee, based on All-India wholesale prices, which is shown in column (4) of Table 1. For the post Independence period, the respective real wage estimates given for each state were derived by applying the ACPI for the state, and the resulting real wage rates were aggregated as described above. The implicit All-India ACPI of our post-Independence series is not reported, but can be readily derived from the money and real wage figures for these years. We have, however reported the wholesale price index for food commodities from 1950 onwards, linked to the Mukherjee series in column (4) of Table 1. DLD/AR-07817/23/84 -6- II Time Series Analysis of Trends in Composite Real Wage Series We first fitted a geometric trend line to the 74 data points for real wages that we have in Table 1. This yielded the following equation: (R.1) RLW(t) = 102.8 (1.0048)t r2 = 0.40; DWS = 0.32 where RLW is the real wage, and t is time. Table 2(I) gives the estimated autocorrelation function, and Table 2(11) gives the computations of the partial autocorrelation function (PACF) for successive autoregressive (AR) model fits, to the residuals of the trend equation R.1. From the correlograms and the use of Quenouille's procedure (namely, where at 95% confidence, VT times the PACF > 1.96) (see Gottman, p.248), it would appear that an AR(2) and possibly and AR(3) process would generate the observed residuals, (as T=74; and VT.PACF(3)=2.32; /T.PACT(2)=-1.29; VT.PACF(1)=6.45). . 5/ We therefore estimated the following AR(2) model of real wage trends: Log RW(t) = a + bt + ut (1) ut plut-I + P2ut-2 + e The results for the whole period and sub-periods are summarized in Table 2.111. (Note that in rows 4 and 6, we have fitted the above model to a data set in which the data for the missing years has been filled by linear interpolation - NAINTERP.) The trend growth rate of real wages over the period 1867-1978, of 0.4% per annum is statistically significant. Figure 1 charts the real wage series, and the computed trend line (for the case where NAINTERP has been used). DLD/AR-078/7/23/84 -7- For the sub-periods the estimated geometric growth rates were 1873-1900: 0.00% p.a. 1919-1947: 0.47% p.a. 1950-1978: 0.36% p.a. (using NAINTERP) Thus real wages would seem to have been stagnant in the late 19th century, and havd!been rising by about about 0.4 to 0.5% per annum in the 20th, with a small fall in the rate of growth in the post Independence period. Trends in Agricultural Output and Male Rural Labour Force We would expect these real wage trends to be related to the trends in agricultural output and the growth of the rural male labour force. Columns (4) and (5) provide the constant price series for agricultural output, and male labbur force, in Table 1, whose notes explain the sources. The geometric trend fitted to the data on agricultural output for the whole period,1868-1979, yielded: (R.2) Y(t) = 57.7(1.0089)t r2 = 0.8425; n = 99; DWS = 0.58 Table 3(I) and (II) show the computed ACF's and PACF's for the residuals from R.2. Using Quenouille's procedure, an AR(2) process is sugg!ested. (as T=99; /T.PACF(3) = 1.61; /T.PACF(2)=4.6; /T.PACF(1)=6.95. Fitting an AR(2) model the agricultural output series yielded the results shown in Table 3.III. For the 1876-1979 near century, agricultural output was growing at the compound rate of about 0.9% per annum. In the sub- periods the rates of growth were 1885-1900: 0.6% p.a. 1919-1946: 0.6% p.a. 1950-1978: 2.4% p.a. DLD/AR-078/7/23/84 -8- Thus there is a marked acceleration of agricultural output growth in the post- Independence period. Whilst, from column (7) of Table 1, we can fit the following geometric trends to the male labour force (rural) data for the same periods: (R.3) 1885-1980 E(t)=82.2(l.01048)t r2=0.91; n=96 (a) 1885-1900: E(t)=95(l.0038)t r2=0.84; n=16 (b) 1919-1946: E(t)=107.7(1.009)t r2=0.98; n=28 (c) 1950-1978: E(t)=149.2(1.019)t r2=0.98; n=29 Table 4, summarises these estimated geometric growth rates in real wages, agricultural output and the rural male labour force, as well as reporting the compound rates of growth for these three variables, as well as for agricultural output per rural male worker (derived in column (6) of Table 1) for the whole 1873-1980 period. The only significant point worth noting is that, in the post-Independence period, there has been no change in the growth rates of real rural wages, but a rough doubling in those for agricultural output and the (male) rural labour force, over their long term trend growth rates. DLD/AR-078/7/23/84 -9- III The Determinants of Real Rural Wage Trends In this section we seek to explain movements in the rural real (and money) wage series compiled in Table 1, for the past century (1880-1980), at the All-India level, in terms of a simple neo-classical model of rural wage- determination. For those with an a priori belief in the validity of such a model, its success in explaining the movements in our derived rural wage series might provide an indirect validation of that series. We proceed to derive a reduced form equation for estimation purposes, from a highly aggregative model, in which agricultural output (X) is produced by one variable input, labour (L) and a fixed input land (N), through a Cobb-Douglas production function: X=A.Lo:.N (2) The money wage rate is denoted by w, the price of agricultural output by p1 and the real wage w/p = W As we will be treating output (X) as exogenously determined, the restricted cost function (C) is given by C = G( w ;N,X) (3) From which the derived demand function for labor (Ld) is L (4) d For the Cobb-Douglas production function in (2) (where A = aP ), the restricted cost function will be (see Varian p. 67): C =X.w'P.NP (5) From (4) and (5) f hd N (6) from which taking logs and rearranging yields the labor demand function as: DLD/AR-078/7/23/84 -10- InL d = Ina + lnX + (a-1)lnw + $ln.N. (7) Labor supply (Ls) is taken to be a function of the real wage rate (w) , and its cpnstant elasticity form is specified as ln.Ls = Tny0 + Y1lnw (8) the reduced form for percentage changes in the real wage (d ln w (t)) is then given from (7) and (8) by: dlnw(t) = a0 + a1.dlnX (t) + a2.dln N(t) (9) Table 5(I) shows the estimated equations for the above model. It is apparent that though the model fits the data for the late 19th century and the Post Independence period faily well for the case with a one period lag on the output (X) variable (for which alone the results are reported), the post World W%r I period's 'fit' is non-existent. As there was considerable .price variability in this period as compared with the ones preceeding and succeeding it, and as it has been found that money wage changes tend to lag those in prices 6/ we fitted another model in which the demand for labour depends on the current real wage as in equation (7), but labour supply depends on last year's real wage 7/ so that, instead of (8) we have: InLs = In* + p1ln w(t-1) (8") From (7) and (8'), the reduced form estimating equation for the second wage determination model is: dlnw(t) = b + b dlnX(t) + b2dlnN(t) - b3d1nw(t-1) (9") The results again for the case where output is lagged one period are reported in Table 5(11). Though the overall fit improves marginally for the 1921-1946 DLD/AR-078/7/23/84 -11- period, the sign on the only significant variable (the lagged real wage) is wrong. This second model in fact only performs well for the 1892-1900 period. Thus all that we can conclude is that our wage determination models seem to work for only the 19th century and post Independence periods. The results reported for the full period 1892-1978 through linear interpretation of the missing data cannot be taken too seriously, whilst that from pooling the data for the three sub-periods (5(I),d(i)) does not provide too good a fit, in part because the length and hence the weight of the inter war and pre- Independence period is over half. As during this period much of the variation in real wages was due to variations in prices, the role of the other two factors (agricultural output and changes in sown area) which were of greater importance during the previous and succeeding periods are understated. Hence, in Table 5(I) and 5(11) we have also reported the results from excluding these years, and these show that the output factor as well as the lagged wage effect on labour supply in'rural wage determination have been significant, in the remaining years of our series. Hence, we would conclude that our computed real wage trends may not be inconsistent with a neoclassical model of rural wage determination, and mutatis mutandis to the extent we believe independently in the validity of such a model, the results of Table 5 could lead us to conclude that our computed rural wage series for 1880-1980, is perhaps not yielding purely spurious trends! As we do not have any data on rural wages by region for this whole period, no attempt has been made to chart regional trends, which of course are concealed (even though they may be considered to be of much greater importance), in the aggregative exercise of this paper. However, for the post-Independence period a recent attempt to explain real wage changes in the DLD/AR-078/7/23/84 -12- Punjab at the district level, using district level AWI data may be noted. 8/ This study by Bhalla, though not based on an explicit model of wage determinaticn, attempts to explain variations in real wage rates of agricultural labourers in the Punjab between 1961 and 1977 in terms of changes in output per rural male worker and in.the price level. She found that: "the rapid growth of farm output tended to push up real wages; the growth of the labour force at rates far above the rate of growth of population tended to pull real wages down. The outcome, it appears, was often decided by inflation, which periodically eroded real wages to levels below even those attained in 1961. As of 1977, for the Punjab as a whole, the forces of rising productivity had won. But it was a narrow victory. Over the entire period, real wages for male agricultural labourers, went up by between six percent (for weeding), and 20 percent (for harvesting)" (Bhalla, p.A-57). However the end point estimates of regional wage trends in the 1970's from the AWI data in Table A.4 show that real wages declined in the Punjab by 4.5% between 1970 and 1979. But as 1979 was an agriculturally poor year, these end point estimates of trend can be treacherous. For this reason the other declines in regional wage rates for 1979 as compared with 1970 for the AWI data shown in Table A.4 cannot be used to draw inferences about either rural wage determination or of the effects of agricultural growth on real wages. Finally, we may note that in so far as the 'surplus labor' model postulates a constant rural real wage, our computed real wage trends for the past century do not provide support. Nor can it be concluded that increasing population pressure makes it likely that such a model will become more applicable in the future. For there has been a positive association between population growth, agricultural growth, and the growth of rural real wages. Whether there is any causal link between population and agricultural growth DLD/AR-078/7/23/84 -13- requires an analysis beyond the scope of this paper. 9/ Our analysis does however support the conclusion, in contradiction of the surplus labour model that ceteris paribus, an increase in the demand for rural labour does lead to a rise in rural real wages. This conclusion is further strengthened by the recent-studies by Bardhan, Rosensweig, and Bingswanger-Evenson of rural labour supply and demand functions based on the micro-economic household data available from the rural household surveys of the NSS and the NCAER. 10/ They find that both the demand and supply curves for rural labour are relatively inelastic, so that even small shifts in either of these curves can lead to greatly amplified changes in real rural wages. The 'surplus labour' model, hopefully, will henceforth be regarded as a theoretical curiosum in Indian economics. IV Our conclusions therefore can be brief. First, our investigations in this paper do lend some support to Rao's claim that the AWI data may be useful 'in charting long term trends. Secondly, as in all time-series analysis, it is important not to draw any conclusions on trends from a short run series. Thirdly, it is important to formulate and if possible estimate an explicit theoretical model of rural wage determination, which has not been common in past Indian discussions. 11/ Fourthly, as a working hypothesis, our analysis would support the use of a 'neo-classical' model of wage determination. A conclusion that is also supported by the recent studies of household rural labour supply and demand functions from Indian micro-economic data. Fifthly, to the extent that rural money wages tend to lag behind prices, and rural real wages are of importance in determining the living standards of the rural poor, inflation is an enemy of the wage earning rural DLD/AR-078/7/23/84 -14- poor. Sixthly, given the likely future growth in the rural labour force, raising the demand for labour remains the single most important means for alleviating rural poverty. Though agricultural development (even in the absence of institutional reform) can and has raised the demand for rural labour in the past, it has not been spectacular in relation to labor supply. The marked increase in the rate of growth of agricultural output in the post- Independence period to 2.4% p.a., over its long term trend rate of just under 1% p.a., has however not led to any acceleration of the rate of growth of rural real wages in the post-Independence period over the long run trend rate of under half a percent per annum. Given the modesty of this agricultural performance relative to the increase in rural labor supply (of about 2% p.a.) it would be unrealistic to expect any significant reduction in the extent of Indian rural poverty. But to the extent that the supply of rural labour is also influenced by the absorption of labour in industry, it is the failure of modern Indian industry to markedly increase its demand for labour, despite relatively high rates of growth of industrial investment, which must also be assigned a major role in the limited poverty redressal in independent India. But that, of course, is another story. DLD/AR-078/7/23/84 -15- NOTES 1. K Bardhan (1973) (1977); P. Bardhan (1970) (1973); Bhalla (1979); Herdt and Baker (1972); Jose (1973), (1974), (1978); Krishnaji (1971); Lal (1976), (1977); Parthasarthy and Adiseshu (1982); Pandey (1973), (1976); Santra (1974). 2. Vernon Ruttan has noted an ideological tone in his survey of the literature on the effects of the Green Revolution on the living standards of the rural poor. He states: "A second set of criticisms has been more ideologically motivated. Hope for the radicalization of the lower peasantry and landless labourers has been viewed as dependent on the continuation of the process of 'immiserizing growth'. The scale neutral green revolution technology, which was equally effective on small and large farms, has been viewed as increasing the political cost of revolutionary change. By offering a prospect for improvements in the welfare of rural people, without revolution, the seed fertiliser technology has become the target of substantial rhetoric." (p. 16). 3. The surplus labour literature is immense. But mention must be made of Lewis (1954) and (1979), in which he seems to view the model of surplus labour now in terms of the segmented labour market models for industrial labour. The latter are surveyed in Lal (1979). On the necessary and sufficient conditions for surplus labour to exist in Lewis' original sense the seminal article is by Sen (1966), who shows that these conditions are provided by a perfectly elastic supply curve for rural labour. An attempt to directly test this assumption with the data available from the NSS 25th round is in Lal (1976a). But the most rigorous empirical critique of the DLD/AR-078/7/23/84 -16- ' surplus labour view is provided by the estimates of household rural labour supply curves by P. Bardhan (1979) and Rosenzwig (1978), (1983). 4. Though oaly incidentally. For the positive effects of real wage rises on alleviating rural poverty see Lal (1976), and on the positive effects of . growth on rural poverty redressal see B. Dutta (1980) for the latest estimates, as well as a brief survey of earlier studies. 5. It should no noted that the stationarity conditions for an AR(2) process are: -1