by A Mukherjee · 1995 · Cited by 71 — seasonal fluctuations in income and imperfect credit markets suggests that tied contracts should dominate casual labor markets. However, empirical observation
33 pages

98 KB – 33 Pages

PAGE – 1 ============
t-!~-~ “-‘~ ~ II ELSEVIER Journal of Development Economics Vol. 47 (1995) 207-239 JOURNAL OF Development ECONOMICS Labor tying Anindita Mukherjee a, Debraj Ray b,, a George Washington University, Washington, DC 20052, USA b Boston University, Institute for Economic Development, 264 Bay State Road Boston, MA 02215, USA Received 15 May 1993; final version received 15 April 1994 Abstract We study labor-tying in a competitive agricultural economy. The co-existence of seasonal fluctuations in income and imperfect credit markets suggests that tied contracts should dominate casual labor markets. However, empirical observation from India suggests that this is far from being the case, and indeed, that there is a declining trend in labor tying. We consider a model that permits deviations ex-post from mutually agreed implicit contracts. In equilibrium, casual labor markets are always active despite the presence of seasonality, and a variety of implications are derived that link economic growth, changing information flows, and the decline of labor tying over time. Keywords: Labor tying; Seasonality; Implicit contracts; Agriculture; Rural labor markets JEL classification: J41; J43; 012 1. Introduction The term ‘tied laborer’ or ‘attached laborer’ is popularly used in the literature to identify any laborer who commits his labor to some particular farmer for an extended period. The period is to be contrasted with that for casual laborers, who are hired by the day and sometimes to complete an operation lasting a few days. We may think of two broad categories of attached laborers. There are those who perform special tasks that require some judgement and precision, and are * Corresponding author 0304-3878/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved SSDI 0304-3878(95)0001 1-9

PAGE – 2 ============
208 A. Mukherjee, D. Ray//Journal of Development Economics 47 (1995) 207-239 difficult to monitor 1. The need for the efficient execution of such tasks is apparent. It is also clear that such laborers must be given appropriate intertemporal incentives 2 In the second category of attached labor, there are no special tasks performed. Attached labor might be used to perform the tasks of casual laborers (see, e.g., Bardhan and Rudra (1981), Bhalla (1976), Rudra (1982a), Breman (1974) or Sundari (1981)). Tied labor contracts need not necessarily carry with them any special duties, but may simply serve to lower wage costs 3 In this paper, we limit our attention to this second category of tied labor, though the framework of the model that we develop can be easily extended to include the first. 4 Three general observations on labor markets in India motivate our inquiry. First, this second category of attachment has been dominant in the past. Second, the incidence of tying has undergone a steep secular decline to low current levels. Finally, there is marked regional variation in the levels of attachment in this category. There is ample evidence in the literature on the organization of villages in social/economic anthropology that in certain Indian villages, the entire agricul- tural labor force could be partitioned into the tied labor pools of different landlords on the basis of the jajmani system. The following description may be found in Lewis and Bamouw (1958). 5″Under this system each caste group within a village is expected to give certain standardized services to the families of other castes. Each man works for a particular family or group of families with which he has hereditary ties . The family or family head served by an individual is known as his jajman, while the man who performs the service is known as the jajman’s kamin or kam karne-wala (literally, worker).” Another widespread form of labor tying, somewhat less structured and formal- ized, may be loosely referred to as patron-client relations (see Beteille, 1979). Under this system, too, the employer is ideally supposed to ensure the general 1 Such tasks include ploughing, regulating the flow of irrigation water from pumpsets, driving and looking after tractors, supervision and recruitment of casual labor, operating threshers, etc. In this context, see, e.g., Bailey (1957), Binswanger et al. (1984), Freeman (1977), Government of India (1960), Reddy (1985), Rudra (1982a,b), Sundari (1981) and Thorner and Thorner (1957), Bhalla (1976), and Mukherjee (1992). 2 See Shapiro and Stiglitz (1984) and especially the work of Eswaran and Kotwal (1985). 3 Basant (1984), Ghosh (1980) and Rudra (1982b) demonstrate on the basis of large-scale survey data on Indian agricultural labor markets that the daily wage equivalents of tied laborers are often lower than the casual wage rate. Similar evidence from surveys of smaller scale may be had from Bardhan and Rudra (1981) and Breman (1974). See also the interesting recent study by Anderson- Schaffner (1992) linking attached labor and servility. 4 We do this in an earlier version of our paper (Mukherjee and Ray, 1992). 5 This description is in agreement with accounts in Srinivas (1955), Srinivas (1960), Beteille (1979), Hopper (1957), and Hopper (1965).

PAGE – 3 ============
A. Mukherjee, D. Ray/Journal of Development Economics 47 (1995) 207-239 Table 1 Secular decline in the proportion of attached laborers in Thanjavur, India. 209 Village Year Percentage of laborers Semi-attached Casual Kumbapettai 1952 52 48 1976 21 79 Kirippur 1952 74 26 1976 20 80 Source: Gough (1981) well-being of the employee, and in particular help the employee out in times of financial crisis such as sickness, death or drought. In return, the labourer is expected to give maximum importance to the needs of this employer as regards the allocation of his time. A large number of the kamins or clients carried out casual tasks, were paid a daily wage (in addition to their traditional payments) and were free to work for others when the jajman or patron did not need them. 6 This corresponds precisely to the latter type of labor tying as defined by us. Hopper (1957), Lewis and Barnouw (1958), Breman (1974) Gough (1981) and Vyas (1964), along with a number of other studies, describe the secular decline of this traditional patron-client system. Table 1 documents proportions of tied laborers in the villages surveyed by Gough in 1952 and 1976. Apart from the secular decline in the proportion of semi-attached laborers, there is also marked regional variation in the proportion of attached laborerers who carry out casual tasks. While surveys in West Bengal (Bardhan and Rudra, 1981) and Tamil Nadu (Sundari, 1981) find significant proportions of attached laborers in fully monitored, or casual tasks, contemporary surveys in other parts of India find a relative absence of such arrangements. This applies to the villages located in the semi-arid parts of India studied by the ICRISAT 7 and also to studies by Chen (1991) and Reddy (1985). The same applies to a more recent survey of agrarian relations in Uttar Pradesh, Bihar and Punjab by Mukherjee (1992). In contrast to these empirical observations, standard economic theory predicts widespread labor tying, whatever the nature of the tasks performed by attached laborers. Given that the technology of traditional agriculture involves steep fluctuations in labor use, and the impoverished state of the agricultural laborers, it is very simple to furnish a theoretical explanation for labor tying in terms of implicit contract theory, s The theory predicts that recruitment of attached labor should always be optimal whenever there is any interseasonal variation in 6 See, e.g., Sundari (1981), Gough (1981), Breman (1974), and Hopper (1957). 7 ICRISAT is an acronym for the International Crops Research Institute for the Semi Arid Tropics. 8 See Azariadis and Stiglitz (1983) or Rosen (1985) for surveys of the theoretical literature on implicit contracts.

PAGE – 4 ============
210 A. Mukherjee, D. Ray~Journal of Development Economics 47 (1995) 207-239 Table 2 Proportions of tied laborers in ICRISAT villages, a Village Type of farm Percentage of farms employing farm servants Aurepalle Type 1 13 Type 2 47 Shirapur Type 1 6 Type 2 7 Kanzara Type 1 0 Type 2 7 a All the attached workers in these villages appear to carry out non-monitored tasks. Type 1 farms are small and medium farms while Type 2 farms are large farms. Source: Pal (1993) economic activity, as there certainly is in agriculture; and if workers are more risk-averse than employers, which is a plausible assumption. 9 Note that perfectly foreseen seasonal variations in labor use do not change this observation. There is scope to tie labor, even if they will only be used in the peak season (this will be formally equivalent to an interlinked credit-labor deal). According to the standard theory, there is no reason why (predicted) peak season labor should be acquired on a casual basis. Of course, in the presence of uncertainty in labor demand the entire labor force may not be tied because keeping a large inventory of labor unused may involve unnecessary expenditure. However, even with this amendment (see Bardhan, 1983), the casual labor market will not function at all unless the state of nature is unexpectedly good. Put another way, while this amendment allows for casual labor, it still does not rule out the possibility of a large tied labor force. How large is large? To form an idea regarding the degree of uncertainty in labor demand, consider the village level studies conducted by the ICRISAT in India. 10 The fluctuations in the percentage use of hired labor over a period of ten years has been measured for three study villages. The gap between the highest and the lowest value is 30 percent for the most drought prone village -Shirapur. In the other two villages, the corresponding ranges are 22 percent and 25 percent respectively. These numbers suggest that the level of labor tying can climb to 70 percent without any risk of loss to the farmer. But the level of labor tying in the ICRISAT villages is nowhere near this figure. See Table 2. 9 Note that the uncertainty explanation may predict low levels of labor tying under the assumption that farmers are themselves risk-averse and face uncertain demand or supply in some other market. But in our view, it is difficult to maintain the assumption that employers are more risk-averse than workers in the particular context of our study. 10 See page 120 of Walker and Ryan (1990).

PAGE – 5 ============
A. Mukherjee, D. Ray/Journal of Development Economics 47 (1995) 207-239 211 In general, the empirical literature on rural labor markets clearly reveals that casual laborers and not tied laborers constitute the majority of agricultural laborers in Indian villages, except in the pockets where agriculture is highly mechanized. 11 Alternative explanations for labor tying stem from the existence of recruitment costs. This is a cost incurred by a farmer due to the effort involved and loss incurred due to possible delays in the process of recruiting labour. This introduces a wedge between the total wages paid by the farmer and the farmer’s total costs ~2 when hiring casual labor. These recruitment costs might induce the farmer to give up a part of such costs to tie labor. Thus recruitment cost models complement the implicit contract theory, and add to the predicted incidence of labor tying. Our paper seeks to bridge this gap between theory and observation. In doing so, we obtain a fairly complete general equilibrium model of labor tying, with other implications that are of interest in themselves. Throughout the paper, we focus on one basic form of labor tying, that which provides insurance or credit against fluctuations, and requires the carrying-out of ordinary production tasks. As noted above, the model can easily be extended to incorporate non-monitored tied labor. Our analysis is based on a natural incentive problem which labor tying creates. Typically, tied wages must be smoothed over time relative to spot market fluctuations. But then there must be periods where the tied wage falls short of the spot wage. 13 In such situations it pays the worker to break the tied contract. There is, in fact, empirical evidence that such contractual non-fulfilment is considered a distinct possibility by farmers. 14 In the case of such default we presume that the renegade is punished by contract termination. Termination of the current contract does not mean, of course, that the laborer will be unable to obtain a tied contract from some other employer in the future. It all depends on the ease with which default histories can be shared among employers. In stagnant societies, where personal histories are easy to keep track of, such information sharing is widespread. This is also true of industrialized economies where information-sharing is of a very high order, though obviously through entirely different channels. It is precisely in ‘intermediate’ societies, which are going through rapid change and growth, yet do not possess an advanced information technology, that it becomes very difficult to keep track of individual histories. Our interest is in such situations. We make, then, the extreme assumption that a renegade laborer faces exactly the same probability of reabsorption into the tied labor market as any other laborer 11 See Bhalla (1987), Bharadwaj (1974), Breman (1974), Gough (1983), Sundari (1981) and others for a discussion of the empirical evidence. 12 See Bardhan (1979). 13 See Malcolmson and Macleod (1989) and the references in this paper for a general theoretical treatment. 14 See, in this context, Section B, Chapter VI, Government of India (1960), or Binswanger et al. (1984).

PAGE – 6 ============
212 A. Mukherjee, D. Ray/Journal of Development Economics 47 (1995) 207-239 currently seeking a job. It follows that the existence of excessive labor tying jeopardizes its existence! A high probability of being reabsorbed into tied employ- ment reduces the scope of punishments that follow a deviation from the agreed terms of the contract. This feature significantly lowers the incidence of labor tying. It is possible to drop the extreme assumption that reabsorption probabilities are independent of past history. The ability to keep tabs on a worker’s history increases the proportion of tied labor (see Section 5.2 for a discussion). Indeed, this last observation and our general approach furnishes a possible explanation for the observed secular decline of the incidence of labor tying arrangements in developing countries such as India, mentioned earlier. Increased growth and change has three effects. First, it reduces the ease of informational flow as personalized ties give way to impersonal, anonymous transactions. This softens the punitive power of contract termination by increasing the chances of reabsorption. To compensate for this, the incidence of labor tying arrangements must fall. We believe that the study of information diffusion in the face of economic change has general implications for the evolution of institutions, and this is only one instance of its effects. Second, growth in economic activity increases the absolute number of all types of arrangements, tied or untied. Thus for the same percentage of tied labor, the probability of reabsorption into tied arrangements goes up. To compensate, the fraction of tied labor must decline. (Section 5.1 provides a more detailed exposi- tion.) Finally, growth might serve to tighten labor market conditions, creating larger fluctuations in the spot wages of a seasonal activity. This feature increases labor tying in equilibrium, and it is only this last feature that is present in other theoretical models (see, e.g., Bardhan (1983) and Eswaran and Kotwal (1985)). By encompassing the other features, our approach sheds some new light on the observations. We now summarize some of the specific findings. We study a full equilibrium system, not a particular employer’s decision. The model incorporates seasonality in agricultural production, but no uncertainty. We show that 1. It is possible to completely characterize economic situations where agriculture is seasonal, workers are risk averse, and employers are risk neutral, yet there is no labor tying in equilibrium (Proposition 1). 2. In contrast, the casual labor market must always be active (Proposition 2), irrespective of the degree of agricultural seasonality, or the degree of risk aversion of workers. 3. Moreover, in no equilibrium is a tied worker given full insurance, in the sense of interseasonal equality of tied wages. Despite the obvious gains from such equality, the employers will always find in their interest to pay seasonal bonuses to attached labor, though, of course, such bonuses are not as large as the seasonal wage gap. 4. Tied laborers are strictly better off than their casual counterparts. 15

PAGE – 8 ============
214 A. Mukherjee, D. Ray/Journal of Development Economics 47 (1995) 207-239 mizers 18. We idealize competition by assuming a large number of identical farmers and by postulating that the labor used by each of these farmers is infinitesimally small with respect to the aggregate. While we do not distinguish between different farmers, it will often be useful to introduce notation that separates the average economy-wide choices from the decisions made by a particular farmer. Denote by C the discounted cost of using one unit of effective labor. We emphasize right away that C will be endogenized in the sequel. Each farmer’s objective is max 8F( n) – nC. The cost of a unit of labor will depend on the type of labor used as input and the wages to each of these types. Accordingly, we now describe the types of labor contract used in this economy. 2.2. Types of labor contract There are two types of labor contracts in this economy: the casual labor contract and the tied labor contract. The casual contract is characterized by spot market hire. The laborer’s relation with the employer ends as soon as the season is over. The tied laborer is assumed to be hired for the entire agricultural year, on a contract that is commonly regarded as potentially renewable. We take it that such offers are made only at the onset of each agricultural year. There are two possibilities. A tied laborer might be required to be available for work for the landlord only in the peak season. In the slack season he gets some payment from the employer, but is free to augment this by earning in the spot market. The other possibility is that the tied worker is paid and his services used the year around. Note that the latter contract (in which work is done for the entire year) can be exactly mimicked in our model by the former contract, in which the landlord ‘additionally’ hires his tied laborer in the slack season as a casual laborer. For ease of notation, then, we stick to the first type of contract. We reiterate that ‘tied labor’ here is fully substitutable with casual labor so far as tasks are concemed. The tied contract spells out terms of payment (x,, x*) to tied laborers, in return for peak season work. The first entry represents a slack season payment; the second represents the peak season payment. 18 Slack and peak seasons are not of the same lengths. Hence the discount factors applied should vary over these two seasons. We do not take this into account for ease of exposition, but the analysis loses nothing if different discount factors were to be accommodated.

PAGE – 9 ============
A. Mukherjee, D. Ray/Journal of Development Economics 47 (1995) 207-239 215 Both laborer and employer are free to not renew the contract at the end of the year. We assume that there is always a flow of quits due to factors not explicitly modelled here. This exogenous quit rate will be denoted by q. 19 It should be observed that q is really an exogenous probability of nonrenewal of contract (even if all goes well during the contractual period), rather than a quit from an ongoing contract. Apart from exogenous quits, deliberate deviations or non-fulfilment of the contract itself is also a possibility. At this point we depart from the standard analysis of tied contracts by recognizing explicitly that an employer might not have access to any extra-economic device to ensure fulfilment of contracts. What economic devices might an employer have access to? Non-renewal of contract in future years is surely one possibility. The farmer might also refuse to pay the laborer for the season in which breach occurred. We suppose that both these instruments are available, though only the former is really important for the argument. The laborer who contemplates a breach of contract must therefore trade these future losses for current gain. We reiterate that the computation of these future losses depends on the macro-environment (including the overall prevalence of tied contracts). A contract which ensures fulfilment by the means of these economic incentives will be called an incentive-compatible contract. We now turn to a description of the characteristics of laborers. 2.3. The laborers We assume that a typical laborer is risk-averse (or more precisely, since this analysis is going to abstract from uncertainty, averse to fluctuations in income). They receive utility from income and leisure, and discount future utilities after each peak or slack season. Fluctuation-aversion is captured, as usual, by postulating a utility function of labor that is strictly concave in income earned. Workers possess one unit of leisure that is supplied indivisibly to the labor market. The utility function of income w conditional on labor supply is given by u(w); u(w) is assumed to be defined on nonnegative incomes, strictly concave and twice continuously differentiable. 20 Laborers enjoy a fixed utility u 0 from uninterrupted leisure for a season. There- 19 An exogenous quit rate is a standard postulate (see, e.g. Shapiro and Stiglitz (1984)) designed to keep some vacancies open in equilibrium for all types of jobs. These may arise due to noneconomic reasons or changing socio-economic circumstances not captured in our model. See Binswanger et al. (1984), Alexander (1973), Sundari (1981) and others for empirical evidence that contracts are frequently not renewed. It should be noted that if the entire economy is growing, the growth factor serves as an approximate substitute for such quits, keeping a pool of vacancies open in equilibrium (see Section 5.1). 20 We normalize u(0) to be 0. It is also assumed that l im ,: ~ 0 u'( w ) = ~ and lim w ~ ~ u'(w ) = 0.

PAGE – 10 ============
216 A. Mukherjee, D. Ray/Journal of Development Economics 47 (1995) 207-239 fore, to work during any season as a casual laborer, the worker must be paid at least a wage w 0, where u(w o) =- u o. So for any wage at least equal to w 0, the laborers supply one unit of labor inelastically. Each laborer’s supply of labor is infinitesimally small with respect to the whole. The total volume of labor supply forthcoming from the economy for any w > w 0 is L 0. Thus the aggregate supply curve of labor may be written as follows: t(w) =L° f°r all w > w° (2.1) = 0 otherwise. At this point we introduce a critical assumption. We postulate that laborers have no access to credit facilities that would permit the smoothing of their consumption over seasons. Indeed, if this were possible, the very motivation for labor-tying would be nonexistent. If the slack season payment may be viewed as a loan, then the potential employer is their only source of credit 21 Indeed, we will make the extreme assumption of equating the incomes of laborers in each season to their consumptions in that season. Of course, the stipulation that no smoothing is possible is an exaggeration, quite unnecessary for the formal analysis. In any case, some smoothing is always possible through savings. But there must be some imperfection in credit and savings markets for our model to have any meaning. 2.4. Availability of different jobs and lifetime utilities Consider the start of any slack season. Two types of jobs are available to a presently unemployed laborer: tied and casual. The chances of obtaining a tied job are, of course, proportional to the ratio of such vacancies to the total number of job seekers. We denote this probability by p; it will be endogenized in the sequel. We have, then, the possibilities of: 1. employment as a tied laborer with associated payments (x., x * ). Such con- tracts are available with probability p. Recall that these contracts permit the laborer to work as a casual laborer in the slack season, possibly with the very same employer providing the tied contract. 2. remaining in the spot market with probability (1 -p). This option means that the worker will receive the utility of a casual laborer over slack and peak seasons (given full casual wage flexibility). If casual wages are given by (w, w–), this utility is u(w) + B u(–w ). 21 Similar, though not identical, ideas have been put forward by Basu (1983), where no lender lends outside a pool of laborers over whom he exercises some ‘power’, and by Ray and Sengupta (1989) and Floro and Yotopoulos (1991) who point out that credit markets may be segmented according to lender and borrower characteristics.

PAGE – 11 ============
A. Mukherjee, D. Ray/Journal of Development Economics 47 (1995) 207-239 217 We reiterate that while these wages and probabilities are treated as exogenous at the moment, later they will be derived endogenously on the basis of the agents’ behaviour given the parameters described in Sections 2.1 and 2.3. let V, denote the lifetime utilities of such a laborer facing these probabilistic options and let W, denote his lifetime utility conditional on being offered a tied labor contract. Assume, without loss of generality, that u, = u(w + x, ) + 6 u( x * ) is not less than u = u(w)+ 6u(-ff), otherwise tied labor contracts will not ever be taken up 22. Now it iseasy to see that the two lifetime utilities are related in the following way: V, =pW, + (1 -p)(u + 82V, ) (2.2) and W, = u, + 82(qV, + (1 – q)W,). (2.3) Eqs. (2.2) and (2.3) are largely self-explanatory, but an additional word of explanation is offered here. The equations presume that no worker wishes to voluntarily quit his job between the slack and peak seasons. This presumption will be borne out in equilibrium, therefore no generality is lost by using this expository simplification. Let z be the vector of variables (w, w; x,, x * ; p, q) facing the agents in the economy. Given (2.2) and (2.3), it is possible to derive closed-form solutions to V, and W, as functions of z. In other words, we only need to know z in order to know the lifetime utilities. 2.5. The employer’s choice of contract An individual employer takes as given the economy-wide prevalence of wages in different contracts, as well as the employment probabilities and the quit rate. That is, he takes z = (w, w; x,, x * ; p, q) as given. The employer’s task is to hire laborers using a contract (or contracts) most advantageous to him. The reader will already see what is to be the basic component of the equilibrium notion. It is that each employer’s choice of contract, given z, must be ‘aggregated’ across employers to give rise to z itself! To work toward this, it is necessary to describe first the behaviour of employers as a response to the going vector z. Consider an individual employer in a regime where the going vector is z. Recall that each farmer is infinitesimally small with respect to the aggregate, so his choices will not affect the going vector. Suppose he is considering the offer of a tied contract to currently unemployed laborers in the slack season. Denote by (x,.~) his offer, and by W the expected lifetime utility of a laborer who accepts that offer. 22 An equivalent assumption is, of course, W, > V,.

98 KB – 33 Pages