UC Berkeley

This dissertation examines barriers to financial stability and human capital accumulation faced by low-income workers across different institutional contexts. A significant challenge that low-income workers, both in poor and rich countries, face is the prevalence of income instability--caused, for instance, by irregular labor demand, or frequent shocks to the ability to work. Chapter I examines the welfare effects of providing low-income, credit-constrained workers with access to short-term liquidity. The motivation is that consumption smoothing maybe hard for households experiencing frequent variation in income and expenses. Even when there are no changes in long-term income perspectives, they face periods in which their income is temporarily lower. In these cases, access to liquidity--by transferring resources from a state with low to one with high marginal utility of consumption--would enhance welfare. However, this is not necessarily the case if households have wrong beliefs about their liquidity needs or their income-generating process, or if they suffer from lapses of self-control problems. We partner with a company providing Earned Wage Access services (EWA), \textit{i.e.} the possibility to withdraw part of their income already earned ahead of the pay date, to study workers' demand for liquidity and whether they exhibit signs of self-control when doing so. Motivated by the evidence that workers in this setting exhibit a substantial variation in income from one pay period to another, which could potentially make the prediction of future resources and liquidity needs more difficult, we also ask whether workers mispredict their future income, and how that affects use and forecasts of demand for liquidity.\We do so through a survey and an experiment with users of our partner's EWA App. From the administrative data, we find evidence of an insurance motive for using the App: when users work less than their average in a given pay period, resulting in lower liquidity on their next pay date, they end up using EWA more in the next pay period. Building on the \citet{allcott2022high} methodology, we then document that users under-predict their demand for EWA (future liquidity) by a substantial amount, and they are willing to pay for an incentive to reduce their future use, consistent with agents' being partially sophisticated about their present focus. We also find that users' demand for incentives to reduce future use depends on their current financial circumstances--when users make predictions in a state with temporarily lower income, they demand less commitment and they under-predict their use by a smaller amount. We interpret this as evidence that users correctly interpret the trade-off between flexibility and commitment that EWA provides. It also implies that the welfare effect of restrictions depends on the prevalence and utility costs of temporarily lower income. We also find a role for over-prediction in income affecting the misprediction in EWA use.

Another consequence of volatile work availability --ubiquitous in low-income countries where many workers are employed in seasonal occupations or in informal labor markets-- is that it prevents long-term skill accumulation. We study this phenomenon in Chapter II with respect to a particular skill: the ability to supply labor reliably. We posit that labor supply is not a function of stable preferences for leisure, but rather is also determined by one's past habituation to work. In existing data, we show that exogenously induced transitory changes in labor supply increase supply in subsequent days---indicating that the inter-temporal labor supply elasticity can actually be positive, rather than negative. To examine this phenomenon in more detail, we undertake a field experiment with casual urban stand workers in Chennai, India, where appearance at the stand in the morning provides a revealed preference measure of labor supply. We randomly provide some workers incentives for attending the stand in the morning in a timely manner for two months (phase 1), and examine persistence after incentives are removed for another two months (phase 2). We find that an increase in labor supply in phase 1 generates a persistent 16\% increase in supply in phase 2---leading to a 22\% increase in employment found at the stand. These findings have relevance for understanding the reasons for irregular work attendance and high worker turnover in formal firms, which impede the transition to formal work in this setting. They also suggest that the effects of unemployment spells may go beyond the income loss: unemployment itself can lower a worker's productivity--offering a potential justification for the ``unemployment scar'' phenomenon discussed extensively in the labor literature, whereby employers prefer not to hire workers out of unemployment.

In Chapter III, we examine the aspect of human capital accumulation of low-income workers from another perspective, namely employers' incentive to train workers in general skills. There is a long-lasting theory in Economics that general skills may be under-provided in equilibrium due to a failure of those who train to sufficiently appropriate the returns of training \citep{Becker2009}. We provide evidence for this theory in the context of new agricultural technologies in Sub-Saharan Africa. We develop a test tightly linked to theory to assess whether concerns over the appropriability of returns to training generate a socially inefficient level of training, limiting the adoption of new technologies. We conduct two exercises to answer this question. First, we document that the level of training may be socially inefficient. We show that inducing employers to train laborers generates externalities, specifically whether employers uninvolved in training subsequently hired these trained workers and adopted more improved technology in their fields. We find evidence for such an externality: in villages with trained workers, not only the trainers but also other farmers uninvolved in training hire more skilled labor. We then test whether allowing farmers to appropriate the returns of this training increases training in equilibrium. In a second treatment arm, we randomize farmers to receive a contract that guarantees them the appropriability of returns from their training. Offering labor insurance to the employer- trainers increase the propensity to train substantially. Moreover, this training is consequential and does not appear to be just driven by demand effects. Overall, we find that, in both arms, adoption of the technology increases, suggesting that lack of skilled labor can be a constraint to technology take-up. We also find substantial benefits accruing to trained laborers, both in terms of labor market earnings and profits on their own land. This suggests that under-provision of training can be consequential for skill growth and earning trajectories of poor workers.