A survey of research on rural employment, wages and labor markets in India

A SURVEY OF RESEARCH ON RURAL E24LPLOYMIENT, WAGES AND LABOR I4AFR&ETS IN ItDI -D - by Kalpana Bardhan Series : Studies in Employment and Rural Development No. 39 Division : Employment and Rural Development Department: Development Economics DeveloFpent Policy Staff International Bank for Reconstruction and Development This paper provides a selective survey of empirical work on rural labor markets in India. Section I su,veys the literature on size, characteristics and composition of the rural working force and rural employment, particularly with a view to changes which have occurred through the decade of the sixties and early seventies. The focus of Section II is on the employment effects of the seed-water- fertiliser innovation and those oL the application of tractor techno- logy to different farm operations. Section I!I covers the empirical work connected with some of the major hypotheses of wage determination in agricultura: the dual labor market hypothesis; the hypotheses o' institutionally-determined and nutritionally-determined wage rates; the seasonality factors in the wage determination process; and the role of supply and demand in determining wages. Finally, Section IV surveys the available data, and the work done so far with these, on labor time disposition in different categories of housenolds on one hand, and the farm-level determination of demand for labor and its composition of family and hired labor on the other. The author prepared this research paper in her capacity as consultant to the Bank. This paper forms part of a research study on Structure of Rural Employment, Inco-me and Labor Markets (RPO 671-30) undertaken in this Division. The views expressed represent those of the author and not necessarily those of the Bank. This report may not be published nor may it be quoted as representing- t-he views of the Bank and its affiliated organizations. Washing ton, D.C., June 1977  A SURVEY OF RESEARCH ON RUTRAL EMPLOY'NT, WAGES AND LABOR iM4ARKETS IN INiDIA TABLE OF CONTENTS Page No. IntroductorY Summary .....................................................i - xiii Section I. Size and Com-oosition of Rural Working Force ...... I I.a. Recent Trends from the Census .................... 1 I.b. Correlates of Change in the Proportion of Agricultural Labourers ......................... 5 I.c. Female Work Participatior in Rural Areas ......... 9 I.d. Principal and Subsidiary Work Combination ............... 10 I.e. Changes in Rural Working Force from National Sample Survey ......................................... . 12 I.f. Migration Pattern of Rural Labour ........................ 13 I.g. Background of Rural Migrants ..................... 18 I.h. Willingness to Migrate for Regular Work. . .......... 19 I.i. Seasonal Migration of Rural Labour ............... 22 Section II Agricultural Growth and Technological Change: Impact on Wages and EmDlov ment ......................... 24 II.a. Nature of Technological Change and Implications for Labour Demand .. ...................... 24 II.b. Changes Over Time in Emplcyment of Rural Labourers ....... ...... ................... 33 II.b.(i) Employment and Underemployment in Rural Areas: Changes Over Time ................................ 34 II.b.(ii) The Surplus Labour Question ...................... 38 II.c.(i) Changes in Wage Rates Over Time ..................41 II.c.(ii) Labour's Share in Production Growth .............. 47 II.c.(iii) Correlates of Cross-Sectional Wage Variations .... 48 Section III Labour MIarket Imperfections and Wage Determination 52 III.a. Hypotheses .................................................. 52 III.b. Empirical Basis: Nutritionally Determined Wages- 54 III.c. Labour Mlarket Dualism .................................... 59 III.d. Some Emerging Employer-Labour Relations ................ 64 III.e. Seasonality in Agriculture and WZage Determinacion. 66 Section IV 'Micro-Level Correlates of Labour Demand and Labour SuDlD. . . ................. ............. 70 IV.a, Data . ............................................ 70 IV.b. Studies on Labour Demand and Supply ............ .71  TABLE OF CONTENTS (continued) Pagce N30. IV.c. A Study of Correlates of Labour Demand and Supply in Rural Areas .................................... 73 Bibliography 1 - 8 LIST OF TABLES Table No. Title Page tio. i Rural TWorker Participation Rates (%), 1961-1971 3 ........................................ 3 2 Percentage Distribution of RURAL Workers b,y Categories of Principal Occupation 1961 and 1971.. 4 3 Percentage Distribution of rnmigr2ant and Non- migrant Resident Workers in Rural Areas by Mlain Occupation .. ................. ................... ..16 4 Change in Agricultural Labourers' Wages and Cost of Living in Punjab and Kerala, 195, -57 to 1970-71 (1971-72) .............................43 5 Change in Average Money and Real Wage Rate for 'Male Agricultural Labourers ......................46  RURAL EMPLOYMENT, WAGES AJND LABOUR IMA,RKETS IN INDIA-^ Introductory Su1mnmarY 1/ 1. To understand the operation of the rural labour market in India, the question of how the employ-ment and wages of the landless or near landless are determi ned must be seen in the overall context of modes of production, types of employment and employer-labour relations. Even if the subject of investigation is a rather narrow one, such as the determination of wages, it is necessary to lcok at the entire structure ofi rural employment. 2. For most parts of the country, it would be a gross over-simplification to see the agricultural economy in terms of separate categories of capitalist farmers: self contained peasants, and landless labourers. The majority of rural households combine self employment and wage employment. In different areas there exist various combinations oL (1) capi- talist farmers producing almost exclusively with hired labour; (2) feudal landowners with sharecropping tenants; (3) peasant households almost exclusively engaged in subsistence produc- tion; (4) farm households employing both family and hired labour, and whose members may at times work for wages else- where; and (5) households who, not havirng enough productive assets for self-employment, are almost fully dependent on wage employment. Households ol the fourth group, who both seek and offer wage employment--either simultaneously or in different seasons of the year--play a very significant role. 3. The non-agricultural sector of the rural economy, traditionally small and dependent on the agrarian activities of the village, now provides more wage employment in public works, private construction, marketing and transDort, and more self-employment in trading and repair services, replac- ing some of the self-employment in declining traditional vocations. These changing patterns of production and income- earning opportunities make a broad approach to analysis of the rural labour market even more necessary. 1/ This is a self-contained intrcduction giving a broad outline of the findings of the review.  4. The characteristics of rural employment and modes of production are still the subject of much controversy. There is a virtually endless series of village surveys and small scale empirical studies, describing i n minute detail the Labric of a rural economy which varies a great deal from one region to another. There are also systematically collected large scale data sets. The present survey of empirical research is selective in coverage and concentrates largely on the period since 1960, which has been marked by significant change in agriculture and in the rural economy in general in many parts of india. Changing Structure of the Rural Labour Force (See Section T, pp. 1-23) 5. The past fifteen vears have seen an increase in the rural proletariat. The proportion of wage labourers in the total rural labour force, stable during the l950s, has grown rapidly since the early 1960s and there has been a corresponding decline in the proportion of cultivators and other largely self-employed workers. The spread of new agricultural technology has accentuated this trend, but the data suggest that the rural poor were already relying increasingly on the market for employment before the Green Revoluti on took hold. 6. The tenurial reform legislation adopted in the late 1950s and early 1960s was intended to abolish intermediary tenure, giving tenants security and eventual rigl to the land they farmed. The results belied the intentions in many areas, particularly in the north, where the new laws led to the outright eviction of many tenants. Large landowners whose property had traditionally been farmed by tenants now began managing their own land with the help of wage labour. Former tenant cultivators swelled the ranks oL the agricultural proletariat. 7. This process was reinforced by the dramatic increases in the profitability of farming being made pos- sible wIth the new technologc, subsidies for agricultural inputs and a tax system that was lenient towards agricultural profits. Public investment in irrigation, fertilizer indus-  tries and in the supply of credit and other vital infra- structural services was considerably increased in the 1960s. At the same time, wit-h the availability of high yielding seed varieties and fertilizers, private investment in irrigation and other elements of technological change was rising rapidly. Fiscal policies--price support, input subsidies, liberalisation of tractor imports, continued under-taxation of agricultural profit--favoured the growth of agricultur.al out put. These policies influenced farmers' investment choices between alternative technologies, some- times with labour-displacing effects. In most areas, the progressive farms were generally large: small farmers and tenant cultivators did not have such easy access to credit, new seeds or other modern production inputs, quite apart from their lesser ability to bear risk, and this ensured that relatively little innovation took place among them. Biases against small farmers in general, and tenants in particular, in the distribution of credit and other infra- structural facilities have directly constrained the growth of employment. 8. In areas (Maharashtra and Gujarat) where land reform was better implemented, the results were inadequate. Without supportive measures, yields on the ex-tenants' tiny new holdings remained stagnant, while the size of the families dependent on them went on rising, and by then it was virtually impossible to lease in additional land. More and more members of small cultivating households became labourers. In some of the agriculturally progressive areas, observers have noted that leasing by large to small farmers has not only declined or ceased, but has even been reversed, with small farmers pushed by their lack of working capital or pulled by rising wage rates to cease operating their own land. In the prime green revolution" area of Punjab, the distribution of opera- tional holdings has become more concentrated than the dis- tribution of ownership holdings. 9. Outside t'ie farm sector, with the decline in tra- ditional means of self employment, many traditional artisans have either migrated to clties or have become agricultural labourers. Though rural-urban population ratios have been declining, the absolute size of the rural population and rural labour force have continued to increase.  - iV - 10. Rural policies and programs need t o be evaluated in the context of these changes in the strructure of the labour force. Growth in the demand for labour, stimulated by agricultural growth or technological changes, must not only be large enough to absorb the gro.'7iing rural labour force, but also directed sufficiently towards wage labourers. The accepted objectives of economic growth wi th dispersed benefits, and employment of the growing rural labour force, must be pursued within an integrated policy Lramework, coordinating land reforim measures, progressive taxation of agricultural incomes and holdings, price policies for inputs and outputs, infrastructural investment and rural works schemes so that the results reinforce each other. Emplovment Effects of Technological Change (See Section II, pp. 24-33) 11. In evaluating the employment effects of recent agricultural growth with technological change, or project- ing the employment efLects of innovation on the basis of alternative policy environments, it is rightly considered important to distinguish betwieen the elements of the tech- nological change "basket" that are not necessarily bound up wiL. each other. The three main elements from this point of view are: adoption of the water-seed-fertilizer innovaation, tractorisation, and use of purely labour-saving devices like harvesters. The last of these obviously has a negative effect on employment. All available studies indi- cate that the seed-water-fertilizer innovation has a clearly positive effect on employment, being land-augmenting and capable of generating large increases in labour demand. The more intensive cropping patterns mean that demand for labour is less drastically reduced in slack seasons, and can thus reduce the seasonal under-eemployment and poverty of the landless and the land-poor. 12. The effect of tractorisation is composite and variable. There remains a discrepancy in the existing literature betw.zeen the macro-level direct effect of tractors on employ-ment, w,hich is strongly negative--  estimated or projected on the assumption that tractors will be used for the multiplicity of operations of which they are technically capable--and the micro-level effert, which is more or less neutral, observed from the use of tractors during the 1960s. However, during the 1960s tractor use was much more limited (beinq mainly concentrated -on seedbed preparation), and hiring out was more common, than is likely to be the case now or in future, given the continued policy of encouraging tractorisation. As the number of tractors rises rapidly, hiring out declines and their owners begin to use them for other more labour- intensive tasks, their labour-displacing effect increases. 13. The new agricultural technology has affected both the amount and the type of labour required. Though the seed-water-fertilizer innovation can be equally successful on large or small farms, most innovative Larms are large, and the employment effects of the new technology have been Qreatest for the hired labour on which these farms depend. Tractorisation is apt to raise the demand for attached labourers or permanent farm servants, rather than casual labour; tractor farms need to be able to rely on a core of workers familiar with the schedule of operations and with the machines. In many of the "green revolution" areas of northwest India, farmers are making increasing use of per- manent labour contracts (often with new tying instruments replacing the traditional ones). We return .o this subject be low. 14. So far, during the initial phase of diffusion or technological change in agriculture, the combined effects or controlled irrigation, high yielding seed varieties and fertilizers in augmenting farm employment have outweighed the negative effect of mechanisation. The employmenL effects of technological change can obviously be inLluenced by policy intervention. Considerable research has been done on the issues of corrective policy concerning credit, agricultural taxation, land consolidation with a view to groundwat-er utilisation, and so forth, which influence the mode of organisation of production and the farmers' choice of technology.  -vi- Levels of Employment (See Section II, pp. 34-41) 15. Conclusions about the actual levels of employment of the growing agricultural population are difficult to draw for India as a whole, since agricultural growth has varied so much between regions. Moreover, comparisons of rates of employment and unemployment over time are dif icult with the available data, and are at best rough approximations. Some of the findings may be summarized as follows: - the rate of open unemployment (the ratio of days without work to days available for work) among both rural labour households (the fifth group in para 2 above) and poor small cultivator households 1/ is low in absolute terms, but this is partly because at any given time the desperately poor may not have a reservation wage below which they choose to remain unemployed. Better inqicators of welfare in their case would be their willingness to take up alterna- tive part time or full time employment, and their level of income. - the rate of open unemployment is relatively lower among poor small cultivator households than among rural labour households. However, there may be more disguised unemployment among the former. - for rural households in general, taking into account all the economic activities they are currently engaged in, underemployment per person per week (defined as less than 28 hours of work during the reference week and availability for more work) appears to have slightly declined over the past fifteen years, but not enough to raise the employ- ment of the average wage labourer significantly. 1/ Here defined as the poorest ten percent of households mainly dependent on self-employment, whether tenants or smallholders.  -vii - 16. A survey of rural households mainly dependent on wage employment in the West Godavari district of Andhra Pradesh showed that their availability for extra wage employment increased as their presen-t incomes rose. There was also an inverse relationship between landholding (which indicated self employment) and availability for extra wage employment. At the time of the survey,, members of the better- off labour households were spending less time on wage emplo-y- ment, but had a higher average wage rate, than members of poorer households. This suggests either that the labour market is stratified, perhaps on the basis of skills, spe- cialisation or labour tying arrangements, or that these workers' reservation wage rises as their incomes increase beyond the level of acute poverty. 17. Data collected in the 27th round (1972-73) of the National Sample Survey will provide a broader basis Lor judging whether willingness for alternative employment is due to too little employment, or insufficient wages for current employment, or both, and thus for designing reme- dial schemes and specifying the amount, duration, location and Liming of the additional employment needed. 18. S-mmarising available information on levels of employment, it can be said that during the initial phase of agricultural growth with technological change: - employment per acre has risen, though not enough to to raise the level of employment per agricul-ural labourer significantly even though increases in employment per acre have mainly affected hired labourers; - with high yielding varieties there are less dramatic seasonal fLluctuations in the demand for labour than with traditional varieties.  - viii - Emplover-Labcar Relations and Agricultural Wages (See Section II,pp.41-51,and Section TII,pp.64-66) 19. As was noted above, technological change, along with changes in land tenure, has been accompanied by an increase in the use of long-term labour contracts (2 years, and sometimes up to 5 years). In areas where the traditional and still dominant category of permanent labourers are paid in crop-shares, payments have changed.. While they used to receive a fixed fraction of gross output, they now receive a fixed share of farm income, from which other production costs have been deducted. The payment has thus become a variable share of gross output; since the more sophisticated the production technology is, the larger is the share in total production costs of elements other than labour, and hence the smaller the share of labour. Other new features of permanent labour contracts, observed in Haryana where permanent labourers often come from households with small rarms, incorporate carefully timed advance payments and managed indebtedness, whereby the employer "relendsIr part of the institutional credit he receives, or "resells" some of the modern inputs to which as a large farmer he has easier access. in an area such as Haryana where the labour supply is tight, the emoloyment of attached workers selected from small cultivating households, on a variable share basis with managed indebtedness built into the contract, assures large farmers of a supply of labour on al.;ourable terms. 20. Labour relations appear to be very different in Tamil Nadu, where there have been violent conrfrontations between organised labour and commercial paddy farmers over wages, Lollowed by government arbitration, and in Kerala. In the latter state, agriculture has grown only slowly but collective bargaining between agricultural unions- organized with support o- major political parties--and organized farmer-employers, is now a regular way of deter- mining wages, working hours, employment procedures, and regulation of tractor use, and of preventing employ,ment of migrant labour.  - lx - 21. As levels of productivity have diverged between states, so have money wage rates. Since the mid l960s real agricultural wage rates have fallen in a few regions, though they have increased in the fast growing highly productive areas of north and north-west India, and in Kerala (Table 5, page 46). 22. Changes in agricultural labourers' wage rates are poorly adjusted to and lag behind rises in their cost of living. Without the organised bargaining power of the industrial wor- ker or the protection of a minimum food ration, their real wages decline whzn there is a sharp rise in foodgrain prices. (Except in Kerala, there has been virtuallv no subsidised public distribution of foodgrains in rural areas.) Payments to workers in kind do not offer adequate protection when the harvest is poor, or when farmers have a strong incentlve to stock up, Lor selling later in the year. 23. The extent of change in agricultural labourers' iving standards since the mid-1960s has not yet been adequately measured. Even in the high growth regions it appears unlikely that increases in agricultural employment and wages were enough to raise the per capita income or level of living of the increasing numbers of agricultural labourers, or to distribute a fair share of production growth in their favour. Several studies have attempted to measure the share accruing to labour of the value of incremental crop output (net of material costs) resulting from the change to high vielding varieties. From scattered estimates it appears that, in a switch from traditional to high yielding varieties of wheat, kharil and rabi paddy, wage labourers usually received only 10%, or slightly less, of the value of incremental output, net of material costs. Considering that from 50 to 90 percent of the work on medium and large farms is done by wage labour, and that .wage labourers account for over a third of the agricultural workforce, this is not an impressive figure.  Labour Market imperfections (See Section III, PP.52-63) 24. The fact that money wage rates have risen or remained stable while most of the wage labour force is underemnloyed suggests that there are imperfections in the labour market. A number of possible explanations for this have been advanced. 25. According to Arthur Lewis's hypothesis, in a labour-surplus rural economy the real wage level is fixed institutionally, and has no impact on the amount of labour offered. However, it i s very difficult to visualise the exact institutional mechanism which could maintain a constant real wage rate except.in a closed system of feudal patron-client relationships or in a totally opposite situation where commercial producers face unionised labour. 26. The second explanation, related to the first, is that wages are based on nutritional needs. Through labor tying arrangements and time-wage adjustments, the few em- ployers who control the market set an equilibrium wage rate tlat at least maintains the required number of work units at the subsistence level. Assuming that better nutrition does increase efficiency, Lewis's constant real wage is thus explainied in terms of employers' economic rationality, which may have become forr.alised over time in institutions of labour tying and pay,nents in kind. 27. However, the subsistence hypothesis is not an adequate basis for a general theory of wage determination. For example, the wages of attached labourers are those most likely to be determined or the basis of nutritional require- ments. The employer's "investment" in their nutrition clearly pays him back. But the decision to engage them as tied labour may rest on the cropping pattern, and on the need to relv on a core of non-seasonal workers, as much as on minimising the cost of work units. Most of the farmers in the highly developed high wage regions of Punjab now use more attached labourers than ever before, while in many areas with low calorie income levels and a less intensive cropping pattern there are very few attached labourers. The subsistence hypothesis is even less applicable  - xi - as a basis for wage determination in the case of casual agricultural labourers, who still constitute the majority of the wage labour force in many regions. 28. The third hypothesis is that the labour market has a dual structure, with the supply price of family labour on small or peasant farms being lower than the market price of wage labour employed on large or capitalist farms. This difference in labour costs between small and large farms may account for the greaLer use of labour and higher production levels per acre observed on smaller farms. For various possible reasons, peasant households do not compete directly with landless labour for wage employment on other farms. They may prefer working on their own farm-- particularly the women members oL the household. Another reason may be that if the peak season absorbs almost all the family's labour supply, they are only available for work elsewhere in the lean season when wages are usually lower. Tiird, even if members of peasant households were available for part-time wage employment during the peak season, large fatters may prefer the full commitment of-fered by pure labour households Lor such activities as harvesting or transplanting whose timely completion is very important. Some peasant house- holds may be able to release one of their members to work lfull time as a wage labourer, by hiring in some part-time workers in the peak season. 29. The empirical work done so far in India has neither conclusively pro;ed nor disproved that labour is cheaper for small farms than large, but if the apparently greater use of labour, and greater output, per acre on small farms has been due to a labour cost advantage (rather than to differences in land quality or irrigation), the question arises as to wny larger farmers do not lease out land in small parcels so that the difference in labour use and productivity per acre disappears, rather than leasing out less land now than they did twenty years ago. T.wo reasons for this trend are imperfections in the land market and counteracting cost dil- ferentials for factors other than labour. As was noted abovle, existing imperfections in the land market were accentuated when the new tenancy legislation was pre-empted by evictions.  - xii - Moreover, large farms can more easily bear the costs of new inputs. Though traditionally smaller holdings were more intensively irrigated than large, the rapid increase in tubewvell irrigation, mainly on larger farms, has been reaucing this gap. Large farms' advantage in capital and access to institutional credit, coupled with the availability of machinerv, has far outweighed the possible labour cost differential of the earlier period. 30. Dualism as it has usually been hypothesised may also be declining because many erstwhile small cultivators have become labourers primarily dependent on the market for employment. However, other imperfections in the labour market may be on the increase. Understanding those arising from new labour tying arrangements may be vital for under- stFnding not only the response of wages and employment to technological changes but also the changing relationships between labour and employers in rural areas. More infor- mation on the types and provisions of labour contracts in various agrarian regions would obviously be very useful. 31. So far not much empirical work has been done in the important area of seasonal fluctuations in employment. How labourers use their peak season bargaining strength, how they migrate into and out of a rural area seasonally, and how employers react to labourers' peak season bargaining strength under different circumstances are basic to under- standing rural labour market behaviour. More disaggrega- tive data are required for the analysis of patterns of seasonality in employment and agricultural wages, seasonal shifts in the ccmposition of various types of employment zor rural households, and seasonal migration patterns. 32. The ideal data base for analysing labour demand and supply behavior in rural areas would be an integrated household schedule that gave both the labour time disposi- tion of its members and the use of family and hired labour in all household enterprises. This is because most rural households act as both employers and employees. How,7 far their members are involved in the labour market, to hire out and/or hire in labour, depends on labour allocation  - xiii - decisions taken at the household level. Data from the 27th round of the National Sample Survey, giving both farm manage- ment details and labour time disposition, should begin to make such analyses possible. 33. The present survey has identified two main priori- ties for further study, both with existing data and with newly collected more detailed data: 1) types of labour contract, and the ways in which contracts are influenced by conditions prevailing in the land, labour and credit markets; 2) the labour supply behavior of rural labour households and the labour demand-supply interactions in small cultivator households under different institutional and technological conditions.  Section I. Size and Comoosition of Rural Working Force The enormous body of Census data on working force--by rural and urban areas, by age and sex, by primary and secondary occupations, 7 y level of education, and by regions (down to the district level)--has provided the chief material for intertemporal as well as cross-sectional comparisons. The National Sample Survey data, for rural households in general and for rural labour households in particular, are used for answering more detailed questions regarding economic activity rates per person, particularly the flow rates of unemployment, and the level and the composition of gainful employment (on own account and on wage, on farm work and on non-farm work). Visaria (1975) has done a general survey of the concepts and definitions used in these two data sources, problems of comparison and of projection of the demographic and the economic characteristics of the population. The object here is to focus more on the conceptual issues and specific pieces of research that have direct bearing on questions regarding changes in size and structure of the economically active in rural areas, the mar- ket allocation of employment under sets of conditions varying regionally and over time, and how this is reflected in certain simple indicators. The Census-based indicators that are relevant here are: changes in rural working force participation rates, occupational distribution of rural workers, the modes of combination of different categories of economic activities in their cases, and the nature and degree of mobility or migration of labour to and from rural areas. I.a Recent Trends from the Census The main difficulties encountered in comparisons over time, cover- ing the three post-Independence Censuses, arise from (a) the basic problem oL defining a worker in the rural context of predominantly household-based economic activity, with use of family labour as the chief mode of employ- mernt, and of seasonal variations in work participation; and (b) related to the consequent search for a suitable criterion, the changes made from time to time in the definition used for describing the worlk er status of a person and the occupational category of a worker. The definitional changes afLect comparability and, unless properly adjusted, might indicate unreal shifts in economic characteristics of the population. The 1951 and the 1961 Census used the "usual status" definition of a worker. It gives an inventory of persons engaced in any economic activity on a more or less regular basis, irrespective of the time intensity of such work participation. The reference period for agricultural and allied rural occupations was broadly defined as "the greater part of the working season'" during which anyone gainfully employed for at least one hour a day was considered to be a worker. The relative importance of different kinds of employment for a wcrker was enumerated in terms of primary and secondary work. The 1951 Census also made a distinction between "tearners" or self- supporting wor'kers and "earning dependents," enumerating primary occupa- tion of the latter and both primary and secondary work of the former.  -2- The 1971 Census made a major departure by using a labour .time dis- position criterion over the reference period of one year (for those engaged in cultivation/livestock raising/household industries and plantations). It defined as workers only those who spent the major part of the reference year in economic activities. Thus, all irregular/marginal/part-time workers according to the time disposition criterion came to be enumerated as non- workers in their main activity, and whatever economic activity these "non- workers" might be doing was enume:ated under their secondary activity. Thus, a measure of total rural working force in 1971 comparable with the preceding Census would have to include the workers according to the main activity criterion and most of the non-workers with secondary work par- ticipation. Subsequently, the Census Commissioner's Office, in a report (1974) based on a sample resurvey designed for the purpose, gave estimates of two sets of adjustment factors for deriving comparable worker partici- pation rates. The sets of adjustment Lactors--one for adjusting 1961 figures in terms of the 1971 concept and the other for 1971 figures in terms of the 1961 concept--were derived from surveying the sample trice, using the respective definition of a worker, length of reference period and the order of canvassing the economic questions. It has been argued that in the rural context of self-employment and use of family labour in seascnal type of productive activities, use of the more liberal concept of a worker like that used in the 1962 Census might be more appropriate and capture the reality better. The argument is justifiable partic- ularly in the case of women and children who, except during the peak agricultural seasons when they are heavily and overtlv drawn into the Tworking force, very often dovetail their participati.on in 'household- based economic activities with domestic work. The borderline separating these is not always very clear either to the respondent or to the enumerator--simply because they are not compartmentalised as in the case of wholetime workers employed on wages and salaries. At the same time, for many purposes it might be useful to separate the hard core of regular, "full time" (relatively speaking, though not necessarily by an absolute time norm) workers from the rest even in an "informal" sector of household-based productive activities. Whichever concept and set of adjustment factors is used, a com- parison between 1961 and 1971 indicates that there was a genuine decline in worker participation rates--a decline that has been more drastic in the rural than in the urban area, and among rural women in particular. The composition of this decline, the underlying reasons and the implica- tions are of obvious interest in understanding the dynamics of rural work force structure in the technological and the institutional contexts of development. Between 1961 and 1971 population had grown at a compound annual rate of 2.23%, as against 1.98% betw.een 1951 and 1961. In 1971, 8bO% of the population ,-as rural and 51"' of the rural pooulation belonged to the prime working age groups. The proportions were not veryi different in  -3 1961: 82%' of the people were rural, of wh'nich about 47% weere in these age groups. 3ut the worker participation rate after adjusting for definitional changes was much lower in 1971 than in 1961. For rural India as a whole, comparable estimates of worker participation rates after adjustment on each of the alternative bases are shown in Table 1. Table 1: RUlRAL WORKER PARTICIPATION RATES (%), 1961-1971 Adjusted Census Census Adjusted 1961 1971 1961 1971 >Iale 57.0 53.5 58.2 54.7 Female 25.7 13.1 31.4 18.9 All Persons 41.6 33.8 45.1 37.3 With the stricter as well as with the broader definition of a rural person's "usual" work status, the worker ratio seems to have declined considerably over the decade, and the decline seems to be much sharper for women than for aen. The distribution of this large decline in rural worker ratio was rather uneven--regionally and occupationally. The decline was relativzely less in Kerala, West Bengal, Punjab and U.P.; and it was relatively large in the more arid western states, the north-central N4adhya Pradesh and Harvana, and the eastern states of Bihar, Orissa and Assam. It is quite significant that the Lirst group of states even initially (in 1961) had a much smaller proportion of their rural women as workers. Because of the smaller weight of women in the rural work force, the further decline over time in female worker ratio in their case had less effect on the overall ratio. In the second group of states, particularly those (Orissa, Assam, MLIadhya Pradesh, Rajasthan and 'aharashtra) rwhich continued to re- main agriculturally back:xward during the decade, the decline even in male worker ratio has been somewhat more than in the agriculturally develoned and growing states (Punjb, 'Da Tamil Nadu, U.P. and Haryana). The state of Kerala, despite very little agricultural growth over the period and a lot of urban unemployment, managed to retain a stable male rural worker ratio (though at a lower absolute level than in other southern states) - a phe- nomenon largelY attributable partly to the relative success of family planning in rural Kerala and partly to migration from its rural areas. The change in occupational distribution of rural workers (on either delinition) appears to be v!ery interesting in the context of the decline in overall workrzer ratio. n.:hile there was a large decline in the proportion of self-employed and unpaid family vorkers (both male and female, but particularly female) engaged in cultivatLon and other house- hold enterprises, the proportion, of those -who worked primarily as hired  -4- agricultural labourers went up quite considerably at the same time (Table 2). The Census Resurvey Report does not separate out the case of rion-agricultural rural labourers; they are combined with artisans and household industry workers. But in all probability, their proportion went up too, while that of the self-employed non-agricultural workers declined more than shows up in the combined figures. Thus, the propor- tion of hired (wage-employed) workers among all rural workers went up from around one-fifth to around one-third. The decline in worker par- ticipation rate was concentrated among various categories of the rural self-employed, those taking part in household-based economic activities. Table 2: PERCENTAGE DISTRIBUTION OF RURAL WORKERS BY CATEGORIES OF PRINCIPAL OCCUPATION, 1962 and 1971 Adjusted Census Census Adjusted Sex/Main Work 1961 1971 1961 1971 Rural Cultivators 62 56 61 56 Male Agricultural Labourers 15 25 16 25 Wo rke r s Other Workers 23 19 23 19 Rural Cultivators 60 33 59 39 Female Agricultural Labourers 23 54 25 48 Workers Other workers 17 13 16 13 All Cultivators 61 51 60 52 Rural Agricultural Labourers 18 31 19 31 Workers Other workers 21 18 21 17 A lot of attention has been attracted by these rather striking changes in the participation rates and in the occupational structure of rural workers over a decade that experienced faster agricultural growth with technological changes as w7ell as higher rate of population growth compared to the earlier ones. A demographic explanation for the decline in the overall par- ticipation rate deri.es from an increase in the proportion of children over the decade. Visaria's (1973) tabulation of age-specific worker rates for five states on the basis of the 1962 Census indicates an inverted U-shape for rural areas with a flat plateau for ages from 20 to 55 %,ears. Though children and teenagers in rural areas are draw.n into productive employment to a much greater extent than in urban areas, and possibly even more so in poor than in better-off households, the wcrkXer rate in  their case still lies below that for adults. A rise in the proportion of children, together with some rise in incomes and in school enrolment that took place even in the rural areas over the decade, might constitute one of the reasons behind the decline in the worker rate for all ages combined. But this can hardly be a major reason. An important explanation may also arise from linkages between changes in occupational distribution of rural workers and those in the over- all rural working force participation rates. The cross-tabulations neces- sary for quantifying these linkages and to identify the correlates of rural occupational structure are not available. However, the question has cropped up in several studies in various contexts, and certain aspects oL changing agrarian structure that have been observed in other connections are relevant to understanding the shifts in the proportion of rural wage labourers rela- tive to the self-employed or own-account workers. L.b. Correlates of Changes in the Proportion of Agricultural Laborers. One plausible explanation for the sharp rise in the proportion of agricultural laborers along with the decline in overall work participation during the sLxties, w;hich `Followed their relative stability over the fifties, could arise from a change in the distribution of cultivated area. It has been argued that, as a consequence partly of the tenurial legislations and partly of the technological changes taking place in agriculture, small peas- ants have been increasingly pushed out of cultivation or self-employment into wage labor. The introduction of irrigation-seed-fertilizer innovation, thnough theoretically scale-neutral, has been very unevenly distributed on account of the prevailing imperfections in credit (capital) markets. The official policies of low taxation of agricultural incomes and assets and subsidisation of modernising inputs, combined with the large farmer's ad- vantage in self-financing investment and in getting credit, have encouiraged big landowners to take a more direct interest in farm management. At the same time, the tenurial legislation discouraged their leasing out. From all accounts, there took place considerable eviction of small tenants, and more large holdings were now operated with hired labor. In some parts of the country, rising land prices migcht also have led some small owners to sell out. But, on the whole there is some indication of a decline in the proportion of the non-landowning rural households, and little indication of net sale of land by small ow,n,ers. But nationwide aRgregative data in this respect average out so many cross-currents that not much can be said on its basis. In some parts of the country, there has been a sfgnificant decline in the proportion of rural households not owning any land. This has been so at least in the western region (Dantwala and Shah, 1971) and in Punjab (Bardhan, 1976). Irrespective of whether the proportion of rural households owning little or no land is growing or declining or remaining unchanged, a more significannt fact is that They are now finding it harder than elver before to lease in land for cultivation.  -6- Available data on tenancy do not give a clear-cut average picture because of considerable regional variations in the nature of tenancy, because registration and reporting of tenancy has been, affected by the land reform laws, anid above all because ofL very significant recent changes in the com- position of tenants. Joshi and Dharm Narain (1969), on the basis of NSS data relating to 1953-56 and 1960-61 and a few village surveys, argued tnat the incidence of agricultural tenancy declined over the fifties and that the decline was more for the larger tenants with long-term contracts than for smaller ones. Though the two rounds of NSS data, on which this argument is based, are not comparable on account of a definitional change (in 1960-61, land leased in on long-term contract was classified as owned) and though several other village surveys indicate an opposite tendency (Bardhan, 1970), it is more or less agreed that there had been some amount of genuine '- cline in the proportion of area under tenancy over the fifties. But subse- quently, during the sixties and thereafter, there seems to have been taking place along with "green revolution" a very significant change in the pattern of leasing in and out. Comparing between 1953-54 and 1970-71 (8th and 26th rounds of NSS), Bardhan (1976) observes that while the leased in proportion of land declined in some states, it increased in others. However, in the two major "green revolution" states of Punjab and Haryana, while the proportion oz rural households not owning any land went down from 40% to 9%, the propor- tion not operating any land went up from 30% to 54% over the period. In contrast with the usual situation in other parts of the country and in the earlier year in this part, the Lorenz curve for distribution of operated area in 1970-1971 was found lying below that for owned area. In other words, the direction of leasing in and out seems to have been reversed, at least in some of the growing areas. The small landowners aire leasing out (obviously to larger ones)-- either because of their disadvantage in terms of resources for investing in irrigation and purchasing the fertilisers and the new seeds (the push factor) or because of the rising wage rates (the pull factor) or both. Bandyopadhayaya (1975) reported a similar phenomenon from recent vil- lage-level data for parts of West Bengal, where large farmers with tubewells lease in land from neighboring small owners for the early summer season in order to raise a high-yielding variety of paddy. In the relatively better- irrigated Birbhum district of West Bengal, Rudra and Newaj (1975) observed sharecropping tenancy declining with agricultural development and owner-cul- tivation with hired labor rising. Thus, it seems that technological change with use of controlled irrigation (involving investment in tubewells which, like tractors, offer a scale advantage) and cash purchased inputs, because of its bias in favor of large farmers and against small tenant operators in particular, has led to a major decline in the proportion of the self- employed and rise in that of the wage-employed in the rural working force. The change chus indicates a polarisation within the agricultural sector, with a faster-growing subsector of medium and large farms generating much of the increasing supply ofL age employment that the previously self-emoloyed land-poor households are now taking up along with landless labour households.  -7- The second explanation for the rising proportion of agricultural labourers is that even if in some or most areas landowning peasants still manage to cultivate the owned bit of land, their incomes from such self- employment are getting increasingly inadequate for supporting their families, as a result of which more of their workers are now taking up wage employment. Collating from various small-scale studies relating to Gujarat, Vyas (1976) partially substantiates this hypothesis. Most of the entrants in the smallest size-group of farms in this area came from below--as a result of relatively more effective implementation of tenancy legislation here--rather than from above through sale or loss of mortgage. The number of small farms (abso- lutely and as a proportion of all farms) had increased in Gujarat over the last few decades and their average farm area remained stable. But, as the size of these farm households grew over time, the uzchanued productivity of their small. holdings (because of a lack of public investment in irrigation in the area and the low and uncertain groundwater level that makes any pri- vate investment harder even for large farms) led to increasing resort to hir- ing out and outmigration of workers. Even for this case, one might argue that if these households could lease in more easily, they could have more self-employ-ent. In the case of some other states like Punjab, Haryana and West U.P., which in contrast with western India have experienced much faster agricultural growth centered around controlled irrigation, and also a more prominent emer- gence of "capitalist"t farmers (however defined in terms of farm size, hired proportion of labour, investment in modern inputs and technology, use of farm machinery, or the rate and mode of reinvestment or farm income), one miaht reasonably argue that a sharp rise in the proportion of agricultural labourers was partly the result of an increased growth in demand for hired farm labour. Any prosperity-induced leisure preference in their case could further rein- force the direct effect of production growth with technological changes on demand for hired labour. The hypothesis of leisure preference among better-off farm house- holds in the context of production growth as an explanation for structural shifts in the composition of rural workers has not been tested adequately yet. From the 9th round (1955) of NSS data for rural households, Visaria (1970) noted an inverse relationship between the rate of female labour force participation and the level of per capita consmption expenditure level (as a proxy for income) of the household. In the case of rural males, however, the relationship was c, clearly positive one. However, the 9th round data arm not nearly as good as those of recent rounds, and, besides, the stock of gainfully employed people such as from this source is much less suitable for testing the leisure preference hypothesis than the flow of employment per person per unit of time from the more recent rounds. From a recent tabulation of grouopd data from twzo subrounds of the 27th round (1972-73) issued by the 'ISS Organization, RaJ Krishna (1976) observes a positive relation between employment and the per capita monthly expenditure group to w-hich a household belongs. If, instea-d of covering the entire range of per capita e:xpernditure  -8- from the poorest to the richest households, one separates out those below Ila poverty line," then perhaps the. income elasticity of leisure preference would come out to be more clearly positive for the rest. Bardhan (1975), after separating out from the 25th round a subset of labour and peasant house- holds--in Punjab, Haryana and U.P.--with their consumption level below min- imum subsistence norm, correlated per capita expenditure of a household with the flow rate of adult employment. He found that there was no significant correlation between the two, because in the case of the very poor, the in- come elasticity of demand for leisure is counteracted by the need for better nutrition to enhance capacity for work. It seems that separate cross-tabulation for the male and the female members of a household, and for the poor and the better-off sets of households is likely to yield a more definitive result by separating the nutrition/work- capacity effect of increases in level of living at the lowest end from its leisure preference effect at the middle and upper end on the degree of work participation in rural areas. The fourth hypothesis, connected with the first one concerning changes in tenancy, for explaining the recent shift in distribution of rural workers from cultivators to labourers is that it could be partly a lack of identification. Apart from small tenants being pressurised by their land- lords into reporting themselves as hired hands, the borderline between a sharecropper and an attached labourer is being rendered thinner as a result of higher turnover of tenants on oral contract along with the landlord taking over more and more farm management decisions. Another plausible hypothesis to explain the shift is linked with the changing age-structure of a growing population. Visaria noted from the 1961 Census that while the proportion of cultivators among male rural workers went up with age, the corresponding proportion of agricultural labourers went down. The proportion of such labourers got steadily higher among young work- ers down to the age-group 10-14 years. If this pattern has any general valldity arising from any stable socio-economic reasons,* then any shift in age-distribution over time in favor of the very young workers tAould have a corresponding effect on the distribution of workers becween these two categories. In the case of rural female workers, however, Visaria did not * Younger male members of rural households with small farms are more often found seeking outside employment on wages while the older men concentrate on the family farm or craft. The strain of wage employ- ment (which may involve daily commuting on foot across villages and is often more muscle power intensive), the lower prestige that goes with it which is less acceptable to the patriarch than to his nephew, the land received as or bought out of dowry that enables the older man to move from wage-labourer to self-cultivation--all of These can be more or less important factors in various cases.  -9- find much of a systematic relation between the age and the proportions en- gaged primarilv in cultivation and agricultural-labour respectively, which he attributes more to misreporting of age in the case of rural women than rural men. And yet in their case there has been a considerable rise (even larger than for male workers) over the decade in the proportion of agricultural labourers and a decline in that of cultivators. While some oL these explanatory hypotheses may hold better than the others in different parts of the country, it would be quite difficult, and perhaps meaningless too in view of the tremendous heterogeneity, to try to assess their relative importance or contribution to the overall shift in the occupational structure of rural workers. Even the initial situation differed from one region to another. The proportion of wage labourers among male agricultural workers was much higher in most of the South Indian states than in the northern states of Punjab, Haryana, U.P. and Rajasthan. These regional differences arlse not only from purely economic factors but also from institutional/historical differences (like, landownership pattern or mode of labour contract) as well as from interaction between these two kinds of factors over time. The dynamics of changing rural labor force structure and their socio-economic correlates are understood better within relatively homogeneous agro-institutional regions. Within a more or less homngeneous region, it Twould be interesting-- even for understanding the macro-level changes--to identify at the household level the allocation of labour supply across kinds of employment, and at the farm level the allocation of labor demand as between family labor and attached and casual hired labour. If this analysis is done for a series of situations over time or across regions characterized by different institutional setups, then one might see better the distinct effects of the economic and the insti- tutional factors as also the effects of institutional changes in the context of shifts in agricultural technologyv This type of studies is discussed in Section IV. I.c. Female Work Particioation in Rural Areas Although work participation rate among rural women is much lower than among rural men, a larger proportion of female wo,rkers is engaged in agriculture and within this group agricultural labourers constitute a larger proportion as compared with the case of male workers. Over the sixties, this pattern has been accentuated. Female work participation has fallen further, the proportion of them working as agricultural labourers has increased and the proportion of cultivators declined more than in the case of the male workr;ers in rural areas. This could possibly be interpreted in at least two ways, of which the strength of one relative to the other cannot be estab- lished adequately from the alvailable macro-level information. Both are valid in varying degrees. One is that this shift in distribution o' rural female -workers as indicated by the Census is a correct description of reality, and is  - 10 - consistent with thse available evidence indicating a rise in rural. income disparity during the sixties. The argument usually runs in the following manner. Assuming that on an average the poorest rural bouseholds without land have more of their women working as wage labourers, that at the next level of small cultivators female labor more or less withdraws from the market and shifts to own-account work on the farm even as some of the male members might still keep on working as wage labourers, and finally at the level of prosperous farmers emplo,ment of female members of the family even on their own farm declines whether due to leisure preference or status considerations, The other plausible interpretation is that self-employment in household enterprise tends to be under-reported in the case of women workers in general and more so among those of cultivating households. Their employ- ment reporting is more subject to recall lapse, oversight and under-recogni- tion than wage employment is. Some of the productive work that the women of farm Lamilies do (like winnowing, handpounding or parboilding of paddy, preparation of dairy products, storage of seeds and produce, and marketing of some of the produce) is sometimes not clearly differentiated from purely domestic work. This sort of under-reporting of female self-employment in rural areas would affect not only the size and distribution of Tworkers in the Census and the NSS, but also the NSS-based estimates of the flow rate of their employment and its distribution as between own-account and wage- employment. Some of the findings of Bardhan's (1975) study of adult female labour time disposition in small cultivating households from the 25th round point in this direction. The elusive nature of female work participation (from both Census and NSS data) brings to us with special force the enormous complexities that are involved in collecting the kind of data and doing the kind of anal- ysis that can capture the reality of work and employment in the "informal" sector, which is characterized by household-level decisions about combining and arranging various economic activities including production for own con- sumption. I.d. Principal and Subsidiarv Work Combination The Census data on principal and subsidiary Twork, and the NSS labour time disposition data, indicate that for any rural worker there is some mixture of the for mal categories of work,- agricultural/non-agricultural, wage-employ- ment/own-account employment. Understanding the operation of rural labour mar- ket essentially involves identifying the correlates of the variations in the pattern or composition of this mixture. The Census categorizes the orimary occupation of a rural rworker in terms of -.T'hat accounts for the major part of this working time.* Then it There is no basic difference in this p,rocedure 'between the 1961 and the 1971 Census. In the 1971 Census, a person not working over the major part of his total time (not the wvorking time) is effectively defined as a "non-worker" by prim,ary occupatior. TUhat he does for the major part of his work time is then regarded as his "secondary" Occupation. For the "worker," the procedure is the same as in 1961.  also enumerates his secondary occupational category in terms of what takes the next largest share of working time. Someonte working on his farm during days and taking up wage-employment in others' farms or a public works site is categorized as prioarily cultivator or agricultural labourer, depending on their relative shares in his total employment. By this criterion, in 1961, 11% of the 100 million rural workers with own-account cultivation as main occupation had agricultural labour as a secondary occupation; and 13% of the 32 million rural workers with agricultural labour as main occupation had cultivation as secondary work. About a third of those who were (rural) non-agricultural workers by primary occupation had secondary employment as cultivators or as agricultural labourers. And undoubtedly a good many of those wno were enumerated as non-workers even by the liberal definition of a worker used in the 1961 Census had some amount of "secondary" own-account employment in agriculture or other household-based productive enterprise. Coming to variations across regions, in 1961, secondary agricultural labour- ers amounted to more than half as many as the primary agricultural labourers in Orissa, Bihar, Tamilnadu and M.P., while in Punjab, Gujarat and Kerala they were of marginal importance. For the 1971 Census, tabulations on secondary work participation are not yet available. However, one can reasonably e:xpect that its more restricted definition of a worker and the long reference period would result in showing not only a very large proportion of rural non-workers as having some secondary employment in agriculture, but probably also a larger propor- tion of the agricultural workers as having secondary employment outside of agriculture on account of the seasonal shifts of rural labour from non- agricultural work to agricultural operations during the peak growing seasons and back again during the lean months. The enmeshed character of primary and secondary participation of different grours of rural workers in various sectoral or occupational cate- gories of work raises certain important questions regarding the operation of the rural labour marklet and the dynamics of the rural working force. How does the composition of employment of a rural worlker change seasonally under different conditions and how are these related to changes in the structure of rural wage rates? How does the combination of types of work vary across regions For a comparable set of rural households, and how is it related to other regional differences? rThat determines the e:tent of secondary inv.,olve- ment of self-employed rural workers in wage-employment? The available Census tabulations do not permit adequate 'handling of these issues, and these are not coordinated with data from other sources relating to the other socio- economic factors, for many of which the unit is the household and not the individual as in Census. Besides, while the "usual stat'us" definition used in the Census captures reasonably well the reality of seasonal work partici- pation in rural areas, it does not tell us anything about the quant=m of em- ployment and its distribution among different types of work. For these two  - 12 - reasons, one has to turn to the intensive household-level (NSS) and farm level (FMS) surveys, if one wants to find out about the economic character- istics of any particular group of rural households in relation to the in- ten5ity and pattern of employ,ment for their members. I.e. Changes in Rural Working Force from National Sample Survey NISS conducted labour force surveys with reference to individuals from the 9th round (1955) up to the 17th round (1961-62). In the subsequent rounds up to the 21st (1966-67), labour force data were based on the integrated household schedule canvassed intensively among a relatively small sample of households. For the earlier series of rounds, a person was considered to be a worker if he was gainfully employed, for however nominal a time, on at least one day during the reference week preceding the date oL en=eration. If he was not gainfully employed even on a single day but was seeking and/or available for work then he was considered unemployed. The workers and the unemployed thus defined constituted the labour force. In the subsequent rounds up to the 21st, though the weekly "usual status" definition of a worker was continued, the time intensity of his/her gainful employment ring the week was the criterion for determining the usual status. Following the recommendation of the Committee of Experts on Unemployment Estimates (1970), the practice of collecting data regarding the numbers of "usual status" workers and unemployed was abandoned. And from the 25th round on- wards data were collected regarding labour time disposition during each day of the reference week for each member of the household, which yield the llow rate of employment (and unemployment, etc.) for each individual and not just for the "usual status" gainfully employed as in the 21st and earlier rounds. In the 27th round (1972-73) along with these, the annual "usual status" of each person was also recorded to give an annual stock rate of worker participation similar but not entirely comparable to that from the Census. Visaria's (1970) comparison of rural labour force participation rates from the 1961 Census and from the 16th round of the LNSS indicates a reasonably close correspondence in the male case; but the female partic- ipation rates from the NSS are considerably lower than from the Census, particularly among younger women below 24 years. This is due to the dif- ference in reference period to which enumeration of women's work (or labor) force participation is particularly sensitive. The NSS uses a shorter reference period of a week, which is intended to minimize recall lapse in enumeration of time disposition. The prevalence of seasonally unstable par- ticipation being more among rural woumen, the shorter reference period would result in some of them not being classified as work'ers with the definition of weekly "usual" status. But manv of them might still be classified as workers by the Census definition of "usual status" over a longer period ofx reference. The 27th round (1972-73) worker participation rates based on "usual" status during the ,ear mizht be definitionally more comparable with the 1971 Census figures. From available tabulations (1975) for two subrounds of the 27th round, the usLial status rural worker participation rates appear to be  - 13 - much higher than those from the 1971 Census. these are 64% for male, 37% for female and 51% for all persons, as against 64%, 13% and 34% respectively from 1971 Census (unadjusted). It would be very useLul if the exact reasons behind this diLference, which is quite large even for male participation, were investigated. As for the proportion of agricultural labourers among all rural wcrkers (an indicator that is more relevant in the present context than the overall worker participation) , the differences betwieen the two sub- rounds of 27th round and 1971 Census are considerable. In 1972-73 from NSS the percentage of agricultural labourers was 34, 31, and 33 for male, female and all rural workers. From 1971 Census, these percentages were 25, 54 and 31 respectively. In a situation -where much of productive emplovment is organized within the household as a unit and takes the form of self-employment and use of family labour and of per-maner.t or semi-permanent farm servants, and wihere all such employment is organized according to a seasonally changing basket of various kinds of economic activities, simply categorizing persons into workers and non-workers or the workers into cultivators and labourers on the basis of the share of work time or on the basis of an absolute cut-off point is not enouah. The broad orders oL magnitudes of chang2s can be meaningful and useful only when we know enough about the whole basket of time disposii- tion--and the nature of its distribution in terms of seasons and in terms of the economic and demographic characteristics of the 'nousehold. Many house- hold workers in rural areas (women, children and the aged in particular and sometimes adult males as well) usually participate in gainful employment to extents vary,ting greatly from one part of the year to another, one type of household to another, and one region to another. .Categorizing them in terms of any cut-off point (whether it is the 1961 Census one "at least one hour of work a day during the greater part of wor'king season," the earlier NSS one of wor'k--irrespective of hours--on at least one day of the reference -week, or the 1971 Census one of work taking up the major part of one's time during the year) results in leaving out a part of work participation and clubbing together all diversities in what is included. There is obviously a great deal of heterogeneity in terms of the time intensity and composition of employment among those above the cut-off point, which have to be analyzed with reference to other variables. Besides, instead of categorizing persons in a shifting and mixed situation it would be more correct to consider the flow rate and time structure of work or employment for everybody in the house- hold. From this point of view, NSS data from 25th round onwards which record the time disposition for each member of the household might prove to be more useful. I.f. .Migration Pattern of Rural Labour How far is the supply of rural labour responsive to spatial or regional differences in work opportunities? This covers a la-r2e number of more specific questions. To what e:tent does labour migrate from one rural area to other rural and urban areas? How much of it is lon-term or semi-permanent and how much short-term or seasonal? VThat are the occupational directions of movement of the out-migrants: i.e., what is the occupational distribution of the migrants before migration from rural areas, and what is it afterwards at the  - 14 - place of enumeration? Are the rural-rural or the rural-urban migrants mainly from agricultural labourers, from cultivators,.or from the non-agricultural rural households? And, in relation to the pre-migration situation, what is their employment status after migration? What is the net inflow or outflow of workers for different rural regions? How have the regional disparities in recent development affected the pattern and directions oL migration? These are some of the most elementary questions of migration characteristics that re- searchers concened with the elasticity of rural labour supply have to contend with. The sources of data at the macro-level are mainly the 1961 Census migration tables and the 13th round of National Sample Survey. Much of the careful analyses of these large-scale data sets have been more useful indi- rectly by revealing their inadequacy for answerinag most of these questions. The 1961 Census considers a. person enumerated in a place different from their place of birth (or origin) to be a migrant. When he migrated, how many times he did so and for what durations are not recorded. Apart from the spatial directions of such "life-time" migration, the Census also gives the distribu- tion of migrants by their present worker status and the (primary) occupational categorv. The Census was not concerned with the reasons of migration or the socio-economic characteristics of the migrants before migration, nor with the duration or the frequency of one's migration for worki as distinct from other reasons. However, some of the findings are still of some interest in the present context. These are reviewed below. Internal migration of women, which constitutes the bigger component of intrastate migration in particular, is largely accounted for by marriages across v-ilages and districts, and though many of them are workers at the place of enumeration their migration reveals nothing about the pulls and pushes of wrork availability. Of the rural male population in 1961, about 14% were migrants, mostly (13%) from other rural areas and a small fraction from urban areas. Of the male rural-rural migrants, 71% originated Lrom other rural areas within the district, 20% from other districts within the state and 9% from other states. Of the urban male population, 40% were migrants: 26.5% from rural areas and 13.5% from other urban areas. Of the male rural- urban migrants, 35% came from rural areas within the district, 34% from rural areas of other districts within the state and 31% from other states. As the urban areas accounted for about a fifth of total male population, it is easy to see that the bulk of their migration was rrom rural to rural (nearly twice as much as that from rural to urban) . Secondly, the bulk of rural-rural migration was intra-district, and at nearly 8% of total male population in the country this constituted the single largest component of migration. The Census data also indicate certain other demographic features of the migrants. One is that the male-female se: ratio for mi-rants of rural origin is much higher (nearly double) for movements acro4,s states than across districts within a state, and the interdistrict one was similarly higher than that for intradiscrict :migration. In other words, migration over longer distances arid from villages to cities features cmore o. male individuals and less of entire families. The importance of internal mi2ration and an adjustment factor for labour supply in the rural econemy are indicated, a-mong oc,ner things, by its quantitative importance, by its directions of mo%vement and by its occupational  - 15 - structure before and after migration. Considering the two available sources of macro-level data on migration, the formats of their tabulations, and the major findings of the research done so far on this basis, the questions may ,e posed and answered in the following manner. First, about the wor'ker participation rates among those who migrated into rural areas, and about the proportionate importance of the outmigrant and the inmigrant workers in the composition of rural workers. According to the 1961 Census, the wor'ker participation rate is much higher for the migrants from rural areas into other rural areas or into urban areas compared with TAnat it is for the non-migrant residents of the respective area--in the case of men as well as women, though obviously more so in the case of men. There could be a variety of reasons for this: the migrant from rural areas to urban or to other rural areas might be poorer than the non-migrants residents in the place of enumeration; they might contain more adults of working age; or they might simply be on an average tougher or more enterprising. The NSS 18th round data do show that the share of young adults was higher among the inmigrants in rural areas (who are almost entirely from other rural areas) and still more so among the inmigrants in urban areas (more than half from rural areas) than among the non-migrants in either. About their being also generally poorer, there are no data at the macro level. From the rural sample of the 18th round of NSS conducted in 1963-64, the overall immigration rate in rural areas was 92 per 10,000 of rural popu- lation. From the urban sample, the immigration rate was 2400 per 10,000 of urban population out of which about 60% came from,rural areas. Using the Census rural-urban population ratio, the outmigrants from rural areas moving to urban areas thus works out for that year to not much more than 3.3/% of the remaining rural population including the inmigrants. There is a very important reason whv the NSS figures indicate so much smaller incidence (pro- portionately spealking) of both inmigration and outmigration for the rural population compared wfith what the Census indicates. The NSS used a specific, shorter reference period. A member of a sample household was considered to be an inmigrant if the usual place of residence one year beLore the date of enumeration had been different from the village/town of residence at the time of enumeration. In contrast, the Census regarded the inmigrant as one whose place of birth (or origin) .Ks different from the place of enumeration. The NSS data thus gives an idea of recent migration (within one year preceding enumeration) , which is naturally much less than a "lifetime migration." But even if as a proportion of rural population, the extent of recent migration from rural to rural and urban areas is rather small, their qu:antitative import- ance is larger among the working population, more so among workers belonging to poorer sections or oarticular occupational grouDs like the acri'cultural labourers. The Census tabulations anal', zed so Far indicate the following pattern according to whether the rural migrant workers camle from within, the district, from rural areas of other districts within the state or froo other staLes. The proportion of agricultural labourers ar:r. inimiarant workers of the First two cate.orie,s was Substantiall1 lar-er (21%) than what it was arnn7 the non-mi,rant ru-ral workers (15% ), though in the case of inmizrant workers from other states it. was slDh;tly csmaller. T-'e proportionate importance of cultivators, on the other nand, ,as smaller among inmigrant workers than what it was among non-mni,rant workers (TabLe 3 upper half). It is .worth noting  - 16 - that the proportion of migrants wno seem to have succeeded in acquiring or leasing in land for cultivation is considerable, even though it is stilL smaller than in the non-migrant case. Tabulations from NSS "recent migration" data also indicate that the composition of the working rural inmigrants was more heavily weighted 'for agricultural labourers, as in the Census data, and for craftsmen and production process workers, than was the composition of the total working rural popula- tion. Correspondingly, the share of cultivators (includina farm managers) was much smaller among working migrants in rural areas than among the rest oLf rural workers (Table 3 lower half). Yet, as in the Census data, even this smaller proportion of cultivators was large enough to match in importance the agricultural labourers among the rural migrant workers. It seems reasonable to say that the inmigrant workers in rural areas, though consisting more (i.e., more than "local" workers) of floating proletarian elements like agricultural labourers, construction workers, also contain contain a good many relatively affluent landowners, cultivators and artisans apart from servicemen. Table 3: PERCEN'TAGE DISTRIBUTION OF I.NMIGRAMIT ALND NION-MIGRANXT RESIDE'NT WORKERS IN; RURAL AREAS BY MAIN OCCUPATION Cultivator/ Household Agric. rentier/farmn Industry Construction Labourer manager worker Worker Others 1961 Census, Male workers Non-migrant 15 66 6 1 12 i-rant (i) Intradistrict 21 45 7 1.5 25.5 (ii) Interdistrict 19 36 7 4 34 (iii) Interstate 12 24 5 7 52 NSS (18th Rou-d) 1963-64, All Workers Agric. Cultivators/ Production process Others lab farm managers work:ers & craftsmen All work.ers 22.5 62 7 8.5 (incl. mig rants) Immi.rant 25 29 15 31 Wdorkers It is worth noting that among the "recent" female mizrant workers in rural areas, the proportion of agricultural labourers was .en larger than the corresponding proportion in the case of maale mio2 rant work-ers. f'n a closer look, it appears that the proportiLon or _agricultural labourers 'V5s particularly higlh among tLhe women mi-ratinq from other rural are1as twir.h-i the same district. Much of the "short-distance" rural-to-rural migration  - 17 - of women workers thus consists of those in search of agricultural wage em- ployment. The phenomenon of tribal and scheduled caste women in paddy areas migrating in the busy season to take up agricuitural work like transplantation in prosperous villages seems to be not an isolated one, but rather of some general validity. NSS provides some cross-tabulations of the previous with the present activities of the inmigrants in rural areas, and their distributions by present activity and by previous activity for each of the various "reasons for migration." From these, several interesting conclusions emerge. First, of all (male and female) the working migrants employed as agricultural labourers at the time of survey, 50%. had it as their main occupa- tion even prior to migration, 9% were cultivators/farm managers, etc., and 14% craftsmen or production process workers. As much as half of all the migrant agricultural labourers were thus previously engaged in economic activities other than agricultural labour, or were unemployed or out of the labour force. Another interesting finding is that among the working migrants in rural areas most were workers even prior to migration: 85% in the case of such men and 76% in the case of w~,omen. Taking all the migrants in rural areas including those working and those not workinng at the time of enumera- tion, 48/%. were working previously in the male case and 251% in the female case. Thirdly, of those who had irmigrated explicitly in search of em- ployment for the first time, the percentage that had found work by the time of enumeration was 62 in the male case and 86 in the female case. Of those wTno came in search of employment not for the first time, more than 80% in both cases were known to have had prior emploment and were also found to be employed at the time of enumeration. Of those who came in order to start their own enterprise, the proportion that was employed prior to migration was nearly of the same order as that employed at the time of enumeration. Thus, the con- clusion is that the major reason for migration in rural areas has not so much been overt unemployment as the search for better employment. And, among the recent rural inmigrants, those who had already found work included not only those who were working earlier but a substantial proportion of those not working earlier and/or lookintg for work for the first time. All this about rural-to-rural m;,-. , which is by far the largest component. rhat about rural-to-urban mig::b.> Comparison between the dis- tributions of these two types of migrants in Let-ms of their previous activity reveals an interesting feature. The proportion of migrants, both male and female who were employ-ed or working even prior to migration was much larger -or rural-to-rural mi-ration than for rural-to-urban micratior. This could be interpreted as an indirect evidence in aFavor of the hypothesis that the second ty;pe of migration compared with the first is undertaken more by young adults seeking worK for the first time or bv better-off household s in winich the economic pressure for wl/;Iomen and children to work is somew.,hat less. But as the available census and -';SS data do not indicate the socio-economic char- acteristics of .householus that the lg-rants belcng to prior to migration, wYe do not have an. definit-ie g-neral ccnclus-ins about the "class" char3cter of rural-to-rural and rural-to-urban mi-ration.  - 18 - I.g. Backzround of Rural Migrants On the socio-economic characteristics of rural households that send out migrants, the nature of the "push" factors, and the process of selection of reglonal or job destinations, which combine at the individual or at the group level to determine the composition of outmigrants and the directions of their movement, there are a few micro studies which partially reveal some aspects of outmigration pattern. A carefully intensive, small-scale survey focused on the household characteristics of rural outmigrants was conducted by the Demographic Section of Bombay University's Economics Department in 1966 for 9 villages of Ratna- giri district in Maharashtra and 14 villages of Kutch district in Gujarat. These two agriculturally backward districts had been characterized by heavy outmigration, slowing down their population growth. The landless had been migrating as entire families and the cultivating households had been sending out selected male workers. Savla (1973) and Visaria (1970) analyzed this set of data and some of their findings are interesting in the present context. About a third of the reported male outmigration was accounted for by non- cultivating households. As the landless families tend to migrate as a whole while the cultivating families retain some to look after and derive income from the farm, the former group probably gets under-represented in a survey conducted at the place of origin. Their actual share in total or even in male outaigration may be even more than what was reported. In Ratnagiri, another 40% of male outmigration was from tiny peasant 'nouseholds cultivating less than an acre of land. For the rest (about a quarter of the total), in both districts the incidence* of male outmigration was inversely related to farm size. Savla's analysis of the Kutch villages points out an interesting diff.rence, namely that while the incidence of male outmigration was gener- ally higher among the households not cultivating any land, in the case of households bel,.ning to the top castes (Brahmins and Garashiyas) it was a reverse situation: the incidence was smaller for the non-cultivating than for the cultivating households. In the case of tnese two castes and also the trading/ business castes (Vaniyas and Luhanas), the incidence of male outmi- gration is higher among the landotw-ning that among the landless households--a very plausible situation that possibly has some general validity. The upper castes in the area have a tradition of male members migrating and sending back their savings for buying land as an investment. Even in their case the migration might have originally been prompted by poverty or lack of acceptable employment, though subsequently it resulted in the remaining members getting richer. A one-point study cannot be expected to capture the entire sequen.ce of events. Savla's study also shows that the outmigration and the subsecuent remittance resulted in varying degrees of increase in per capita income for the resident members of diffearent castes. There is a close positive associ- ation between the upper caste and per capita household income and the in- crease in income resulting from remittance. The remittance depend-s on the incidence defined as outmi-rants as a proportion of all rEsidents (in- cluding non-mi-rants, return migrants andl inmigrants) plus ouumigrants.  - 19 - kind of work which the migrant takes up at the destination, which--as any casual observer knows--is far from a random process and depends very much on the caste- or community-transmitted information and contacts guiding the migrants to specific types of work. Not only will the migrating Vaniya enter trade and business guided and aided by those who had already migrated from his kinship group, and the 14uslim or the Harijan migrant become a fac- tory worker possibly helped similarly by his kin in the city, but the process will also determine for him the exact type and location of his new Twork. It seems quite plausible that even with a strong push factor (unless it is extreme as in an epidemic or an acute drought), outmigration from rural areas is selectively-determined in terms of persons, timing and destination on the basis of the main available sy,stem of information and contact cumulated through past migration from kinship groups. This hypothesis is also corroborated by the studies relating to unskilled factory workers in urban areas, their composition and background [Morris (1965), Papola and Subrahmaniyan (1973), Bhattacharya and Chatterjee (1973)]. Most of these studies indicate that the process of migration from rural to urban areas is structured, connecting specific groups of rural migrants (from specific re-ions and/or specific groups or communities) to particular types of urban work. The mechanism of such structuring is a com- bination of factors: specific needs (like the need for seasonal jobs) and skills of the migrants, personalized or community based system of contacts and information, and the use of gang-leaders and other intermediaries for labour recruitment in industry and in civil construction works. Workers employed in public sector construction like canals and dams have usually consisted predominantly of landless agricultural labourers (Raj, 1960; Dandekar, 1Q56). Even there, mobilization through established links with contractors and labour intermediaries is one of the factors that produce concentration of particular groups of landless labourers from particular lo- calities not necessarily in the neighborhood of the project site. The famine relief projects are exceptions to this. Though we do not have enough data regarding the operation of all the factors normally involved, the available evidence -ndicates that the shape-ups of migrant labour tend to be group-or- community-specific, that the markets for different types of even unskilled manual labour are not integrated and that the pattern-of rural migration to other rural and urban areas are determinled not just by differential w.ages and employment opportunities. I.h. Willinr-ness to WY4Crate for ReP2-ular 7;iork NISS in the 25th and the 27th rounds canvassed some probing questions about whether the rural household had any member wi11ing to tak-e up regular full-time employ,ment on wvages, and about their preference for the location or the type of such employent and the expected wa.age. Some researchers (Raj Kriashna, 1976) have interpreted the answers to indicaLe readiness to migrate. The 25th round gathered the response of non-cultivating labor households and from small* cultivating households. IThe 27th round posed the Cuestionls to all * The lowest decile of rural households accordin- to the size of cul- tivated area.  - 20 - rural households, and grouped the answers by whether those willing to migrate were regular farm workers, non-farm workers, irregular workers. Observations based on this type of data corroborates that the pro- portion of Dersons willing to leavze the village for full-time employment is generally higher in the case of landless households than in cultivating house- holds with small farms. Secondly, among both types of households readiness to leave village for full-time work is much more widespread than actual unemployment during the reference week. In the case of male workers in par- ticular, the proportion willing to leave village for work is several times those reporting some unemployment during reference week. This observation is consistent with the earlier one based on NSS cross-tabulation of present and previous activity status of the actual inmigrants in rural areas. And both are consistent with the fact that the poor cannot afford to remain without work, that in order to survive they would take up in the short rtn any work however little it might offer. Eventually, some o-f them manage to migrate in order to improve their earniags. The overwhelming majority of those willing to migrate in this sense are in the prime working age group and most oL them are illiterate. From the answers (25th round) regarding the expected income at whi1ch they are willing to ta-e up alternative full-time work outside the village, Raj Kri-hna points out that on an average it is about 60% higher than the wage rate they reported having earned during reference week--a margin that he interprets as the prem- ium lor migration. Considering the percentage distribution of those willing to migrate according to the ranges of income e;pected, he suggests that at the all-India level there are three backward-bending migrant labour supply curves in three broad income ranges, possibly relating to three distinct classes with very little different skill/endowment levels and also present income levels. The hypothesis though plausible needs more detailed testing. Lal (1974), using the same data set at the state level and a similar interpretation of response to the rather vague question about expected annual income, reaches the somewhat different conclusion that the supply curve is upward sloping in a normal fashion--particularly for the non-cultivating wage-earning households. The main reasons for the difference is that Lal considers cumulative percent- ages of those willing to migrate at different wage ranges Twhile Raj Krishna takes the absolute percent implicitly assuming that those available at a certain wage range are not available for a higher wage range. However, in v%iew of the highly appro:ximate nature of the inccme e:Dectation figures from the 25th round, it would be better not to read too much in them. In this connection, we may refer to an intensive micro study of agricultural labour households in W7est Godavari district bv Parthasarathv and G. D. Rama Rao (1973), which indicates that, for the male as well as the female labourers prepared to work outside the village, the expected wage rate was 20-30o% higher th,an their actual wage rate, and for those who wvanted to go for non-agricul- tural work the wage ex:ectation was somewhat higher. The supply curve for work outside the village was found mostly rising forward, with a backw;ard turn at the End. Bardhan's (1976) analysis of the poor (households below a bare minimum of per capita consunption) subsets of small cultivating and non-cul- tivating labour hlou-seholds shows that in 1970-71 even in the agriculturally growing areas of Punjab, Harvana and WIestern U.P. as manv as eir-60 craent  - 21 - of these households reported having at least one person willing to take up regular full-time emplovment outside the village. Bardhan does not, however, interpret this entirely as readiness to outmigrate semi-Dermanently. For an employed and yet very poor household, the w-illingness to spare any member to take up regular full-time work outside the village can definitely be interpreted as a very meaningful indicator of disguised unemploy- ment (if it is a cultivating household) or low productivity/exploitative em- ployment (if it is a labour household). But how far it can be interpreted as an indicator of willingness to miarate is not so clear always. One can com- mute from the village or one may be thinkling of short-term stays away from it. In 1970-71, at the all-India level, 7% of all men belonging to small cultiv,ator households and 10% of those belonging to non-cultivating rural labour households reported as willing to take up full-time employment outside the village, particularly in public works at suitable locations. There was considerable regional variation. As a proportion of the adult male population of the trwo sets of households, they were more than 20% in ',est Bengal for both sets and in Punjab and Haryana for the non-cultivating households. It is doubtful whether such high proportions of the rural poor are entirely to be regarded as ready to migrate. In the absence of more specific information in the 25th round regarding the distance they are prepared to move away from their village and whether they are thinking in terms of a brief period or a long peried, it would not be right to regard all of them as prepared to migrate. For one district (West Godavari), Parthasarathy and Rao (1973) based on an independent intensive survey prov,ide the needed breakdown, in the specific context, in terms of the durations for which the labourer is willing to take up the full-time or the supplenentary employment, in terms of his locational preference, and the kind of work he is prepared or able to do. One of the findings is that of all the labourers with or without land who are willing to take e:tra or alternative employment 47% preferred it inside the village, 52% outside the village but within the district and only 1% outside the district. As this is the case of one district and as the willin2ness question in this study relates to supplementary as well as full-time employ- ment, one cannot quite gencralize on its basis or compare with ',SS. However, the ex-ample indicates the importance of the need for detailed characteriza- tion of the willingness to take more employment. The 27th round of N'SS was intended to fill up at the nationwide level an important gap on these issues. It canvassed fairly detailed questions about preferences for farm or non-farm work, inside or outside the village, in rural or in urban areas. And these questiorn were asK:ed not only from small cultivator and labour househ-olds but ali other rural households as well. Zhen the tabulations from these become available at the state level, it would be possible to separate out the cases of those willing to miarate for long periods, those willing to migrate seasonally, and those wanting supplementary or alternative full-time emplovmenc within the village. Together with their stated work preferences, wage e-pectations and hlousehold economic characteristics, this set of data -would constitute e:xcellent material for studying (a) the connected phenomena of lo,.i-income employment (or unemployment) and willingne--s to m igrate for diLferent durA- tions or stay. on for -uitably l ocated alte-rnative (or e-x-ra) employ.ment,  - 22 - (b) the centers of "pull" for the difLerent rural groups in different areas, and (c) the policy issues concerning how to structure rural works and other programs in term of the location, the timing and the kind of employment to be generated so as to match the specific needs and preferences of the poorer sections of rural population. I.i. Seasonal iMigration of Rural Labour In sections I.f and I.g, we have discussed long-term, more or less permanent migration of rural labourers on the basis of Census, NSS, and other survey data. It is possible that the 18th round NSS migration data contain an element of seasonal migration, but there it is not separated out. Seasonal migration of labour into and out of the village econom; is, from all observers' accounts, an important mechanism of adjustment of labour supply to seasonal changes in labour demand for agricultural operations. Seasonal inmigration of labour to rural areas is of two types: inflow of pure wage labour from outside, from relatively depressed areas with a single cropping season and without much animal husbandry; and the seasonal return of members of cultivating hlouseholds to help out at harvesting time on the family farim. How far the latter constitutes a necessary contribution to the to the required labour input during the peak season is not entirely clear. It may be partly a necessary one, though as Sen (1975) pointed out, "work,ing is not merely a method of producing output, it is also a way of establishing one's right to a snare of the family income from a joint farm." But the in- migration of outside wage labourers in the peak season and the corresponding outmigration from their own villages certainly constitute more genuine changes in labour supply in response to seasonal changes in demand or employment oppor- tunities in one area relative to another. During the recent period of rapid agricultural growth in certain regions and consequent sharpening of regional disparity, the importance of rural-to-rural seasonal flows as a mechanism of labour supply adjustment is likely to have increased. The data necessary for intertemporal comparison to check this are not available. The manner in which the seasonal inflow of farm labour is organized and offered in the local market matters very much in wage determination. Observers report different modes and degrees of organization of the inmigrant casual labourers in different areas. A prevalent mode seems to be through an intermediary or a labour gang-leader who negotiates the wage rate and assesses the amount of labour supply required. The seasonally migrating labourers usually move in groups. It would thus be reasonable to tnin'k that often the inflow, and by -implicaLion, also the outflow of rural labour for seasonal agricultural operations is more or less prestructured rather than a completely unoreanized convergence of ndependentlv movino labourers.  - 23 - Apart from moving seasonally to other rural areas for farm work and to tne urban areas for temporary employment in the informal industrial or services sector, seasonally underemployed rural labour is also absorbed in rural construction works. Since the Third Plan inititated a Rural Works Programme, there has been some increase in organized public sector efforts to generate rural employment through labour intensive construction in order to absorb seasonal underemployment. The Crash Scheme of Rural Employment, started in 1971, was designed to generate 10,000 man-months of employment in each district at specified minimum wage rates. Following a series of field evaluation of performance, these programs have been subsequently modified. There is now more emphasis on locating the pockets of underemployment, on adjusting the timings, and on providing more continuous employment and creat- ing assets to sustain a core of extra employment. Without going into the many details of policy issues of o.X ganizing seasonally complementary eTmploy- ment generation, we may note certain interesting aspects of the needs and preferences of seasonally underemployed labourers, as revealed by some of the field studies undertaken for evaluating performiance of the programs. It has often been noted (Rodgers, 1973; Dantwala, 1975) that in spite of apparently considerablc underemployment and/or reported availability for alternative employment, the rural poor do not seem to utilize fully the employment offered through public construction works. Some of the studies used questionnaires to probe into the reasons behind this apparent inconsist- ency. The 25th round of NSS indicated that a very high proportion cf persons from small cultivator and labor households willing to leave village for work were quite keen about employment in public works provided it was suitably loca'.ed and timed. The qualifying clause is important. Later on Dantwala poinced out, on the basis of field studies in two areas of Gujarat and one of U.P., that the major reasons why many of the low-income workers considered employment in public works unsuitable for them were: (1) being undernourished and used to agricultural operations or traditional vocations, they considered heavy earthwork and stonebreaking as too strenuous or requiring a skill they did not have, (b) the timing clashed with their seasonal employment in agri- culture, (c) the usually ad hoc nature and short duration of employment in public ,-orks. Rodgers too found these to be important reasons for rural labourers not participating in the lccal public w~ork;s. Other reasons included the wages offered being considered as too low. Thouch on par with or slightly higher than the local off-seaDon wage rates, these can still be too low for those who are prepared to migrate to other areas for harvesting and other peak operations. X'rother factor is the use of outside contractors and inter- mediaries rather than local institutions for mobilizing labour groups for the public works. Formal or informal tie or contract of a rural labour household to the farmer-employer is sometimes quoted as another reason, tend-ng to depress labour mobility in general in the non-cultivating rural households normallv considered to be a floating population.  24 - Section II. Agricultural Growth and Technological Change: Impact on WTages and Employment During the sixties, particularly since the mix-sixties when massive public investment took place in the New Agricultural Strategy to create the infrastructure for a seed-fertilizer centered breakthrough, in. many parts of the country an unprecedented pace and kind of agricultural change was ex:perienced. The irrigation-seed-fertilizer innovation is normally considered to be land-augmenting and labour-using, capable of generating large increases in employment and also distributing these more evenly over the year, thus reducing the seasonal under-employment and poverty of the landless and the landpoor. With tne use of perennial groundwater irrigation and shorter duration seed varieties, the usually lean seasons (from November to April in the single-monsoon kharif belt of east and central India and from MIay to October in the rabi areas in the north and the north-west) have become more active over the decade. Along with the shifting production possibilities with adoption of the biological-chemical innovation, related technical changes have also been taking place, some of them labour-saving and thus modiLying its effect on employment. Barring a Lewisian theoretical extreme of the supply curve of labour being perfectly elastic over a wide range, the large increases in demand for labour would normally be expected to have significant positive impact on both employment and wage rates for labour. What have the actual changes been like? What has been the impact of agricultural growth with the technical changes on emplovment? How has the structure oL wage rates and the composition of labour emDloyment changed in response to the shifts in supply and demand? Considerable amounts of data collection and empirical analysis relating to these issues have taken place meanwhile. Though there still are major gaps in both, these enable us to understand some of the employrent and distributive implications of recent agricultural growth. This section surveys the macro-level studies concerning the labour demand aspects of recent agricultural growth with technical changes, the actual changes in employment, for hired agricultural labourers in particular, and their wages. Secti-n III concentrates on the studies relating to the questions of market determination of wa,ges. Section IV surveys the work concerning micro-level issues of allocation of labour supply at the household level, and those oL composition of labour demand at the farm level. II.a. Nature of Technological Change and Implicarions for Labour Demand Perennial, controllable irrigation is rro,wth-inducinj and labour- using in two ways it extends the effective cropping period th.is increasinq the productive absorption of labour on an acre of l.an.d; and it w:ork-s as a vehicle Lor introducing i-nnovations embodied in the neo fertilizer-resDonsive  - 25 - seed varieties which raises productivity of labour use on each cropped acre. While most of the major public sector projects-for irrigation, flood control and drainage have made indispensable infLastructural contributions in this direction, the immediate impetus came, in much of north and north-west India, Lrom private tubewells that gave farmers the kind of irrigation that encouraged them to invest in a fertiliser-intensive crop rotation with the high-yielding seed varieties. Incidentally, a similar thing took place also in West Pakistan (Nulty, 1972). Although the water-seed-fertiliser innovation is normally considered to be scale-neutral, its adoption rate in India so far has effectively been rather unequal across size-groups. This has been due partly to the prevailing imperfections in credit (capital) markets. Though the initial installation cost of tubewells as well as the working capital for investment in improved prac- tices came in quite large measure from institutional sources of credit and not just from private capital, the actual distribution of even such credit in most cases strongly favoured the laraer farms. Some small cultivators eventually managed to use some of this type of irrigation--through buying from tubewell owners or mana-in- to stay on as cropsharing tenants paying higher rental shares in exchange of the landoTwner getting the tubewell installed. But this has been very limited. Buying tubewell irrigation from large farmers requires locational proximity and having to pay exorbitant monopoly price. By and large, "green revolution" has been predominantly a medium and large farmer phenomenon. The second feature of this phase oL agricultural growth has been its concentration in certain regions--regions that were even initially better endoa,;ed in terms of rainfall and canal irrigation. The reaional disparities in agricultural productivity are sharper today than they were in the late fifties and the early sixties. The regional and the size-group disparities in the growth of the leading inputs have been rightly emphasized in most of the evaluation of the impact of recent production growth on employment and levels of living of the rural poor. In most studies of the far= employment implications of technological changes, a distinction is made betw,een the effects of the water-seed-fertiliser innovation, of tractor ploughing, and of mechanical harvesting/threshing/reaping. This is obviously important not only for understanding -,:hat has been happening but also for luture policy purposes. For, use of the growth-promoting innova- tions, even though acccmpani ed by some of the labour-saving technical changes, may still be a basically distinct and independently viable process. Their concurrence, it has been argued, is largely a product of concentrated distri- bution of cultivated area, the fiscal policies, and the biases in the credit market fal.vourinfg large farms' investment in machines. Analytically and estimation-.ly the emiployment efLect of the see-~,a tter-fertlliser innol,vation Must be separated at the macro-level from those of the m-echanical innovations.  - 26 - WXnhen these two effects are not separated from each other, then the seed-fertiliser technology might seem to be labour-saving (Hayami and Ruttan, 1970, in the Japanese and the American conte:t). Hanumantha Rao (1975), using aggregative data for 265 districts in 1964-65, points out a lack of positive association between fertiliser use and number of agricul- tural workers per net sown acre, which he interprets as an indication that use of fertilisers and irrigation might have on balance been labour-saving. But this is not a conclusive indicator. An overall inverse association at the macro-level between per acre intensity of irrigation and fertiliser use on one hand and of labour input on the other could be the result of labour-saving mechanization that in some areas has accompanied the use of the new seed and fertiliser. Secondly, other important interdistrict variations (like those in cropping pattern, rainfall, tenurial and size distributions of land) have not been controlled in this comparison. One has to compare either intertemporally for the same region or cross-sectionally within a relatively homogeneous region in order to abstract from these basic variations. Thirdly, even Twith an unchanged or declining number oL agricultural workers per acre, labour input per acre could have gone up a lot if earlier the workers were only partly engaged in agriculture, if more of their labour time were now shifted from other kinds oF employment or idleness to agricultural work, or if there were an increased seasonal inflow of labour from other areas or decreased seasonal outflow. Most of the farm-level data (F'1S) indicate a consistently positive correlation between use of labour input and use of other material inputs per acre. Some estimates have been attemoted in the Indian context reaarding the effect of different types of mechanization on farm employment of labour. They are not very conclusive, the reason being partly the non-uniform procedures of the estimation and partly the inadequate data base and the approximation used. The major micro-level studies, using farm level data, are by Hanumantha Rao (1975), Lockwood (1972), Rudra (1971), Sharma (1972), Sapre (1969), Acharya (1973), Singh and Singh (1972) and Vashishtha (1975). The major macro-level studies on the direct employment effects of technical change in Indian agriculture are by Billings and Singh (1970, 1971), Sethiuraman (1971) and Raj Krishna (1974). On the indirect (or non-farm) employment effects, estimates have been attempted by Sethuraman (1971) and Raj Krishna (1974). The micro estimates--from farm-level data--of the coefficients of labour input per hectare (for a given crop pattern and for a given cateoory of farms, and for an improved crop pattern--with and without the specific -echnical changes) constitute the building blocks. With these, and taking account of the relative importarnce of different crop patterns, different categories of farms and the specified elements of a technological change "basket," one proceeds to4ards estimating the macro level coe.ficients. The element of the "basket" that has received the most attention is the tractor.  - 27 - MIost of the studies (both micro and macro level) on the direct employment effect of tractorization end up, as-Raj Krishna points out, by attributing to tractors some of the eflects that are really due to other innovations and farm practices. MIost of the differences in their findings, and the conflicting claims made on their basis, stem partly from differences in estimation procedure, how well they control for other concurrent--related but independent--changes in farm technology from which the effect of tractor use is sought to be separated out. Part of the differences are also due to regional or sequential differences in the phase of tractor use. Given other things, the direct employment effect of tractors would differ depending on how far they are used--seedbed preparation apart--for the more labour-intensive operations like harvesting, drilling seeds, irrigating, and transporting crops. Hanumantha Rao (1975), using Larm-level data for a progressive district (Ferozepur) of Punjab during 1954/55-1956/57 when use of tractor power was negligible, and during 1968/69-1969/70 when it was quite consider- able, attempts an estimar.e of the employment effect of tractor use. First, he compares between tract:or farms and bullock farms of similar size and irrigation levels, and Linds that the tractor farms have higher croppinig intensity and employment per hectare. This being a combined effect of tube- well irrigation, high-yielding seeds, fertilizer and tractors, he tries to separate out the effect of tractor use. He uses a multiple regression analysis in nwhich labour employment is a function of farm size, irrigated proportion, proportion under HYV seeds, fertilizer input, investment in tubewell, and two sets of dummy variables--one for tractor ownershinp and the other for thresher ownership. As the coefficients of the tractor dummy turn out as insignificant for both output and employment, he concludes that "the techno1ocical displacement of labour consequent on the use of tractors is roughly coImoensated" by the positive employment effects of the change in cropping pattern and intensity associated with tractor use. But, unless the increase in cropping intensity and in use of HYV seeds is essentially dependent on the tractor use, this conclusion though plausible is not quite established. Vashishtha (1975), using the same (Ferozepur 1968-69) data set, estimates the effect of tractorization on the intensity of cropping, on output per acre as well as on employment per acre. His method consists of working out Lrom a variance-covariance model the adjusted mean levels of each of these 3 variables for farms with tractors and farms without tractors, after standardizing for the effects of far-m size, irrigation intensity, inputs of seed-lertilizer-manure-pesticide, tubewell/purnpset, and thresher. Compar- ing the adjusted mean levels with each other and with actual ones, he finds that tractorization had no slgniificant effect on either cropping intensitv or output per acre. The observable differences between tractor farm:;s and ncr- tractor farms in these two respects were mainly the results of differences in i rrigation and seed-fertilizer inDut rate. As for farm labour employment per acre, while the standardized effect of tubewell/p=pset .,as positive and that of threshers negative, the effect of tractorization turned out to be neutral. The adjuszed mean 1, vels of employment per acre on farms with tractor and those without it when comDared with each other and w-ith rtheir observed mean levels shorwed that ver',' little of the ovlerall difference wvas attributable to the tractor.  - 28 - Both the studies, being based on Larm management data, are admittedly limited by having to analyse in terms of tractor ownership rather than tractor use. Besides, unless the breakdow9>n of total use of tractor in terms of various operations (particularly its distribution be-ween seedbed preparation, irrigation and threshing) is known, one cannot be sure about the stability of the employment cbefficient of tractorization. For, the employment effect would change with a different (operationwise) composition of tractor use. The third limitation of tractor studies based on FMS data is that the estimatud effects on cropping intensity, output and employment per acre are worked out: not for a specific crop, but for all crops put together, many of mhich may not: be using the tractor at all. Sharma (1972), with cross-sectional data from a pilot study of HYV Programme in the Karnal district of Haryana, and using the method of comparing between tractor and non-tractor Larms similar in terms of size and irrigation, concludes that the labour used per acre is about 7% higher for bullock-farms, which is not a very large difference considering that it might include the effect of the difference in fertilizer use. The studies by Singh and Singh (1972), Grewal and Kahlon are more or less similar, relating to western U. P. and parts of Punjab respectively. Sapre's study, relating to a relatively backward region (viz., Dhulia district of Maharashtra), takes output per acre as an indicator of employment per acre. It is perfectly possible that to the extent tractor use contributes to cropping intensity it also increases employment, with the labour saved on ploughing outweighed by the labour needed for more weeding, harvesting, threshing, etc., unless any of these are also mechanised with the help of a tractor. But even then the main problem is to separate out the effect of tractor use on cropping intensity from the effects of other related but independent factors like irrigation. Acharya (1973), with farm-level data from two districts of Rajasthan, compared labour use per cropped acre between a tractor-operated HYV user and a non- mechanized HYV user, and found that it was higher for the first one. This, of course, assumes cropping intensity to be entirely independent of tractor use, which may be so for a particular crop rotation and not so for another. Lockwood's study, though it does not measure the effect of tractor use, is interesting for at least two reasons. One is that he manages to study the features of all tractor using farms--those owning it and those hiring it. Secondly, he considers separately the employment effects of different operational uses of tractors. He uses the farm surviey data collec- ted by the Programme Evaluation Organization, during 1967-68 to 1969-70, for samoles of f.arms using HYV seeds and those not using them in 8 districts f-rom Punjab, Haryana and western U.P. In the sample of HYV using farmers, while 9% oawned tractors, 40% hired or concracted its service, althoughn the7 used it for smaller proportions of area. The farm size-distribution of tractor users (whiich included some non-HYV farms too) was much more broad-based than th.at of owners. In most of the projection exercises (as in Billings and Singh), it is assump-d that the tractors would be used in the multiplicity of farm tasks for which they are desi-gned and hence that there would be no time for hiring out, "unless the farmcer wants to use his tractor beycnd break-e-ven point." But in the samDles studied by Lockw;ood, mostly representing an initial phase  29 - of tractorization, the tractors twere actually being used "mainly for land preparation, including a good amount of contract work, and were not used for many additional operations" except for carting during harvest season and as "automobiles" during off season. As the number of tractors increases and their owners apply them more and more to additional operations like weeding and harvesting Twith the use of various attachments, the demand for labour is likely to be affected much more seriously. Large farms' investmenc in these attachments is morc common now than it was five years back. Since there is a definite large farm bias in the use of tractor (though less than the bias in its ownership), partly because of his greater access to working capital and partly because of some economy of scale involved in tractor use, it is of some interest to break doTwn the employment effect in terms of family labour, hired casual labour and permanent farm servants. Rudra's study (1971), based on a sample survey; of large farms (above 20 acres) from 11 districts of Punjab in 1968-69, makes some interesting comparisons regarding the composition of labour use. Considering the relative frequency distribu -ons of these large farms by the number of farm servants and the number of mandays of casual labour used for 3 mechanization categories--(a) farms having tubewells, pumps and tractors, (b) those having tubewells and pumps but no tractors, and (c) those having none, he observes that the mech- anized farms (particularly group a) save some casual labour in per acre terms compared to the non-mechanized farms but the intensity of use of permanent farm servants is more or less the same (about 26-29 acres per servant) across these categories.* Given a certain number of male family workers on a large rarm, the addition of tubewells and pumps reduces (at the cross-sectional level) the demand for permanent farm servants somewhat but increases that for casual labour. The further addition of tractors then increases the demand for farm servants and leaves that for casual labour unchanged. Pumps and tube- wells seem to displace permanent farm senrants, who presumably wvere used earlier for lifting water and maintaining the bullocks, and create demand for casual labour. Tractors seem to raise demand for Dermanent farm servants and reduce demand for casual labour. The net effect on total use of hired labour in farm category (a) is, however, not clear from Rudra's study. Considering all sources of farm labour, he finds that as the number of male family workers on a large farm increases, the demand for permanent farm servants declines within all the three categories, while that for casual labour declines in (a) and (c) but not in (b). Vashishtha's analysis of Ferozepur farms (mostly large) also indicates a slightly lower proportion of casual to all hired labour on tractor farms than on non-tractor farms, wnich is consistent with Rudra's point. * Rudra's mechanization effect on labour use per acre is admittedly mi:ed up with the farm size effect, if there is any even within the large Size-roup.  -30- The available micro-level studies in the context of north-west India tend towards the conclusion that tractorization at least in its initial phase did not have any Independent effect on cropping intensity. * Nor did the use of tractor significantly affect the use of hired labour per acre, as its use was confined mostly to seedbed preparation. It seems to have some effect on the casual-attached composition of hired labour, raising the attached component. However, this could be partly an effect of the larger cropping intensity itself. Lastly, in this region at least, tractor use reduced the use of family labour at the tillage/sowing stage. Bell (1972) studied farms in a somewhat different region--the Kosi canal area in Bihar. His conclusion is that although tractors used for land preparation alone could be called land-augmenting, their overall net effect on labour use would be negative or positive depending on the adoption of labour-intensive intercultural practices. Comparing between a traditional paddy-jute rotation with a 3-crop (wheat and two crops of paddy) rotation with high-yielding seeds, he observes that the constraint for the switch is not labour shortage but shortage of draught power which is required for two more or less simultaneous and critically time-bound operations (threshing the harvested crop and preparing land for the next one). Besides, he argues, the bullock power needed for the switch would be much more expensive for the farmer than tractor power--not only in terms of relative costs of comparable eEficiencv units but also the land area that would have to go in raising the requisite fodder. Given the particular time schedule of the improved rotation, and given the supply and the efficiency of local bullocks, use of tractor for land preparation is thus regarded as a concomitant of the labour-using seed-fertiliser innovation. For ploughing and threshing, human labour complementary with bullock labour is displaced by tractor. To get the net effect on employment, the labour displacement on account of these two operations is to be offset against the increased frequency of major opera- tions and the larger-sized crop of each for the harvesting operation. The question of trade-off between a somewhat smaller total employment and its greater seasonal stability (and shortening of total slack period), Bell points out, is also to be taken account of in considering the effect of tractor use for land preparation alone. Then it comes to projection of the employment effect of tractoriza- tion, one must take account of the likely or the planned composition -- operationwise -- of tractor use. The employment effect experienced in the initial phase, and analysed by the micro-studies discussed earlier, is specific to the particular operational composition observed at that tlme. Given the official encouragement of tractorization (liberal import, subsequent delicen- sing of its domestic production, subsidised credit for buying it), its continued rapid e:xpansion is likely to ma'ke the labour-displacing effect get stronger as it is increasingly used for more labour-intensive operations. Adjusting the- * However, together wvith tubewells and pi-ips, the use of tractors affects the cropping pattern to the e.-tent it does -away wFith the need to grow fodder for bullocks for lifting water and plough.Ping.  -31- growth in tractor stock to what would be needed for just seedbed preparation, coupled with institutional arrangement for handling its dispersed use; can be a policy issue of considerable importance. For macro-level projection of the direct employment effect of technological change, Raj Krishna (1974) carries out an algebraic decomposition in terms of the effects attributable to its different elements. In the same paper he surveys the other macro-level estimates of direct and indirect employment eLfects of mechanization. The basic procedure used in macro-level projection of the direct employment effect is to apply the coefficients of labour input for each crop, with the old and the new techniques of raising it, to the projected areas under each crop grown with the traditional technique and with the new technique, respectively. A major study of this type is by Billings and Singh (1971), who conclude that over the period from 1968-69 to 1983-84, farm labour demand in Punjab would decline by 17%, assuming that in the terminal year 20% of all cropped acreage will be tractor ploughed, the entire wheat crop mechanically threshed, all corn mechanically shelled, 50' of wheat crop mechanically harvested, and tubewells and pumpsets will cover 60% of irrigated area. They assume a 14% increase in cropped area and a cropping pattern determined by past trends and optimality conditions. Of the total direct labour displacement in the model of Billings and Singh about half rwould occur in the peak harvest- ing months on account of mechanised harvesting and threshing, affecting the casual labourers most. The remaining, on account of pump irrigation and tractor ploughing, would be spread over the year. In Raj Krishna's model, which clears up some of the common algebraic mixup in decomposition of incremental erfects, the direct employment effects in the Punjab context are worked out separacely (for wheat alone and then for all crop production) in terms of the effects of cropping pattern, cropping intensity, specified items of mechanization, and the interaction element. Assming a termlinal year pattern of technology similar to that assumed by Billings and Singh, his model indicates for Punjab over 1968-69 to 1973-74 a sizeable negative direct employment effect * for wheat and rice combined, which is composed of a large negatilve one on account of the assumed rates of mechanization, partially compensated by the positive effects on account of rise in cropping intensity and on account of change in rice-wheat mix. He then separates out the direct employment effect of tractors with their multiple uses. For wheat, assuming that ploughing and threshing are entirely tractori- zed and half of the pumping is done with tractor, the total direct employment effect of tractor is indicated as - 90 manhours per hectare, composed largely of the effect of mechanical threshing. For rice, assuming only the ploughing 0 .7 Raj Krisnna's (1974) model indicates effects of technology, crop-mix and intensity on labour use (manhours) per net hectare for wrLeat arnd rice combined (Punjab, 1968-69 to 1973-74) as -74.51, +17.04, +27.67 respec- tively. After allowing for a small interaction effect, the total direct amplo;-ment erzect oL the entire "basket" of innovations would be -30.14 mannours per net hectare.  - 32 - and half of the pumping are done by tractor, the corresponding employment effect would be less adverse (-27 manhours per hectare). For -heat and rice combined, the employment effect of tractor technology (comprising mechanized ploughing, threshing and irrigation) works out as -95 manhours per hectare ,which, set against the +64 manhours/hectare arising from the other elements of change (including those in pattern and intensitv of cropping arising from reasons other than the use of tractors), is the only reason in this model for the negative direct employment effect of technological change. The negative risubstitutionri effect of the technological change; Raj Krishna concludes, though outweighed by the positive "scale" effect of growth, would thus mean slower farm employment growth rate corresponding to a given agricultural production growth. It has been argued that the manufacture, servicing and operating of agricultural machinery increase non-agricultural employment which must be taken account of in evaluating the overall employment effect of tractor technology. Raj Krishna's illustrative calculations indicate that the growth in non-agricultural employment induced by something like a 5% rate of growth in agricultural production with labour-displacing technological change Twill not fully compensate for the slower growth in agricultural employment. To avoid rural unemployment, he argues, the policy maker would have to adjust the "growth-innovation-demand mix." In the existing literature on the emplovment effect of tractor, there remains a discrepancy between the macro-level direct effect (negative) estimated or projected on the assumption that tractor is used for the multi- plicity of operations for which it is designed and the micro-level effect (more or less neutral) derived from its earlier actual use which, from avail- able accounts, consisted during the sixties mostly of ploughing and very little of threshing. Jechanical harvesting, reaping, shelling, threshing are more strongly labour-displacing than mechanical ploughir.g. How far, from a policy point of view, tractor use has necessarily to be an all-or-none choice is not altogether clear. So far, during what has been an initial phase of diffusion of technological change in agriculture, the combined effect of controlled irri- gation, high-yielding seed varieties and fertilizers on flarm employment has dominated over that of mechanization. When it comes to macro-level estimation of likely employment growth, from the planning point of view of what is required lor utilizing rural labour supply, it would be interesting to e:-plore the role of institutional and policy alternatives in achieving a normative combination of the tTwio elements of technological change: the water-seed- fertilizer innovat on and the mechanization of labou-r-using operations. The explicit bias of the New., Agricultural Strategy towards regions with developed irrigat!on potenrtial, and the institutional biases against small farmers in general and tenants in particular in the distribution of credit and other infrastructural facilities have been noted in many studies not only for their direct dampening effect on employment but also for the  - 33 - indirect o-ne through affecting the very choice the technoloaical. alterna- tive for investment by large farmers. The issues of corrective policy concern- ing credit, agricultural taxation, land consolidation with a view to ground- water utilisation, etc., which have been engaging considerable research interest in the Indian context, are important from the point of view of influencing the mode of organization of agricultural production and the farmers' choice.of technology. II.b. Changes Over Time in Employment of Rural Labourers How has the labour demand effect of the given pattern of agricultural growth shown up in the levels of employment/underemployment for the major groups of suppliers of rural labour and in their wage rates? What has been the share of labour in the benefits of increasing production? How has the composition of farm labour input (in terms of unpaid family labour, wage labour, and, within the latter, casual and attached labour) been changing and what are its correlates? Besides, how does the non-agricultural component of the rural labourers' total employment relate to their farm employment, in a static and then in a changing setup? Understanding the operation of rural labour market, the shapes of the macro-level supply and demand curves and the mechanism of alloca- tion, in a siruation where decisions regarding supply of labour as between marklet and own-account work and demand as between family labour and hired labour are taken by a large number of households operating simultaneously on both sides of the market, is obviously Lar more difficult than in a hypothe- tical situation with capitalist farmers purely hiring in, non-cultivating labour households purely hiring out, and peasants completely outside the labour market. The mixed category of rural households, with own productive enterprise and entering the market to demand as well as to supply labour in various combinations, which by all eoDirical account is very sizeable in any region of the country, seriously limits the use of any pure-category analysis of rural labour market. Since a lot of small cultivators hire out some of their family labour and hire in some, since a part of the labour hired by large farm-ers is in the form of attached labour, since both small cultivators and landless households derive part of their wage employment from outside agriculture within the rural area, and since the rural labour market interacts with the urban labour market, it becomes difficult to generalise in termns of the available theoretical formulations. Fragmentation and imperfections in the rural labour :narket might result in rising wage rates for some suppliers of labour and low-productivity employment or unemployment among others even within a region. Similarly, there may be a lack of adjustment across seasons. Across regions, of course, institutional differences acdd an important dimension to the question of the impact of agricultural grow-th on employment and wage rates 'or rural labour.  - 34 - The rest of this section surveys the major findings of existing empirical studies on the structure and the changes over time in rural employ- ment/unemployment/underemployment and on the correlates of differences in employment/underemployment. Following that we move on to the studies on changes in wage rates for agricultural labour, labour's share, and on corre- lates of cross-sectional wage variations. II.b. (i). Emplo-ment and Underemployment in Rural Areas: Changes over Time The Report of the Committee of Experts on Unemployment was followed by a chain of intensive discussion on the conceptual and the empirical problems of measuring rural employment and unemployment, on interpretation of available estimates, and identification of qualitative characteristics of un(under)employ- ment. Raj Krishna's survey (1976) of this discussion is an adequate reference for the conceptual issues and methods of measurement. Recent rounds (25th and 27th) of NSS collected the time disposition data for each member of a household that are necessary for measuring flow rates of employment or unemployment for the household or some part of it (like, the working age men). The earlier rounds give the flow rates for only those members of a household whose employment during the reference week exceeds a speciLied minimum. Rounds of NSS up to the 21st, and then the 27th provide data for measuring weekly stock rates of employment and unemployment. 1971 Census and the 27th round of NSS yield the annual stock rates. Raj Krishna's survey conveniently puts together these measures in a comparable form. The annual stock rate of "usual status" or chronic unemployment increased over the sixties from 0.35% in 1961 to 1.07% in 1971 (adjusted with 1961 Census definition) and 1.01% in 1972-73 (from 27th round of NSS). However, this is not a meaningful indicator of employment/unemplovment for the major suppliers of rural labour, because the poor cannot afford to remain chronically unemployed at a reservation price. Stock rate of "current" unemployment is somewhat more relevant. It shows a declining trend over the decade (Raj Krishna, 1976). Definlng unemployment for a person as less than one hour of work during the reference week and availability for work, the rate was 5.6% in 1958-59, 3.8% in 1964-65, and 3.4% in 1972-73. Defining severe underemployment as less than 28 hours of work in the reference week and availability for more work, the current or w7eeksly stock rate of the unemploy'ed and the severely underemployed in rural areas seems to have declined from 11.7% in 1958-59 to 6.5% in 1964-65. The corresponding rate for moderate underemplovment (Tworking 28-42 hours in the week and available for more) dropped from 16.1% to 9.2,%. A rise in the stock rate of "chronic" unemployment alona with a decline in the stock rate of "current'" weekly unemployment is not diflficult to ex1lain. It is because the former relates more exclusively to the rural rich. Educat onal levels  - 35 - of the chronically unemployed are higher than those of the rest of rural people, particularly in the case of women (Raj'Krishna, 1976). The educated being usually better off can afford to be voluntarily unemployed while looking or simply waiting for a suitable job. Ahuja (1975) pointed out that if the composition of persons working below a specified cut-off level kept changing from week to week, then an average estimate based on weekl1y stock rates might be misleading. But if one considers separately the subrounds that comprise any rounds of NSS, then the seasonal fluctuation which causes this problem will be minimised. Besides, even in more aggregative estimates of weekly employment/unemployment rates, this problem will not arise if one uses Llow rates (mandays or hours per person per unit of time). These are not only less ambiguous but also more meaningful for the puroose of comparison across regions with heterogenous production and seasonality conditions. In the rural context wnere almost every member of a household (except perhaps the pure labour households) does some economic activity at some time or other in a year, for measuring the flow rates of employment/ unemployment of an entire household or any aroup of households or a state one must consider the time disposition data for all household members. This is what the 25th and the 27th rounds collected, but not the earlier rounds. Thus, comparisons over the sixties are difficult--particularly for aggregative rural situations comprising various categories of households. However, as the prevalence of part-time; seasonally varying, work participation is likely to be somewhat less among the labour households than among cultivating households, a comparison over time for the "workin-" members of these households might make more sense. 'NSS collected such data regarding not only employment but also wages/income/consumption Lor agricul- tural labour households in 1950-51 and 1956-57 and Lor rural labour households (including those engaged primarily in non-agricultural labour) in 1963-65. There is a definitional non-comparability problem between 1950-51 and the later ones: an agricultural labour household was defined in 1950-51 as one with the major part of its members' working time taken up in agricultural wage employment; later it was accordiog to w,hether the major part of household income came from wage employment (in agriculture for 1956-57 and all kinds for 1963-65). For 1970-71, 25th round collected data for the "`weaker sections of rural households," comprising two groups: non-cultivating .wage-earnings households and the lowest decile of cultivatina households. These tTwo groups, though not alway-s precisely representative of the population of agricultural labour households, do consist predominantly of this category (about 60-80% of their employed weekly mandays on an average were in agricultural wage-employ- ment, and the proportion is even more if one considers non-farma wage employ- ment) . There are a fewv other differences too w.hich affect comparisons over time; f'or example, the first two labouir enquiries (1950-51 and 1956-57) made a distinction between casual and attached laborers whi ch was not made in the later ones. How:ever, broad comparisons over time are still possible on s.Lome common basis for just those household members who are mainly working as  - 36 - agricultural labourers in 1950-51, 1956-57, 1963-65 and 1970-71. Aks the available tabulations are at the state level (except the earliest labour en- quiry which gave the data for regions within a state), it is not possible to use them for more detailed micro-level analysis. Not much use has been made of these labour enquiry data sources except to find out whether their earnings and level of livina had risen or declined. The data can be used also for analysing the pattern of cross-sectional variation in their employment derived from various sources, wages, modes of wage payment, nature of labour contract, etc. Comparing the case of the relatively full-time agricultural labourer between 1950-51, 1956-57 and 1964-65, picked out from the somewhat divergent sets of labour households covered in the three years, it seems that his total employment (on wage and on own account) in a year had on an average declined a bit from 267 8-hour days to 255 and then risen back to 267: a remarkable stability over a 14-year period, a slight improvement if one takes the more comparable years. The 12-day increase in per workers total employment from 1956-57 to 1964-65 was composed of a 20-day increase in wage employment and an 8-day decrease in self-employment, indicating that the trend towards increased market dependence oL the rural poor for work started even before the new technology that accentuated it. Unemployment comparison between the two earlier years and the later one is difficult because the distinction between casual and attached labourers was given up. The unemployment question is obviously relevant more to casual labourers. In 1950-51, a casual agricultural labourer was unemployed due to want of work Lfor 67 8-hour days in a year (i.e., for 20% of his days of labour supply). In 1964-65, unemployment was 48 days (i.e., 15% of labour supply) per worker for casual and attached agricultural labourers combined. In 1970-71, from the 25th round, unemployment during the reference week was 7.18% of labour supply of working age men (15-44 years) of non- cultivating wage-earning households and 11.75% for working age women. It was 5.54%' and 9.79%". of labour supply of working age men and women respectively of small cultivator households. How far the difference between 1964-65 and 1970-71 is a genuine decline in unemployment, and how far it is on account of differences in reference period, sampling design and sample coverage, have not been researched yet. Thus exact nu2erical comparisons of rates of employment and unemploy- ment over time are rather difficult with available data and are at best rou2h approximations. Cross-sectional comparisons are likely to be more accurate. It is also possible to compare ranks (across states) over time in respect of emplo-ment, unemploy=ent and wage rates for agricultural labour. An interesting difference, from 25th round, is that the flow rate of unemployment is significantly higher among the landless labour households than among the small cultivator households. .A.cross regions too there are significant differences in the unemployment rate among the non-cultivating  - 37 - labour households, which was considerably higher than the all-India average in Tamil 'Nadu, Kerala, Orissa and Rajasthan, and was considerably lower than average in N,aharashtra, Haryana and ASsam. Most of these differences can be better understood after one has considered them in relation to certain other characteristics of the rural households or the persons concerned. NSS 25th and 27th rounds collected data regarding some of the micro-level correlates of employment/under-employment, like farm size, consumption level of the house- hold, the age-sex distribution and educational levels of its members, the availability for full-time work, etc. Some household-level cross-tabulation has been done with 25th round for a few states by Bardhan (1975), in connection with identifying economic/demographic characteristics of the "poor" households. Parthasarathy and his co-workers (1973) have done interestina cross-tabulations for a sample of rural labour households in West Godavari district. The labour household survey. give an idea of that part of total rural labour supply which comes from h useholds with too little means of self-employment. The market supply of, and the market demand for, labour in most rural areas of the country is still not the major part of total labour demand or supply. The 27th round fills up an important data gap in this respect by collecting labour time disposition data for the primarily cultivating households as well as for the primarily non-cultivating wage- earning households. However, the relevant tabulations are not yet available for analysis. To summarise some of the research findings discussed so far concern- ing employment/unemployment of rural labour, particularly changes over time, at the macro level: The proportion of rural workers primarily employed as wage labourers has been going up irn the recent period in contrast with its stability over the fifties; and the importance of all self-employment in their total employment has been going down. The incidence of open unemployment (in terms of days without work but available for work) per person is more among the rural labour households than among other rural households. Correspondingly, disguised unemployment must be less in their case than in the case of the self-employed cultiv/ating households. Even in the case of labour households, the flow rate of open unem- ployment is quite small, and from available evidence, is going dow-n on an average on er tine. But this is partly because at any point of time the des- perately poor may not have a reservation price at which he chooses to remain unemployed. Bec:cer indicators of unemploy.ment in their case would be their willingness to take up alternative part-time or full-tlme elp1pvlment, aad their level of income. Their answers to probing aquestions regarding willing- ness to takle up alternative full-timme e:-.plovrMent have to be cross-tabulated with ot'her characteristics to identify r,inethe: the reason i s too little em- plo ,yrlent or too low w.;age rates or both.  - 38 - To get a complete idea of actual employment effect of the recent agricultural changaes, labour time disposition data for non-cultiVating labour households have to be studied along with those for self-employed cultivating households. Available NSS data do not permit this type of study over time. But the 27th round would permit a more integrated one-point study oL the labour supply and demand situation covering all major groups that generate them in rural areas. IT.b (ii). The Surplus Labour Question It has often been pointed out that, except for the landless lab- ourers, much of rural unemployment may be disguised rather than open. It may be disguised as employment: a given amount of work iln the household enterp.ise may be spread or stretched over longer hours or shared among a large number of family members, with consequent over-reporting of self- employment (in time units or in person units). Unemployment may also go unre-ogniied and get reported as outside the labour force altogether (Sen, 1975; Bardhan, 1975). When unemployment is defined as the time for which one is available for more work, the respondent may interpret the question as one relating to availability for insecure or otherwqise unsuitable work. Or, if he regards the prospect bleak anyway, he may simply not care to talk about it or too resigned to it to recognize that he is unemployed. Empirical studies on the surplus labour question in Indian agricul- ture have mainly been of two types. One involves direct estimation of the stock of labour time (with reference to a specified norm of working cime per adult -worker engaged in cultivation) and the required labour time for main- taining existing production level with existing technoloay, the difference between the two being a stock estimate of surplus labour. The second approach is through estimation of marginal productivity of labour in agriculture in order to find out if it: is zero, and if positive whether it is significantly different from the prevailing market wage rate for hired agricultural labour. The major attempts at direct estimation are by 'Mehra (1966), Ahuja (1973) and Rudra (1973). Using FIS data, MIehra calculates the required number of workers per acre for each size-group of farms by assuminr that it equals the actual number of workers per acre in the highest size-group (where the proportion of hired labour is the largest) multiplied by the mandays of labour input per acre in a size-group as proportion of mandays of labour i nput per acre in the highest one. The sum of differences between the actual nunber of wor'kers and the required number obtained in this way for the different size-groups is the surplus. This assines that the mandays of labour put in by a worker in the highest size group can be regarded as the norm for calculating the required aumber of whole-time wor'k-ers in the remaining size-groups. Tf there are major diLferences in cropping pattern among the various size-groups, and in the extent of employment outside crop production or farming, then this assum ption may not hold. Ahuja uses coefficients of labour requirement per acre for each of the major crops calcula ted as averages on the basis of farm nanazement data relat-ing to a number of states, and also a standard labour oeffl cieft for  - 39 - maintenance of animals. Applying these to the actual cropping pattern in Rajasthan, she works out for each of the two major crop seasons, the labour requirement. The difference between the actual total number of workers en- gaged in farm production multiplied by a 2400 hours a year work norm and the requirement is the stock of surplus labour days which is taken as the combi- nation of disguised and open underemplovment in this exercise. The use of standardised labour coefficient for a crop--in the form of an average of actual labour coefficients over a number of varied regions--implies that more labour is used for raising wheat (i.e., employment productivity slower) in a Rajasthan district than on an average in a number,of other major wheat-growing states. Rudra's estimate for a district in West Bengal, on the other hand, is simply that of open underemployment in farm households. He takes the actual amount of labour used in cultivation as equal to the required amount. Then he deflates the hours worked on the farm by family workers of farm households by a 2400 hours annual work norm to derive the required number of whole-time family workers. The rest of actual family workers is taKen to constitute the surplus. As the peak season work load is much higher than the average, Rudra calculates in the same way the surplus manhours for the peak season, and also for different size*-groups of farms separately. The smaller peak season surplus still works out to a significant proportion of male family workers actually engaged in farming. This is simply a measure of the difference oL agricultural work from the specified norm of 300 days in a year, which tells us nothing about how much of it is unemployment and how much is leisure preference or outside the labour force or spent in essential non-economic activiti-es. It is not surprising that Rudra's estimate shows up more surplus labour defined in this way among the larger farms where it may be largely voluntary. Among the smaller farms, the measure reflects worlk sharing among household members, as a concomitant of income sharina, which is an important but not the sole indicator of their underemplovment. Ahuja's and Mehra's estimates are intended to indicate low-produc- tivity employment along with open underemployment oL the work-sharing type. But the criterion for determining efficient employment that they use--per worker employment in large :arrms or an average of regional labour inputs per acre for each crop--can give misleading results for a number of reasons. Production functions even for the same crop can differ across regions and farm size-groups, depending, on soil, water and other conditions. The crop-mix can differ very much even within a region, among size-groups. Another major limitation of all of these estimates lies in the artificial separation of agricultural work from an essentially interate-d totality of rural employment. Ahuja takes account of labour input for animal husbandry wnich is important in Rajasthan. There are various other kinds of tiork as well--farm, non-farm, on own account, as an employee, inside the village, outside the village--that the family workers of a farm household combine in seasonally var zing proportions depending on the environment and their ingenui ry in utilising it. Calculating underemployment or labour surplus based on farm requirements alone is likely to give a gross over- estim.ate in maost rases. For planning the vollume, the timing and the location  - 40 - of additional employment streams to be generated through rural public works, it would be more useful to have ar estimate of the surplus mandays left over after taking account of all the different types of economic activities that the members of the household engage in and the labour that is required to maintain all the ongoing work. And when it comes to estimating the require- ment of labour, it might be safer after all to consider the actual uses of labour input in the various activities than to apply any standardized norm of average labour productivity taken from other regions or from a particular Larm type. But then it will not tell uis about disguised unemployment for the rural self-employed. The marginal productivity approach to measuring disguised employment has led to quite a Lew estimates with Indian data, the usual purpose being to prove or disprove the Nurksian hvpothesis that the marginal productivity of labour in an overpopulated, traditional agriculture is not significantly dif- ferent from zero, or to find out if in the case of family labour-based small farms it is significantlv different from the prevailing market wage rate. For this, the problems and the limitations of the marginal productivity estimation approach are well known, most of them pointed out by the authors themselves. These are discussed in details in Section IIT. c. Here we shall briefly men- tion the major ones. First, the estimated marginal productivity of labour includes the marginal productivities of all those inputs (like equipments, bullock labour) that are complementary to labour in production, but are left out of the production function estimate as they have to be in order to avoid the problem of multicollinearity. Secondly, family labour and hired labour may be different from each other in terms of efficiency per unit of time. Hence, an estimate based on some Iefficiency' units of labour input might give a somewhat higher marginal productivity than Twhat would seem to be the case Twith all labour regarded as homogeneous. The question of possible heterogen- eity of labour used for different farm operations is also relevant in a related way. Thirdly, the complementarity between farm labour used in dif- ferent seasons lor different time-bound operations with inflexiblL sequence makes it unrealistic to think in terms of marginal inputs of time-units of labour, as is asstned in the usual non-seasonal interpretation of production functions. Unless the marginal unit of labour is so defined as to incorporate the entire chain sequence of operational inputs of labour, the seasonal com- plementarity aspects may not be taken care of. This would be hardly possible with available data,* in view of the seasonal complementarity of labour used in agriculture being less than perfect and regionally varying. T The FNS data are not readily available Lor each farm in the form of labour inputs for the different crop-operations and/or for the different operation-periods, even though the data were often collected from the field in that form. S.K. Nath (see Oxford Economic Papers, Nov 1974) estimated, for Ferozepur district 1967-68, the marginal productivity of labour during slack and bus+> season separately. He found it to be zero for the slack season and positive for the bu-sy one, indicating seasonal undere-mployment. He took casual labor input as an approximation for the busy season and labour days of permanent workers as an approxizatio-n for the slackL-: season. The validity of this procedure is limited by the fact that the permanent workers intensifE the sha,re of their work tine on the farm during the bus:r season.  The operational significance of indicating the presence or absence of surplus labour in agriculture by measuring its marginal productivity is less than the attention it has received. tMore relevant questions are about its extent in different regions and for different sections of rural people. The issue of physical transferability of surplus rural labourers without affecting production is irrelevant from the policy point of view in a context where unskilled labour supply is not generally a bottleneck for industry growth, where there already are substantial migration `lows from rural to urban and other rural areas ar