UC Berkeley

The central theme of this dissertation is applying causal inference methods to the analysis of issues of policy importance in labor and development economics, using knowledge in econometrics and quantitative economics learned throughout the PhD. Recent advances in causal inference---rigorously studying how changes in certain factors affect the outcome---have greatly enriched econometrics and economics. In this work I apply several methods to study policy questions that affect businesses, workers, and consumers worldwide across a variety of contexts to demonstrate their applications to real-world economics issues.Chapter 1 of this dissertation applies a shift-share instrumental variables (SSIV) approach to study the effect of labor demand shocks on local consumer prices in Mexico. Policymakers have been interested in how local labor demand shocks affect welfare. While many studies have examined the effects of local shocks on nominal income, much less attention has been paid to prices, the denominator of real income. Using data on the universe of CPI price quotes, workers, and firms in Mexico, I study the effects of local labor demand shocks on local consumer prices. To base the estimation on plausibly exogenous variation, I propose an Instrumental Variables strategy that leverages national industry-level changes in labor demand in combination with pre-existing employment shares. I find that a 1 percentage point increase in the growth rate of local labor demand leads to a 0.061 percentage point increase in services inflation, but a 0.099 percentage point decrease in goods inflation. The effects are concentrated in products that exhibit larger geographic price variation and are more subject to local markups. The decrease in product prices is consistent with increased product entry as firms introduce new varieties in locations with increasing market size, while services are more exposed to increases in local wages. To guide the interpretation of empirical results, I develop a general equilibrium model that features variable markups and endogenous firm entry. Comparative statics show that an increase in market size driven by local labor demand shocks leads to relatively lower markups and prices on continuing varieties. The overall effect on non-housing inflation is close to zero as lower inflation in tradables cancels out higher inflation in services, but poorer households see their price indices fall more than richer households, who spend more on services.Chapter 2 of this dissertation, joint with Uyanga Bambaa, Edward Miguel, and Michael Walker, estimates the gender wage gap and examines the underlying factors that cause the gap in Kenya. The gender wage gap remains persistent across the world, especially in low- and middle-income countries. We estimate the magnitude of the gender wage gap and contributors to it in Kenya utilizing a rich dataset from the Kenya Life Panel Survey (KLPS) covering 5,878 adults. The data measure typically unobservable individual characteristics that could potentially determine wages, such as cognition, personality traits, job task indices, and economic preferences, allowing for estimation of the gender wage gap among prime-age adult workers in Kenya controlling for extensive covariates. We use both regression and causal machine learning methods to estimate the gender wage gap. We find that women earn 78 log points less than men (54%) without adjustments, and 41 log points less (34%) after controlling for education, experience, demographics, occupation, cognition, personality, preferences, and job task indices. The latter four account for 18% of the residual gap unexplained by education and occupation. The large gender wage gap after accounting for a rich set of potential confounders potentially suggests the role of discrimination and social norms in hindering women's access to opportunities in the labor market.Chapter 3 of this dissertation, joint with Matthew Grant and Meredith Startz, combines causal inference with structural modeling in a setting where rigorous identification is challenging to yield insights on the gains from firm consolidation among small retailers in Nigeria. We ask: Why are consumer goods in low-income countries often sold in a physical market area with many small firms side-by-side, rather than by a large, integrated retailer as is more common in high-income countries? To what extent does this matter for prices? We begin by documenting a set of novel empirical patterns about markets in Lagos, Nigeria, using original survey data among 1,500 firms. We show that wholesale/retail firms incur large fixed costs when sourcing goods for resale from distant suppliers; as a result, they often source from very nearby suppliers within their own market. We also document that labor and supervision costs are increasing in the number of employees, consistent with span of control or agency problems. We build a structural model in which these countervailing economies and diseconomies of scale lead to a large number of co-located, independent wholesalers and retailers who transact with one another but maintain boundaries between firms as the equilibrium market structure. Compared to standard models accounting for the firm size distribution in developing countries, this leads to substantively different conclusions about the welfare costs of small firms. Quantifying this comparison is still in progress, but the qualitative insight is that small firms that co-locate and source within a market capture some of the gains from scale in sourcing while aligning labor incentives. In essence, there is less gain to be had from turning micro-entrepreneurs in the retail sector in a Nigerian market into cashiers in a large firm than might be expected.