Working Papers

In low- and middle-income countries, internal labor migration remains widespread even when destination markets have high unemployment among local workers. A common view is that firms hire migrants because their likelihood of accepting a lower posted wage is higher than local workers. We test an alternative demand-side explanation using a field experiment in which firms receive offers to hire either local or migrant workers. Firms’ demand for migrant workers is 81.67% higher than for locals at the common market wage, and firms are willing to pay a wage premium of 12.3% on average to hire a migrant rather than a local worker. To understand why, we experimentally vary migrants’ residence status during the job. When firms do not have the flexibility to have workers reside on site, the migrant–local hiring gap shrinks. Higher demand for onsite residence is consistent with two mechanisms: lower absenteeism risk when workers stay on site and greater scope to extract working hours. An incentivized belief elicitation indicates that firms believe migrant workers are more likely to work longer hours for the same pay, less likely to be absent, and more likely to be socially and economically vulnerable than local workers and these beliefs are stronger for migrant workers living on site. We then assess whether migrants anticipate these conditions or instead hold systematically biased beliefs, using incentivized belief elicitation and a contract-variation experiment that reduces vulnerability. In the belief-elicitation exercise, both migrant and local workers report expectations about hours, absence, and vulnerability that closely track firms’ beliefs. In the complementary experiment, workers exhibit higher willingness to accept lower wages under contracts that mitigate these vulnerabilities.

Informal labor markets in low- and middle-income countries are characterized by high unemployment and job turnover, despite high labor demand. I conduct three field experiments to examine how frictions arising from wage theft, worker reneging and liquidity constraints lead to a mismatch in workers' labor supply and firms' labor demand in these markets. In the first experiment, 1,360 workers receive real job offers on either flat contracts (paid daily), which reduce wage theft concerns and worker liquidity constraints, or back-loaded contracts (paid at end of contract), which don't. To isolate the effects of wage theft concerns I cross-randomize these contracts with an insurance contract. I find that labor supply is three times larger for flat contracts. This preference is driven by workers' fears of wage theft by firms (28\%), liquidity constraints (22\%), and demand for flexibility to break contracts (50\%). This flexibility captures an option value, protecting workers from income loss in the event of future shocks or excess work exaction by firms under back-loaded contracts. The costs of these frictions are high—72\% of workers who reject job offers end up earning less income than those offers would have provided. In the second experiment, I make real offers to hire workers on either flat or back-loaded contracts to 349 firms. The firms' labor demand doubles for back-loaded contracts, driven by liquidity constraints (37\%) and costs associated with worker reneging (63\%) in flat contracts. In a third experiment, firms and workers who accepted contracts in the first two experiments are matched with each other. Workers are 69\% more likely to complete back-loaded contracts compared to flat contracts, which have a completion rate of 34\% and thereby impose significant costs on firms. Firms exact longer working hours from workers in back-loaded contracts, validating the workers' concerns. My findings suggest that improving contract enforcement and providing credit to firms can reduce welfare losses in equilibrium and improve efficiency.

Millions of workers in low- and middle-income countries seek low-skilled employment through informal spot markets in urban areas. Using detailed micro-data from spot markets in India, this paper examines firm–worker wage negotiations and the determinants of undercutting behavior. Despite high unemployment, workers undercut each other in only 27% of negotiations. The likelihood of undercutting varies systematically with contract type, which is strongly correlated with job type: multi-task jobs are predominantly offered under fixed-hour contracts, while single-task jobs are typically offered under  Fixed Work contracts, where pay is based on task completion regardless of hours worked. Workers are nearly twice as likely to undercut in negotiations for Fixed Work contracts compared to Fixed Hour contracts. We explain this pattern using a model in which workers’ reservation wages depend on the value of leisure, with effort choices, shaped by ability, determining attainable leisure time. This mechanism generates greater dispersion in reservation wages for contracts without fixed hours. The model predicts, and the data confirm, that undercutting is more likely for jobs in which the reservation wage distribution is more dispersed, as measured by the variance. These findings shed light on wage-setting mechanisms in low-wage labor markets and show how job characteristics and workers’ belief structures jointly shape negotiation outcomes.

We study how schools embed themselves in communities and, in doing so, function as instruments of social engineering that influence political and cultural change. Our focus is on schools run by the Rashtriya Swayamsevak Sangh (RSS), a Hindu nationalist volunteer organization. We compile new data on the establishment years and locations of these schools and link it to constituency-level electoral outcomes for the Bharatiya Janata Party (BJP), India’s dominant right-wing party. Using a staggered difference-in-differences design, we  find that the presence of RSS schools leads to a persistent rise in BJP vote share and a higher probability of BJP victories in subsequent elections. These schools also lead to cultural change. Vegetarianism rises in nearby villages, especially among upper-caste households, the traditional supporters of the BJP. The share of children given Hindu deity names, particularly those of the Vaishnavite pantheon, increases in affected areas after the setting up of schools. Together, these patterns point to a process of gradual identity activation through sustained community engagement. To examine the mechanisms in greater detail, we survey a random subsample of RSS-run and non-RSS private schools across four districts in Bihar. The survey results indicate that, relative to close-by non-RSS private schools, RSS schools are substantially more likely to engage in Hindu religious programming  as well as community engagement that involves not only students but also parents and local communities.

Sand, the backbone of modern construction, makes up 85% of global mineral extraction. Its large-scale removal—especially from fragile river ecosystems—has raised serious environmental concerns, particularly in developing countries experiencing construction booms and lacking strong environmental regulation. Yet, the full extent of these effects, their social and economic consequences, and how they are distributed across communities remain poorly understood. This project combines novel administrative data with satellite imagery in a difference-in-differences framework to estimate the causal impacts of sand mining in Bihar, India. We find that sand mining increases flood inundation rates and flood frequency by 1.5 and 2 percentage points, respectively, from a baseline of 10%. Agricultural output, measured by vegetation index, declines by 5%. Despite these environmental costs, sand mining raises local income—proxied by nightlight intensity—by 10–20% in nearby villages. Income gains are broadly distributed within villages, reaching both ex-ante wealthier and poorer households, but are concentrated among socially dominant groups, particularly upper castes and OBCs. These findings help explain why sand mining continues with local support despite environmental harm and regulatory efforts. The unequal balance of economic benefits on social groups highlights the need for targeted policies—those that mitigate flooding, ensure sustainable mining, and compensate marginalized communities bearing the brunt of net harm.

We study the distributional consequences of colonial institutions by analyzing the effect of direct and indirect rule on the outcomes of historically disadvantaged and advantaged caste groups across the border of Hyderabad, the largest princely state in colonial India, using a border discontinuity design. We find that disadvantaged caste groups in indirectly ruled areas have 1.4 fewer years of schooling and significantly less wealth than similar groups in direct rule areas. A similar analysis for Advantaged caste suggests that they have less wealth and small but insignificant differences in education. Cross-caste wealth inequality and land inequality are higher in indirectly ruled areas. Differences in extent of agricultural labor repression, as measured by agricultural wages and bonded labor, led to the emergence of these differences during the colonial era. Differential growth in credit institutions after independence negatively affected productivity in indirectly ruled areas. As a result, farmers in these areas have larger non-institutional debt than directly ruled areas, and disadvantaged caste farmers drive this effect.

With increasing urbanization, road transport has become a major contributor to pollution in many low-income countries, where privately provided public transit remains a common mode of travel. In such settings, affordable electric public transit vehicles offer a potential pathway to reducing pollution. We study the introduction of Electric rickshaws (ERs), a low-cost, slower alternative to combustion based private transit, in Indian cities, where they now dominate the last-mile connectivity market. Compared to their closest substitutes, such as auto-rickshaws, ERs are cheaper but only half as fast. On the narrow roads of Indian cities, they can slow down other vehicles, thereby generating congestion and pollution externalities. To analyze these effects, we build a city-level model in which, in addition to the price of ERs, city characteristics such as population density, income, commuting distance, and the relative speed of ER determine demand and, in turn, the rate at which they are adopted by private providers of public transit. The social planner faces a clear trade-off: improving technology raises ER speed, which reduces congestion and improves welfare, but also increases prices which lowers adoption rates. The model shows (and the data confirms) that, in equilibrium, ER adoption at the city level increases with population density, decreases with average speed, and decreases with average commuting distance. To evaluate the pollution externalities of ER adoption, we analyze district-level pollution using a staggered adoption design. Analysis indicates that NO2 levels rise after adoption in districts with high uptake, consistent with congestion effects outweighing local emissions benefits from electric vehicles. Combining this elasticity with the model, we show that the planner can improve technology to increase ER speed, which mitigates congestion externality, even though higher costs reduce adoption and supply. The net result might be fewer but faster ERs on the road, leading to lower congestion and, in turn, lower pollution.