Poverty, Occupational Choice and Social Networks:

In rural labor markets, a tied-labor contract involves a long-term relationship between an employer and a worker where the employer provides a steady but low wage to the worker (relative to a casual labor contract that offers a high wage rate during the harvest season). The role of labor-tying on terms of labor contracts has been stud- ied extensively in theoretical studies (Bardhan (1983), Eswaran and Kotwal (1985), Mukherjee and Ray (1995)) and the empirical relevance of tied-labor has been shown, particularly in South Asia1 (Bardhan and Rudra (1978)). In developing countries where poor households face substantial amounts of risk and limited insurance opportunities, labor-tying is likely to be an important channel through which they smooth their in- come, hence their consumption2 (Morduch (1995)). Yet recent empirical studies have mainly focused on other mechanisms of consumption-smoothing such as informal in- surance and pre-cautionary savings3. Using original survey data from Bangladesh, I show that labor-tying is an important mechanism through which poor workers smooth their consumption. Furthermore, I test the effects of an experiment that increases the expected income of the poor women living in rural Bangladesh on their involvement in tied labor. In particular, I show that an exogenous improvement in the outside option of poor workers decreases their participation in tied-labor, and allows them to enter labor contracts with higher return but higher income volatility. This change in the level and composition of labor supply within the village has different general equilibrium effects on the returns to tied and casual labor in the male and female labor markets within the village. Finally, I provide evidence that suggests that the treated poor households are changing the mechanisms through which they smooth their con- sumption. In particular, the households that are exogenously made wealthier are less likely to engage in tied-labor arrangements, but more likely to form reciprocal transfer links with other villagers. Taken all together, the findings show that as poor households (exogenously) get richer, they move from second-best labor contracts (that yield a low return but insure them against risks) to more profitable yet riskier income generating activities, accompanied with reciprocal transfer arrangements that help smooth their consumption.

In order to formalize the incentives of workers and employers in entering tied-labor arrangements, I adopt the risk-sharing model of labor-tying developed by Bardhan (1983) where a risk-averse worker enters into a tied labor arrangement with a risk- neutral employer in order to smooth her income during the lean and peak seasons. Alternatively, the worker can choose to settle down for her expected outside option, which will be a function of her wealth and vulnerability (proneness to risks). The model assumes that tied workers and casual laborers are perfect substitutes in the farm production function during the peak season. Hence, the employer’s only incentive in offering tied-labor contracts is to ensure supply of cheap labor during the peak season. In equilibrium, it will be the poorest and most vulnerable workers that enter into tied- labor contracts, while better-off workers will choose to remain self-employed and work for the employer as a casual worker whenever the realized village market wage rate exceeds their expected outside option. This automatically implies that casual workers will receive a higher wage rate on average.

I use this theoretical framework to test the effects of an exogenous increase in the outside option of poorest workers on their participation in tied labor and on the terms of labor contracts in the village economy. The exogenous variation I exploit is the randomized roll-out of the “ultra poor” program in Bangladesh. The “ultra poor” program was pioneered by BRAC4 and targets the poorest women living in villages. It involves a combination of a large asset transfer (livestock or trees), enterprize training and weekly visits by program officers to ensure that the treated females are able to generate income from the assets that they receive. In short, the program improves the self-employment opportunities of treated women. The data used in this paper comes from the randomized evaluation of BRAC’s ultra poor program in Bangladesh. The program identifies the poorest females living in rural villages, who are often landless laborers. They rely primarily on finding work as agricultural day-laborers or maids, and on the transfers they receive from the rest of the community. This is a setting where seasonal fluctuations in wage earnings are very significant (see Figure 1) and a large proportion of the targeted poor households enter into tied-labor contracts that provide a smoother income profile but lower average wage.

The theoretical model gives the following predictions with respect to an exogenous shock to the outside options of the poorest workers in the economy:

1. In partial equilibrium (assuming there is no effect on the returns to tied or casual labor)
   
   1. (a)  Treated workers will be less likely to be working for a wage. This depends on two factors: (i) whether the amount of increase in the outside option of the treated worker is large enough (ii) the initial level of the outside option of the worker.
      
   2. (b)  Conditional on remaining in wage-employment, treated workers will be less likely to be in tied-labor contracts and more likely to be in casual labor contracts.
      
2. In general equilibrium, depending on how the program affects the aggregate dis- tribution of workers’ outside options, wages for both tied and casual laborers may increase. In that case, the threshold level of outside option below which workers enter into tied contracts also increases.
   
3. A corollary of prediction (2) is that the effect of the program on whether treated workers remain in wage-work and the type of contracts they enter will be am- biguous in general equilibrium. The direct effect on their outside options and the GE effects through the labor market have opposing effects on their labor market participation.
   
4. Finally, if workers are matched assortatively by their outside options in reciprocal transfer arrangements, then treated workers will be more likely to enter reciprocal arrangements with wealthier workers to smooth their consumption. This will increase their likelihood to switch from tied to casual labor contracts.
   

In order to test the predictions of the model empirically, I make use of two key characteristics of the evaluation strategy: First, in order to identify tied and casual workers empirically, I use data on the identity of workers’ employers and their food transfer links. The data is unique in the sense that for every business activity that the respondents were engaged in, they were asked to report the identity of their employer and as long as the employer was within the same village (as the respondent), their household ID number was recorded. Similarly, respondents were asked to identify the most important 3 households they would borrow food from at times of need. Using these two pieces of information, I can identify which employers were also a borrowing source for the worker: 25% of the poor workers report their employer as a source of food transfers in times of need. I show that this definition of tied labor contracts also correlates with having lower average wage rate and lower wage volatility, in line with the definition of tied labor contracts in the theoretical framework5.

Second, in order to identify the direct effects of the program on the treated house- holds and the indirect spillover effects on non-treated households via the labor market, I make use of the fact that the program was randomized at the village level and the sample includes both treated and non-treated workers in treatment and control villages. Comparison of treated workers in treatment villages to those workers that were selected for treatment but were not treated in control villages (henceforth “selected workers”) allows me to identify the direct effect of the program combined with any indirect general equilibrium effects. By comparing the non-treated workers in treatment villages to the relevant group of workers in control villages, I identify the general equilibrium effects of the program on the rest of the community.

I start by analyzing the effects of the program on the treated women. I find that the program has a negative impact on the participation of treated females in the female labor market in the village. They are 10% less likely to be working for another household in the village at followup relative to eligible women in control villages. This suggests that there is an overall fall in the labor supply in the female labor market in the village. In line with prediction 2, conditional on being in wage employment, treated females are 20% less likely to be in tied-labor contracts. Hence there is a greater fall in the supply of tied female workers relative to casual workers. Furthermore, this suggests that the direct effect of the program on the outside options of treated women dominates any indirect GE effects through the labor markets.