CCT;>.RACIV.UJ ARRNGE2ENTS, ETLOY T AND WAGES IN RURAL LABOR MLA?XKTS: A CRITICAL REVIEW By Hans P. Binswanger and Mark R. RosenzweIg Series: Studies in Employment and Rural Development No. 67 Division: Employment and Rural Development Department: Development Economics Development Policy Staff International Bank for Reconstruction and Development This paper has been prepared as the introduction for a volume edited by the authors t4t'ed Rural Labor lMarkets in Asia: Contractual Arrangements, Employment and Wages. The volume assembles selected papers from a conference held in Hyderabad, India, August 22-24, 1979 sponsored by the Agricultural Development Council, and the International Crops Research Institute for the Semi-Arid Tropics. The views expressed are those of the authors and not necessarily those of the Bank. This report may not be published nor may it be quoted as representing the views of the Ban'K and its affiliated organizations. Washington, D. C. June 1981 RODUC'ION TEE PEASAT HOUSIOL AIM TE LABOR MART tiuri:stiO and Efficiancy Wages Labor SuppJy and Markat Structure: The Duality HypotheSis TJIZCT, SECL RPPJG AD OTME CONMCT AL ARUL GMETS Racenc Teucsy Xodels Var±taous ad Choanje in Cutractual Arangemets Mp±iic.al Studies of the Eff±ici.cy of Sbarscsopping Tenau.cy and T=ovation TES leteCrogeae±t of Labor, Wages, and Easr?ings Geographic Dispersion of Wages a;ut Mobilict Soucesc of Market Power Risk, the Cradit Marker, ad Labor RalatiD±u Testiug the Dmandi-Supp1y Frsanerk CONCLUSIONS In racent years the litarature concarned with economic development has shifted its focus to the rural sector of developing countries. Attention has turned from describing the broad outlines of the macro process of accnmi development to attempting to understand the institutions and the behavior of Individuals and fanilies Lu rural araas. In part this concern reflects the underlying reality that the bu1' of developing country populations reside Lu the rural sectrr. The recent generation of a large arrav of data sets - fs intensiTe village studies to large-scale nat±anal surveys - along with advances in computer tecbology 'ave also enabled more precise descriptiins of these settings as well as provided the material -which can be used to tast assumptions, models and hypocheses, and thus promote sc e±f oiSc progrsas. The davelopment of many of the models and theories of the rural econcmy Lu developing countries, hampered until recently by the absence of good empirical information, bave generally sought eithir to account for the existence of certain "stylized facts" which appear-ad t contradict the implicatIons of competitive models, or to provide the thaoretical underpinnings for the assumptions characterizing the macro "surplus labor" models (Lewis, 1954, Ranis and ?ei, 1961) which have dcminated the development 14terature. Among the -most popular stylized facts taken as data are (1) the coexistence of high unemployment and rigid wages (or discrep- ancies between the marginal product of labor and wage rates), (2) negative correlations between output per acre and farM size, and (3) the existence of share tenancy. -2- The essays assembled in: this volume depart from this tradition scmewhat. Thaemajor focus is on the determination ot the earnings of indi- viduals Wit the rural sectors of Sao4th- and South-East Asian societias, societies which are not so different as to bar a. coherent collection of studias, yet sufficiently varied to provida the contrasts which illuminate understanding of the more fundamental regularitias which may characterize rural labor markecs. Jhila the approaches and perspectives vary widaly, a unifying the is the characterization of the ectent to which markats for the lmportant factors of production in tha agrarian settiug, in particular labor markets, operate according to the prlnc±ples of the supply- deand, competitive mdel. A second theme is the issue of whether the same- times unique institutional features of rural markets - sharecropping or other contractual arrangements or even "institutiona±" wage races - reprasnt. barriars to the efficien: operaticn of markets, are optlmal. (second-best) responses to exogenous, tachuically determined constraints on mar- kats (market failure), or are reflections of collustva or otherwise ePloitative p r rtionships. The flaxibility of institutional arrangemeLts as ell as wages and earnings in labor markets when there are fundamental changes in the supply of or demand for factors, an important policy question, is thus addressed through an ezamination of the underlying causes of institutional arrangenents and their variability across different eccnomic anviromments in the rural sector. Much of the variations in rural wages and contractual terms considered in the conference papers relate to periods of five to fifteen years. General theories of development such as those of Lewis or Ranis and Fei are, however, concerned with very much longer-term trends of real rural wages. 'Wile such models are silent on the behavioral and institutional features of the rural sector, thae age determinacion and contractual choice theorias of rural ma-- kacs do not and cannot explain very long-term trends, as they treat the structura of technology and the sectoral composition of output as fixed. They also largely7 ignore the determinants of reproductive proce3ses and in- vesments in h:an capital which underlie population growth and the growth and comnsition of the labor force. The confarence on which this volume is based specifica11y excluded these very important topics from consideration, areas of research which must ultimately provide the linkage between the de- tailed short-rn models and evidence that is the focus here and a richer general theory of economic development. While the major focua of this volume is on the labor market, it is clear that labor earnings and employment are affected by the cnaracteristics of markees not directly involving labor. in a world with parfact markats for all factors of productin (including c.redit and insurance) an individual's income in each year would simp17 be thb emloyment of 'ais or her factor endowments valued at the market rate per unit. n this world, the initial distibuticn of endowmnts. among individuals, for given tastas and aggregate quantities of each factor, would uniquely determine the individual distribution of income. Moreover, production - total output - would be at a zaLum and would be un- related to the distr±buticn of the ownership of factors; production techniques ovtld be identical on all farms facing the same markat environment and oper- atisg the sme quality of land, e.g. output and employment per ace -would be unrelated to farm si2e; therefore productive efficiency could not be improved by any rearrangement of factor uses or distributions, barring scale economies. -4- To ae=lain labor earaings in such a world, with a given endowmenc discribution, would require an exulanation of the retuzns to each factor (wage rates, rent), for which the competitive supply-demand =odel has proved a powerful coal. Failures of one or markets, howaver, not olly would have important implica- tion for the distribution of earnings and productive eff±ciency, but is likely to require more complex models in ordar to understand earnings de- termination. An Important, unresolved question is whether such models ca. outpredict the simplar competitive models when only some markets are imper- fect or absent. Attention to market failure, however, is not oul7 important for under- standing tdie-detarmination of earnings and the achievement of productive VIdficiency, bxcg, as discussed below and in many of the essays in this vol.-, may aid in und rstandag as well the existence of and changes in the many and diverse institut4ional arrangements the t7pe of contracts, labor recruitment stratagies and the "1inteiinking" of labor and one or more factors of pro- I/ duction within one tansaCtio 1-- that characterize labor "markets". :ndeed because of the garal mon-independence - the "intarrelatedness" - of all factor markets, market failures anywhere in the rural sector may importantly affect labor market earnings or arrangements even if the market for labor operates perfactly. To explain earings then requires information beyand the determination of wages and labor supply. In this essay, we critically review the existing literature pertaining to labor and other factor markets in rural areas to place in perspective the contributions oa the essays in this volume. In particular we look at the various models and theories of labor makasts and tenancy with attention to 1/ Braverman and Srinivasan define interlinked contracts "as transactions in more than one comodit7 or service made between the same pair of individuals and linked in an essential way... delinking the contracts would be infeasible or costly for at least one party" (p. 4). the issues of absent markets, market failure, collusive -.power and the in- tardependence of markat3. A cantral theme arising from the review of modelling efforts in these tvo areas is that an understanding of institution- al arrangements or imperfections in any one market (e.g. labor) requires at- tantion to the imperfections in or constraints on other markets (e.g. land, credit). We will not attempt to give an impartial treatment to all ideas but will emphasize the major traditions of thought and recent empirical studies which are closely connected to the themas of this volume. We will also discuss some important themes which we think have been neglected in the literatre as well as what the essays as a whole may suggest for future work on rural labor markets. THE PEASANT OUSEaOLD AND THE LABOR MR&?K$T T'he theoretical and emvirical Literature dealing with the employment of labor and the determination of wage rates in rural areas of developing couu- tries has been shaped by concern with two major issues 1) how sensitive is aggregate agricultural output to the removal of agricultural laborers and 2) are there aspects of the rural labor market which bar the attainment of productive afficianc7 within agriculture. The first issu has taken onu 1mm portamce because an essential ass=wpt±on of smoe popula macro development models is that, in the initial stages of industrialization, agricultural output will be inariant to the transfor of labor fioz agriculture to in- dustry.-/ The theoratical literature has. thus focused for the mosiv -art on positing rural labor mar kt structures and/or madels of peasant behavior which could make this "surplus labor" assumption true. Likewise, the et- pirical literature on rural labor markets has been concerned with either 1/ The Ulteracu=e on planning models and benefit--cost analysis of projects is also concerned centrally with the opportunity cost of withdrawing Labor froam agriculcure. t-'stang the validiry of this develooment model asstupcion or with testing those assumtions of the theoretical models of rural labor market behavior which were developed to racionalize the assuptions of the development models.. (Tor a recent review for India see Lal 1976.) Two basic strands of thought characterize the "surplus labor" models of rural agricultura. The first assumes that rural labor can be withdrawn without cost because of the ex±stence of large pools of unemployed or tundereployed` rural workers. The theoretical problem is to reconcila large-scal unemployment or underutilization of laborers with a non-zero wage for labor. The second theoretical approach seeks to explain, the assued insensitivity of agricultural output to the number of laborers by distinguishing bet'ween the number of laborers and total labor supplied,focusing on the labor supply behavior of the peasant household as well as on labor market structurse. a. Nuturition and Efficiency rWag:s The most iufluential surplus labor model expla-inig the coexistence of idleness and constant wages is based on the efficiecmy or autritonal wage hypothesis elaborated by Laibenstain (1975) and later by Mazudar t1959), Wounacott (1962) and Stiglitz (1976). This model assumes that there is a te bnically deterMiMed/ positive rel3atioaship at low levels of income between nutritional levels and labor effort per unit of time (or per laborer). G±ven certain (weak) assumptions abour the shape of the income-effort relationship, thera is a unique wage - the "eifitaiency wage" - which minimizes the cost per unit of labor effortZ/. Maximization of profits 1/ As Bliss and Stern (1978) stress, if the relationship is not technical it will be almost impossible to test empirically. 2/ The thaory assumes that labor can be measured in efficienc7 units. That this may not be possible when there are more than two factors of production has been discussed in the appendix to .chapter 5 of Binswanger and Ruttan. -7- by farmers in surplus labor economies entails hiring laborers until the margial value product of total effort (or efficiancy units) hired is equal to the efficiency wage. Some woricars would be left unezployed in such a setting; although they might be willing to work at a wage lower than the efficiency wage, it is not profitable for farmers to hire them, no matter hay uch time wages ara bid down, due to the reduction in elfort associated with lowered wages. Given profit maxizization, large ==bers of laborers relative to land and the assumed (non-behavioral) relationship between the level of nut:rition anLd effort, it is clear that unemplo7ment and positive, even high, wage ratas will coexist; moreover, roval of laborers from this systen will neither affect output nor the wage. Tf, of course, the marginal value product of the last laborer hired exceds the efficiency wage and no more idle laborers are available, the wage rate will rise and the conventional supply-demand fraewcrk pertains. The problen with the profitsaxImizing utritional wage, model, of course, is chat in tha abject poverty, conditions which are assmed, the une,4loyuent equilibiru. described cannot be long run, as unemployed workers would have no means of survival. Two additional assuptions are thus made to avoid a Malthusian result. Leibenstein assumed tha; because of social pressuras landlords collude to lower wages, sacrificin3 profits and total output in order to support more peopla in the economy. Ea thus provides a rigorous definition of underemployment, if not a reaistic model of the rural econmy - those additional people who obtain wage work as a result of the collusive wage reduction belaw the profit- maximiz!ng (efficlency-wage) level are underemployed in the sense that their transfer out of agriculture would not lower output. Stiglitz achieves the same result by assumi.g that fami 7 farms have egalitarian con- carms about the earnings (cowsumption) of family members; although the model does not consider what happens to wage workers with no land. In both versions of the nutritional wage theory (because consumopion determines labor effort at any given time supplied to worlk), utility =axi- m±zat±on and profit m-aim±-ation are thus not compatible. Equalizing con- SI tiou among workars (Laibenstain) or betveen workers and nonworkers in the family (Stiglitz) lowers their effort and thus profits.. Moreover, in both models the subtraction of laborers actuall7 increases output since the wage paid, and thus total effort, rises; the margimal product of labor is thus negative even though wages are positive. Another way out of the dilema posed by the starvation of unemployed landless workers 4s to assume that the unemployed have a fall-back option in self-ezployment activities such as bunting and gathering or a nonagri- cultural entarprize (mat weaving) which assures them survival at a lover utility than the agecultural wage. Such a model can indeed account for the coe-4 tence of =employment and a positive wage. However, because the marginal product of the "unemployed" workers must be positive they thus caot be withdrawn to the industrial sector at zero cost. -Curthermore their supply vould not be infinitely elastic at a wage equal to their - ginal product in hunting and gathering since liMits on the hunting and gathering groxnds would impl7 that withdrawal of some of the self employed would lad to a higher marginal product in self employment of the remaining ones. The third means of "saving" the nutritional wage model is to assume that it holds seasonally CRodgers, 1975) - during times of the year when labor demand is very low the efficiency wage is then the floor below which wage races do not fall despite high (seasonal) unemployment. At other times, when labor damand is relatively high, the conventional suppqy-demand frame- work holds. However, Bliss and Stern (1978)., based on a careful review of the nucrition literature, conclude chat the reLationship between hu±r tional intaka and effort must be weak within short periods of time because the huan body acts as a store of nutrients. Any nutrition-effort association is heX-sef ort unlikely to apply to daily wage contracts. Furthezore, the relegation of the nutritional wage theory to slack season labor market phenomenon clearly limits its empirical importance and increases the dif- ficult7 of detection in the absence of estimtes of the technical wwage- nutritio£-eff ort relationship. Tmpirically, year-long open unemployment appears not to be a marked phenomenon in rural Labor markets (?aglin, 1965; Hansen, 1969; RosenJzeig, 1980), although this would appear to .be consistent with both efficiency wage and conventional competitive market theories. Rodgers (1M75) has atptempud the most a=bitious tests of thIs f-amework, exploiting an implication of the thecory not emphasized by its advocates, anel-y that employers would pay attention to the actual conSUMptIoU of the workers. Thus, wokers 'w ith dependents would by necessity have higher nutritional wages thn. uzattached workers; sin.il1a , workers from landed households - with rental incme from land - would be better fed than landless workers and thus would supply the same effort as the latter for a lower wage.- Rodgers finds that in the group of villages in 3ihar from which he has data, average area wage rates are higher where 1/ Bliss and Stern (1978), hovever, show that thi3spredict±on holds only if applied to te regions with separate labor markets; in one region all workers have land and in the other all workers are landless. f landless and landed workers coexist in the same region, the prediction is not unambiguous. households are primarily Moslem, i.e., where women tend nao to be workers. While this evidence is consistent with employers paying higher wages to males with more dependencs, it is also predicted by the supply-demand modal in which man and v.Raen laborers are substitute factors in production. Rosenweig's stud7 of Indian district-level data based on a demand-supply framework in this volume indicates that both male and female wage racas in agriculture are nigher where 'Hoslem households are prevalent. The hypothesis of individual heterogeneity in wages as a function of land ownership or uber of dependents is rejected in Rosenzweig (1980). It is also incoasistent with the uniformitt7 of wages for daily paid adult worka=s (of a given sex and for a given operation) found by Bardhan and Rudra (1981) in villages in West Bengal, Binswanger at. al. in semi-arid ladia and Whit, and Makali in West Java villages. Rodgers also njOits out t1lat the nutcrtiona wage theory might account for labor t7ing arrangements if effort is a function of sustained nutritional intake. Long-term employment ccntracts exceeding several weeks, however, are not very frequent in the South and South-East Asian context (Bardban and Rudra 1981, Binswanger et. al., Whita and Makali). Furthezore they seen to b- related mcre closely to the demad for specialized skills om an assured basis (such as bullock driving, herdsmen) and the need for credit and problems of adequate collateral of the workers. (Bardhan and Rudra (1981), Bhalla, 1976, Binswanger at. al.). Moreover, nutritional considerations canot explain the absence of long term contracts for fmale workrs.-L/ 1/ Sheila Bhalla notes an exception in Haryana where, altar the green revolution and substancial wage and income rises, feales were also offered longer term contracts. 3ut this obviously cannot be attributed to the wage-nucritiou relationshin. Given the generally lower wage raca for remales, nutritional considerations should apply.to them as well as to men. Efficienc7 wage relations can also be based on other than nutritional grounds and have been applied primarily to the nonagricultural sector. Can these versions be applied to agriculture as well? With workers of equal pro- ductivity the efficiency wage relationship could be based on "moral" effects where workers receiving higher wage provide more effort. In such a model survival of unemDloyed workers could be assurad through the sharing of income by employed workers with unemployed members of landless households, because the efficiency of the worker does not depend on his own nutritional intake. To our 1mowladgs this is therefore the only "model" which could account for both the coexisteuct of unemployment and positive wages and for a zero cott of withdrawing labor from agriculture. It has not been subject to either careful theoretical or erzpirical analysis, however.- 1J Tom ocher modeli which rely on a relationship berveen wages and efficiency to ecplain the coexistence of uemployment with positive wages are based an labor turnover and on inzormation constraints: The labor turnover model (Stiglitz 1974b) is based on the notion that firms paying higher wages have lower labor turnover. Raising the wage from a low level could thus initially reduce turnover sufficiently to reduca efficiency labor cost. TIn agricultnral labor markaet3 casiial or daily labor provides the bulk of labor inpvt and such laborers have a ver7 high turnavex anyway (Whita. and lakali, Binwanger ec. al.). The turnover model would appear to have Limited applicabilit7 where turnover costs are evidently very low. Another reason for an efficiency-wage relat±onship is the possible sceauins f'mction of wages (Weiss 1980). If laborers are heterogeneous i.e. have dfeare-at lesvels of inherent efficiency, then high quality workars should have high opportunity costs in scme self-employment acti- viyt7, while low wage workers should have low opportunity costs. At very low wages the pool of applicants therefore consists only of the lowest quality workers. In raising its wage to progressively higher levels, the pcol of applant3 -will start to contain worke-s of higher and higher qualit-7. If f±im' have no way of distingUi3hhin between' high and low ouality workers and paying them accordingly, they draw at random out of -.the pool of applicants. Raising the wage leads to a higher quality work force and we are back to the results of the efficiency wage theory. In the context of casual agricultural wage ;orkers, in a peasant agricultural setting, this model can hardly apply since information about the quality of laborers resident in the village acc=ulaces over time and is widely sharad. It can also be acquired at a very low cost, for example by hiring a worker for one day only. Note that in this model unemployed (but self- employed) workers also have a positive mar3±nal product and thus cannot be withdra'wn from rural areas at zero cost. In the tenancy section we will return to a similar model by Newbery and Stiglic: about screening of poetential tenants. -12- The assumpriorns and implications of the nutrition and morale versions of the efficiency wage hypothesis are suarized in figure 1. The first two colUi present the outlines of the nutritional, efficiency wage models of Sciglitz and Leibensteia. Boch of these model3 rely on either landlord or worker altruism to obtain a steady-state ecuilibriumi. This equilibrium is characterized by "underemployment" rather than by open unemployment, with rural output increasing as rural labor is -ithdrawn. When efficiency wage relations are justified on nutcrtional grounds, a steady-state equilibrii= with excess applicants can exist only if the landless unemployed have some non agric=ltural self employment opportunities in the rural areas, which would prevent them from starving, colu= 3. If the efficiency-wage rela- tionship is based on morale effects, however, the landless unemployed could either £) rely on such self employment (colu= 3) or ii) share in the consumption of their employed family members (colun 4), which is not oossible under the nutritional wage version without a loss in out=u.t Wb.ila in each of these variants, e.cess applicant equilibrium is possible, rural output (both agricultural and rural nonagricul.tural) would be in- variant to the withdrawal of workers only under the morale-.age model vith sharing. It is thus the only true "surplus labor" modal. he nutrition- wage mdal variant is unlikely to be an important deter.inant o0 rural wages in Asia because nutrMiton is u-likely to affect aglort in the short- n and only short-te= (daily) labor arn-angemants are cconly observed, especially -or wcmen. The wage-moralef-elfort relation is very di-f'cul: to detect. Since wages vary sharply over the season this association is either not stabla, i.a. vi ly untestable, or applies only to the slackc 1/ Furthersre, if the wage-morale-effort relationship exists but is unstable, the relevant question becomes what factors determine the relationship at any time. Unless such factors exist and havre a stable impact on the wage- effort-nutrition relationship, the theory is not testable and/or useless. b. Labor Supply and 'Market Structurs: The Duality Hypaches-is 'he alternative route to rationalizing the possibilit7 of a zero marginal product of laborers in agriculture rel±es on the distinction betvaen labor tize supplied and labcrers. If the withdrawal of one family meber always (or on average) leads to an increase in the anount of work supplied by ather family members which is just equal to the work formerly supplied by the departed family work=, total output would remain constant. The marinal product of labor time thus could be positive while the margiual product of a laborer could be zero. This view is evidently subscribed to by Lewis (1972). The theoretical issue then is what model of the peasant family could account for this behavior; the ampirical. focus is oan family labor supply determnats. ?olowing a long tradition of models of peasant farm families (Chayanov), Sen (1966) proposes a nodel it which the fami1y unit consists of workers and =Workes AW0 jointLy maimizze a £family welfare functin conutainig as arguments the leisure time of each worker and the consumption of each family member. The family owns a fixed plot of land, with farm output and thus, incom solely a function of total family labor time, as family labor time cannot be sold in the market nor labor time purchased. Assuming the separability of the family -tility function, Sea shows that total 1/ ..Ryan and Godhake measure both wages and uemployment rates over the season f or the same panel of households. They find that wages and unemployment rates both vary, with unemployment rates much higher and wages lower in the slack season. Refined analytical models and very careful empirical work is required to sort out the relative roles of adjustment to season- ality via unemployment and via wage rate changes. -14- ?±gure 1: EMCIENCY-WAGE THEORIES AID SURPtUS LABOR IN AGRICrLTUHZ- BASED ON BASED ON N~rrzoz. ~MORAL~E RMATIONSHIS F. CTS A11 landless workars (l)No landless (2)Unemployed (3) Employed Landless employed because workers exist have non- workers can share employers sacn±flce and farm families agricJtural consuption with profits (Laibenstain) share work and/or self employmnt unemployed fa-ily cousiption (Stiglitz) opportunity members without effecta on effor: qO E%.3RM WT. ?OSSIt LE EQUMI3RIU£ WITH EXCESS APICANTS EXCESS APPLICANTS (recorded unemployment) MIRAL OUT?PT RURAL OU=P RURAL OIITPUT INCR2ASES DECREASES INVA3.IAT HET WOas wHIEN WORKERS WEMN WOR RS LEAVE LEAVE LEAVE nutritional relation- real wages var7 across ship is not short regions and over years as well as seasonally. long term contracts wage-morale relation- Limited in agriculture ship either not expecially for women stable or conifned to slack season 1/ Turnover and screening versions are unlikely to apply in a peasant agricultural setting. See footnote 1 p. 10. far outpuc rmains invariant when there is a reduction in the number of workers only 4- the marginal utilities of both consumption and leisure are constant in the relevant range; so that the marginal rate of substi- tucion between Leisure and consumption is a constant. While the San coad±tions may or mayr not be plausible, it is important to note that they pertaiu to a farm household totally isolated from che labor markat; the nodtl indeed assumes the absence of any rural labor markat whatsoever. For households in which (identical) family workers work for wages, however, the marginal rate of substitution between leisure and constmption is equal to the market wage ratea, which is in turn invariant to the behavior of the individual household. The Sea conditicn of tha constancy of margzal utilities of leisure and consumption would thus lead to an indeterinate solution within the relevant rarge for a household with wage wmrkers. The inapplicability of the San conditions to hcuseholds with wage earners, or for economies in which all households, participate in the labor market., illustrates the Importsance of assuaptious about the rural labor markat in pradicting peasant behavior, rather than the impossibility of completely c=mpensazinzg labor supply behavior by the family. In a household with no d which. consists of identical workers with dependents (non-workers) a reduction in 'family members (cr the workc time of ona family member) would reduce the income (consuption) of every umber. but not the real value of tlme (cost of leisure relative to the price of goods). Ia this case, the empirical issue is the magnitude of the income effect on leisure. For non-identical wage workers (male and female, for example), £t can be easily shown that the response of male (female) work time to an exogenous change in the work time of females (males) is equal to the ratio of the cross-zompensacad wage effact to tba own coapensated wage effect; i.e., the direction of response depends on whecher mala and female leisure time are substitutes or complements in the family welfare functio. Boeh U.S. labor supply studies (Schultz, 1980) and the Zndia= study by Rosenzveiz in this volume suggest that this cross wage aiffect on labor supply is negative; the craation of ano-farm enployment would thus result in an inc-ease in the work tIme in agriculture by family embers of the opposite sex, i.e. some compensation takas place. None of these studies, nor the recent study of 3arnuin and Squire (1979) (which assumes family members are ident±cal) indicates, hovever, that family labor supply respcnse would fully ccmpensate for the reduction in the family labor force. The assumticn of Mon-Carticipation by households in-the labor Market, i! true, would have a number of iMportant imolications. Fi=st, among households with identical plots (size and quality) of land, the margina product of labor will differ according to the prefereuce orderings and demographic stZucture Of the households. Because of these diffr-ntial "sub ective" equilibria, productive efficiency is not attained, in the absence of a weLl develoced market.f or land. A, second Importanj feature of households which cannot trade labor is that increases i the price of ouc;uc could lead to reductions in tise worked (the dominance of th. income effect) and thus :I output supplied. This is in contrast to the case of well-functioning markets for all inputs (including labor), where the price-elasticity of output on indiiidual fars must be positive. If some farm. housaholds can participata in the labor market and others cannot, then a "dualistic" agricultural economy results as long as the land rental market is also absent. This extreme dualistic assumption, that some farms maimize profits and ucilize hired landless laborers and others, small farms, use only family labor and do not participate in the labor markat (work on other fam.s), is one popular e=lanat±on for the comon and well documetad obser7ation that small arms- employ more labor per acre than large farms.- In this mdel, the cost of labor (the wage) to big farms is likely to exceed the marginal rata of substitution between leisure and goods on the isolated small farms. As Sen points out, howeveres, this "strong" version of labor m-rket dualism requies tha: in addition there be no trans- actions in land, since large land-ioners could increase profits by leasing out small plots of land to families, taking advantage of their low oppor- tunity cost of labor. Giveu the failure of t-om markats, land and labor, a dualistic agzriculutral structure clea-ly is inefficient. The empircal evidence, as presented byy ?aglin (1965) for India and aas.n (1969) for ZE3t, and as confired by almost all the empirical Studies in this volume, strongly rejects in the Asian context this extreme form of duallsm. Members of small fa households appear to parttcipate substantiall7 in the labor market as both buyers and sellers. Given the pervasive evidence on inter-farm labor mobility within labor markets, more subtle hypotheses - wnich we will call "weak" dualism - have been put forward to account for size diffirentiais in. output per acra in 1/ For a recent review see Rudra and Sen (1980). terms of labor cost discrepancies across large and small armf. The argument is that, given the uncertainty of agricultural production, employers of labor or sellers of labor from small fars (who may be the same persons) are unwilling to make large future labor co=ients; there appears to be only a limitad market for contingent labor contracts (Bardhan and Rudra 1981). As a result, most labor is hired, on a daily basis. Z!oreover, because of differences in the timing of operation across fars, workers work for many employers during the year, a:s shown in Binswanger, et. al. and Whita and Makali. Transactions costs associatad with job search are thus quita nigh in tha "casual" labor markat, irth substantial resorting of of employers and employees ea-ch day. It vould thus not be surprising if the probability of finding employment or finding enough laborers on any given day was not equal to one. On small farms where individuals are primarilyr sellers of labor, days when workers cannot find market employment nigh: be spent wcrk:!=g on the land, up to the point where the marginal value of the utilit-y of leisura, rather than the markez wage, equalled the marginal product of labor effort. On large farms, for those days, in which not enough laborers could be found at the market wage rate, the marginal product of lAbor would exceed the market wage. Due to these frictions - transaction costs - in tha labor market, therefore, over the year the marginal product of labor could be lower an small than on large farms. One of the important contributions of the Ryan and Ghodake and Pranab 3ardhan papers in this volume is the cozputation of uneployment prqbabilities for wage labor. They find that the percentage of "working" days in which sellers of daily agricultural labor reported that they could not find wage work was 14 and Z3 percent respectively for maLe wor-kers (2 per- cant for femnla workars in the Ryan study). P. Bardhan (1979a) also has found evidence thit tha probability of wage empaoyment tends to influence the market participation behavior of wcmen, although not that of men. Ryan construccs a dtLect test of the "weak" dualism hypothesis - that average "opportunity" costs for labcr ara greater on large than ao smal_l fars. wherc the opportunity cost of labor for small far households is the product of the employmet probability and the wage - and finds mixed results for male labor, but confirmation for feale labor because of the greater measured ±nvoluntar7 unemployment of the latter group. The existence of transaction costs associated with the problem of job matching are, of course, only ose of the possible sources of departure of the rzral labor market from the perfectly cOMpetit±ve Model. Others ne-tioned, cnsizstent with the weak dualistic or "`iage gap" assumption, are the "dislike" by small farmers of or social pressures against wage employment (possibly for cer.tan 3-roups or castes) or flied costs associated wich (traveL time), or, on the side of employers, the costs associated with supervising hired workers. The latter is akin to scale diseconcmias under the assumption that hired workaers are utilized only if famfly labor time is insufficient to drive down the marginal product to the market wage. Again, however, the ineficiancies associated with these labor market imperfections could be circumvented if the market for land were funccioning perfectly. Indeed, as discussed below, a aajor motivation for sharecropping is the problem associatad with superVising hired labor (in the absence of a land sales market). -20- To suarize we note thae strong dualism, i.e. the absence of a rural labor markat, is not necessary co achieve the labor surplus result of the Sen model. A maximizing model of the farm household in a perectly compec±t:ve market setting can theoretically lead to fully compensatory labor supply behavior. The issue is entirely an empirical one, and the astimates by Roaenzweig and by Squire and Barn=u i.ndcata less than full compensation. Strong or weak dualism are therafore seen to be aeithar necassary nor sufficient to yield the surplus labor result. Either assuMption only leads to divergent opportuity costs of labor, and therefore to differential factor use across farm sizes. While there is little evidence to support stroug dualism both strong and weak duali: must assume that there exist at least some imperfections in the land rental market, a topic addressed in the tenanc7 literatures which we now discuss. -tUTA=7, SE-:'0?'tC .MD Oa C0Nc7NC;xE A.iRMM.C=ES Whila the literature on the datermination of wages and employ- =ant just reviawed takes the distribution of la1d as ecogenously g±ven, i.a., the sales or rental market for land is assumed to be absent, it is an empirical reality of rural economies that s=e labor i3 cmbined with land, not by the taporary sale of labor services 'but through the tempora acquisition of land. It is clear that the terms and arrangements associated with the markat for land importantly affect the earnings of rural households and the production of aggregate oucput. The first four of the papers in this volue, two theoretical (Jaynes, 3raverman and Srimivasan) and two empirical (Jodha, Raumasset), are part of a rapidly expanding literature concerned with the role of contracts which combine labor (and or other factors of production) with land when certain factor markets are absent or incomplete. -21- The major focus of the many theoretical models which have been proposed in this it.erature is on tew issues: (1) what are the efficiency character4stics of a contract which provides labcrers with a share of t-otal agricultural output, an important contractual arrangement in the rural economy, and (2) how do welfare levels or earnings of such sharecroppers c=mpare to those of laabrers who work oly.r for wages, i-e. what daternines the contractual tarms. (aJ Recent Tenanca Hodels./ T only the sales markeat of land were absent or invol7ed very high transactions costs, a land ower could hire optimal quantities (relative to h.s own land) of all cooperating factors of production, including bullocks and magement, and rent out an,y e.xcess (non-land) factors he owns over and above the optImal amount. Productive efficiency, i.e. equal factor ratios on all far;s with ecual quality land, could still be achiered. An absent salas markat for laud is thus not sufficient to force the u3e Of tenancy. Tae institution of tenacy and the markat for tenancies subst±:ute for the absent sales makaet. At least one other factor market must be absent (whae there- are no scale economies) before tamporar7 rentals of lad becCMe a nacassary tool to achieve the most efficient factor rat±os for a"l factors of production and &lI agent. The absent or incomplete markets (inolving high risks or transaction costs) could be those for insurance, fAmily labor, bullocks or anagerial skills. 1/ For other recent reviews covering similar materials see . and Zusan (1980) and Ntewberry and Stiglitz (1979). -22- Whether sharecropping, one form of tenancy, is a productively efficient system of cultivation ralative to (a) self cultivation with the help of wage labor or (b) fixed rent tanancias,has occupied economi-ts sinca Ada Snith. heQ classical economists, including M!arx, understood sharacropping as an adjustent to the absence of markets or markat failure, in particular of the markets for credit and capital. Within a setting of imperfact markats, and given that it is difficult to supervise labor, they 7±eewd sharacropping as an Improvenent over wage labor, because of its positive incentive effect. Given that tha worker shares in the output, he has an incentive to provide more labor than he might provida under a wage contract if not supervised very closely. Eowever, they also recognized that share cropping provides the workers with less ir-entive to work than a fixed rate tenant or an owner cultivator would have. As discussed for=a1ly in fashall's famous footnota, an a= axcgemousl7 gi-e area c: .and and with a given share, the worker rece.ives only a fractiaa of his argiual product, with the fraction aqual to his share.- The classical economists also understood that the sae incentives problem applies to a1l other iuputs and especially long term invest=ents iu land quality. They therefore regardfed long term tenancias at fixed rates as a superior systan to sharecropping, if the level of development of a country or region permitted it. . . .. .. ....... / As Jaynes shows in this volume, Marshalls thaory of sharecropping was a complete classical theory and the footnote only ilustrates the incentives problem. Cheung's work (1969) set the stage for the recent share-cropping litarature in terms of the major issues to be ,;.Adressed, and in carms of the major reasons for share tenancy. His work attacked the negative efficiec7cy (incentive) implication of sharecropping and also broadened the scope of enquiry of the sharecropping litarature to discuss the 'ssue of aow and at what level the size of the tenancy and the share of the crop are dacarmin. All the litarature frcm Cheung onwards has regarded both the tenancy siza and the level of the shara as endogenous to the respect:ve models, while taking the wage rata as exogenously Siven. Contractual terms, but not the wage rate, are thus determined by economic forcss, with the equilibriu solucion to the contract choice problem involving =a-41i- zatiou both. by the laudlord and by the workers. 'Workers equilibrtu: requires that "of the set of contracts available in the economy, there eaxists none which the indIvidual worcer -rpefers to the one which he has." And landlord equiLibrium impi.tes that ".. . there exists no subsat (of the available cautracts) which the Landlord prefers to the subset which he employs." (Stiglitz, 1974, p. 222). Ch.mg also assigned to r'sk and risk aversion a much larger role as a determianarof share tenancy, although not in his for:mal model. It can readily be seen that under a wage labor systen all the risks of cultivat±-on are borme by the ovner-cultivator. Owner-cultivator income is the residual after paying production costs at fixed wages. Under a fixed rate tcancy, all risks are borne by the tenant since tenant's nicome is the residual after pa7ment of a fixed rent; but under share tenancy the risks are divided between the two in proportion to the crop share. -24- AS Jaynes' paper in this volume shows, however, Cheung's model achieves its efficiency outcome because it assumes away the problem of the negative incatizves of shariag as well as the difficulty of monitoring affort; undar those a3sumptions, share tenancy would not be observed. Cheung indaed needs to introduce risk, risk aversiou and "transactions" costs informally to explain the very existence of tha contracts whosa terms he explores in a formal model where such sources of market imperfections don't exist. This is an approach similar to the one advocated in the last section of the Eoumasset paper in this volume. With respect to the risk aversion mocivation for share-cropping, Newbarry (a975) and Raid (1976) showed that (with constant returns to scale) shara cropping provides no risk sharing benefits which landlord and worker could not achieve by dividing a plot of land "into tro subplots, one of which is rented out at a fixed rental R and the other is operatad by the Landlord who hires labor at a wage W ...." (Newbery and Stigl4tz, 1979, p. 314). Therefore a model in the Chaungian tradition,J i.e. without probl ems of worker incemtives, does not provide an explanatiou for the existeace of share tanacy, even if there is production risk and risk aversion (Figure 1, ==clusion C). Sharecroppiug can also be a means oi risk avoidance under more complex characterizatious of risks. Newbery and Stiglit= demonstsatad that with a second Independent source of risk, such as wage rate risk in the labor market, share contracts are superidr to micing wage and fixed rent contracts.. Jf there are no incentive (monitoring) problems or econcies of scale but there are multiple sources of risk, the sharecropping contract acts as the necessary instrument to achieve productive efficiency, i.e. it prevents an inefficient allocation of resources rather than creating one (Figure 1, ConcLusion D). Another class of cenancy models focuses on the costliness af labor supervision as a cause of sharecropping i.e. the ",lrshallian" inefficiency. Stiglitz (1974) developed cne model (among several) which assuines coscly super7ision, with the landlord setting the size of the tenancy just like the share, taking into account the impact of tenancy size on the tenant's input decision. The landlord can prevent the tenant from renting any other land or ftlam working for wages, or he can stipulate these provisions in the contract and monitor and enforce them.-/ The landlord thus has an extra control inst-u=mnt and can - via his maximization - control. the contractual terms in such a way as to limit the tenant to his reser7ation utilit7 i.e. the wage rate. Productive efficiency is not achievable in this model, of course, given the effort mcnitoring problem. The contribution of Braveman and Sr±invasan in this volume is an extansiaon of the Stiglitz model with costly supevrision in which the tenant and landlord are allowed to engage in a simultaneous share aum credit contract, with the credit being used for the tenant's consumption. Such a tied contract becoes superior to an untied contract if the landlord bas access to credit from thi-d paries at lower rates of interest than the tenant. The landlord sets four contractual terms, amely the crop share, the tenancy size, the rate of interest which he charges 1 Bardhan and Sri±ivasan (1971) 'had developed a si=ilar model. Eowever they did not allow the landlord to control the size of the tenancy. Newbery (1974) quickly pointed out that, in their model with incentive effects, full employment equilibriu could not ecist since tenants would attempt to rent land until its Marginal product was zero, i.e. there would be excess demand for land. 3raverman and Srinivasan in this volume shows that in the model of Stiglitz, when landlords cnot control plot size, there may in fact be an optimal share level which elicits the lavel of effort at which landlord profint ars maximized, but which give sharecroppers a higner utility level than ia the labor market. Just as in the efficiency-wage models, an equilibri= with excess supply of tenauts may therefore exist. -Z6- the tenant, and the proportion of credit recuirements which the cenan: borows frem the landlord. Given that the landlord hnas two extra instrumencs available, the landlord can almost always hold the tenant to the Utility level which he would obtain as a wage laborar. ?olicies such as teenrcy reform or provision of credit to the enuant at lower than market rates therefore cannot Improve the tenant's utility level. Nothing short of land red±.stribution, intervention in several markets, or rising alternative wage levels can effect the tenants welfare.- Costly supervision arises in these models because of imperfect information. The landlord car-ot know at a sufficiently low cost how =ich effort the tenant provides, o7ly the tenant can know it. Information is asyetrically distributed between then. A central planner subject to the same informational asyme=ry conld not improve on the existing allocation. Such improvements could only. be achieved if the central planner 'ad cheaper means of monitoring available than the landlord (an u".like. situation in the case of agriculture). Alternavely the central. plamer would hae to redistribute land to the tenants to overcome their inability to purchase land in the land market (which leads to tenancy in the first place). But such a policy would also Improve efficiency in a decentralized economy. As long as the underlying constraints on information or land transfei are not removed, the share tenancyr equilibrium achieved is optImal with respect to the inforation and land market constraints, i.e. it is a second best optImuz, ralative to the set of infor:tional constraints assumed in tha model. Th}is point is !mportant recurrent th-ee. A problem not addressed explicitly by the models discussed so far is the coexistence in the same region of all forms of contracts: owner cultivation, share contracts and fixed rent contracts (see Jodha, this voluMe, for an example). 'toreover, "tenancy ladders" appear to be i=portant in developed and developing countries, with workers first becoming -27- sharecroppers, then fixed rent tamants and finally acquiring land of thair own. (For a'discussion in the T.S. contat, see Reid, 1979.) There are three "atxplanations" for the coexistence. of tenurial contractual arrangent: (i) differences in risk aversion, (ii) screening of workars of different quality and (iii) market imperfectians for inputs other than labor. Differential risk aversion, however, cannot account for the tenancy ladder, since there is little reason to expect the samr individual to become completely risk aeutral as he becomes older, even if he accumulates assets .-! It is therefore neceassar7 to recogni ze that workers differ ia other respects, such as ability, management skil or capital ndowments. If productivity per hour of work differs among otherwise hcmozemeous workaerz but t81 productiity differences are mown only to the workers and cannot cost- lassly be observed by the landlord, landowners or workers face a screening 2/ cost. Zn this case -allagan (1978) and independently Newbe r d Saigl±tz (1979) show that the choice of cont-ac: conveys infa ation about the workers' perception of his ability. "IndiVidual3 who believe they are most product- ive. [as workers] will choose the rental contract; individuals who beliewe they are very unproductive will choose the wage contract ard those in between will choose the share contract." (TNewbery and Stiglitz p. 323). Each of the classes of workers prefers its respective contract. Utility levels for the Mora able workers are higher than they could achieve in a labor market without screening. Since information is asymet:ically distributed between landlord and workers, productive ef!icienc7 cannot be achieved. The implicit screening via contract LI Complece risk aeutrality appears to be extremely rare among rural populacions in developing countries, as indicated by several experimental studies (Binswanger 1980, Sillers 1980, 'Walkar 1980). 2/ The problem here is similar to screening models im the wage labor market which can result in an efficiency wage relationship for average worker quality as discussed in footnote of p. tC . Excepc hare the scraening instr:ment is a contract with more coplaex terms racher than a wage rate. choice again rapresents a second best improvement in efficiency over the situation without tenancy contracts. This model leads to coecistence of contracts but not to a tenancy ladder, unless werkers graduate from one 1 efficiency ci.ass to the other as they grow older. The clearest route to the ex±sCanca of a tenanc7 ladder and social differentiation of laborers and different types of tenants is via absent markets or Imperfect markats for inputs other than labor. dtdiiadbilities of inputs laad to economies of scale and Newbery and Stiglitz point out that eccomies of scale can make share cropping attractive in the absence of incentive effects even if there are oi.Ly r±sks iu production Iadivisibilities arise in the casa of bullocks or other capital. equipments i rental markets for them are deficient. Rantal maraets iu bullocks may be absemt or poorly developed if there is lack of flexibility in the timing of bullock operations (e.g. seeding)-/. Such inflcxibility makes reliance on rented bullocks too risky. ndivi.sibiltties may also arise in management skills as in the case of the 3Be.-Zusman models (of which we consider only the latast 1980 version below) where landlords can get access to the tenants' managerial skils only by, renting land to then. T/ he model does not consider managerial efficiancy but only efficiency as raw labor. Managerial skills tend to increase with age. 2/ Bliss and Stern (1981) attributa the, ecistance of sharecropping in the-r North ndian village lar3ely to imperfect bullock markets and show that bullocks ownership is an essential requirement for renting land. Jodha's pape in this volume confirms this for semi-arid India by showing that tenancy is closely associated with the equalization of bullock/land ratios across the farms studied. Sheila Bhalla (1976) shows that villages in Earyana distinguish between tenants and other long term labor contracts paid by a crop share depending on whether the laborer or the landlord provides the bullocks. -29- Credit market imperfections are the chird major reason and are built into a capital rationing model by Jaynes discussed briefly in this volume. la Jaynas' modeL eacb. contractual f£or (wage cul=ivation, sharacroppin& or fixed rent tenancy) requires a fixed amount of landlord's time for supervision per unit of land allocated to the cont=act in order to overcome incentives.problems. Nevertheless imperfect markets lead to the possibllity of co-ccistance of all forms of contracts and to a difexentiat±o of the taers of contracts among different tenant-landlord pairs becausa of diffarencas in their respective factor endowments. Tn sharp contrast to models whnre tenats are not diversified by labor 'skills, untradeabla managemet skIls or capital endowments, the welfare level of tenants is no longer the exogenously-Siven reservation utility level offered in tha wage. labor market (or by a completely cradeable endcwment of all factors, as in Jaynes' model). Different forms of contracts are available to different indivriduals because they allow them to make bettear usa of their unique endowments wh±6c (because of market Imperfections) could.oth.zise only be used in a less effict way. These con=acts thus improve tenants' utility levels. In their most receat bargaining theoretic model * Bell and Zusmar. (1980) cosider risk, incentive effects and four factors of production: land aot cradeabla in a salas market but only via tenancv; labor which is freely nobila between share and fixed rent tenanc and the outside labor market; fertilizers or dther modern inputs, also freely tradeable in a perfect fart:iioer-acum-scredit market and finially management capacit7 of the tenant, which can. differ across tenants and which is compleely noncradeable. A landlord can get access to it only by ranting land to tenants. Clearly this fourth factor could also stand for other -30- nontradeabla components of rhe tenant's endowment such as bullocks, or even female family labor (as in Bangladesh wnere the Cain a,.. al. study in this volume. suggests that female workers can workc outside their own farms only to a very limited extent). Output share, input share, Afied rental rate and tenancy size are all decermined in a bilateral mcnopolistic bargai-inz process in which both landlords and tenants have some extent of power, delcaxm±md jotly by the levels of their respective endowments and by their relative number. Thus this model can accomodate all levels of landlord power, from pure onopoly to large number ccmpetitiou. Tenants c deal with many landlords, a frequent phenomenon in many parts of the world. Equilibrium is reached when no landlords or no laborer can improve his utility by signing other contracts or changing landlords. Labor, fetilizer and ma enmet input are tenants' discrationar7 variables. Clearly, tnants' welfare in this :odel vii tlbe u.igher -hati`ha .ti'ity levi.l a-ciaveales W -a- pure wage workars. The possession '- management skills gtves then some bargai3ing power except if there is only one landlord and thsupply of ta is infinitely elastic, in which case management is no longer a scarce factor. As sumarized ia Figure 2, the theoretical literature of tenancy suggests that there are several alternative combinations of reasons wnich can explain the institution of sharecropping. En order to explain some form of tenancy, of coursa, the land sales market must be riddled with ±nper!ac:tons and a second markat imper'ection is raquired (Canclusion A). WhMen risks are present and crop insurance is absent the mixing of fixed rent contract3 and cultivation with wage labor can substitute for the absent insurance market (Conclusion C). Only if the mixing of contracts is Impossible or uneconomical (because of economies of scale in cultivation) does it become necessary to FIGUME 2 : E ROUTES TO SHARECROPP?SG .-BSEXT SALES 4ARKTZT FOR LAND (A) Insu:fficient to cause tenancy ADDMSONAAL MAR= T wEICETON (3) Some Form of TenAcy ABSM?CZ or CROP 2ISURANCZ (C) T=sufficient for Share Tenancy H LMZLZ NO MI=DG OF .kS41ZETIC INCENTVE E3EECT 2DEPTDIET CONTRACTS W?ORMATION PROBLLS ISPUT MARIZTS, RISMS (ECONOMS OF ABOUT LABOR NONTADEABLE SCALE) QUALTT =?PCT$, DtS- ECONOIES OF SCALE (D) 7CR) 0F) ^(G) Sharecropping Shsecropping Sharecroppiecrhrecropping Sharecropping as iasurance as insu;ance for screeninst as insurance to get access to and incencive nontradaable