UC Berkeley

Perhaps the most central insight of development economics is that, absent a complete set of perfect markets, households' economic activities cannot be neatly ``separated" into those of a profit-maximizing firm and a utility-maximizing consumer (Singh et al., 1986) In particular, risk-averse farm households face a tradeoff between maximizing farm profits and smoothing consumption across states of the world. Balancing these motives is important not only for these households, who constitute a massive share of the world's poor, but for aggregate productivity as well. However, little is known about how to diagnose the market failures that create these tradeoffs, quantify their costs, and prescribe robust policies to address them. This dissertation seeks to provide methodological and empirical progress from the micro to the macro levels.

Chapter 1 focuses on identifying how distinct market failures affect aggregate productivity in Thai agriculture. Agricultural markets often fail to allocate resources efficiently across farm households in developing countries. However, policymakers require knowledge of which markets fail and how the distortions they generate are correlated. In this chapter, I use data from rural Thailand to characterize how distortions in land, labor, credit, and insurance markets each contribute to factor misallocation. I use moments in household consumption and production data to separately identify these distortions and then quantify their impacts on aggregate productivity through an equilibrium model of misallocation. I find that the efficient allocation would increase aggregate productivity by 31% relative to the status quo, while only 15% (7%) gains could be achieved by eliminating financial (input) distortions in isolation. Positive interaction effects from addressing multiple distortions simultaneously account for the remaining 9% TFP gains. Meanwhile, other common methods would produce larger estimates of misallocation and suggest that a financial market intervention would decrease aggregate productivity. Accounting for multiple correlated distortions is therefore crucial for measuring misallocation and designing policies to address it.

In Chapter 2, coauthored with Ethan Ligon, we move from Thailand to Northeastern Nigeria and move from the growing season to the lean season spanning harvests to study another important tradeoff between consumption smoothing and investment. In particular, we conduct a randomized control trial offering postharvest loans (PHLs) to farm households in Gombe State. The purpose of these loans is to enable households to shift from exhausting grain stocks and buying them back at high prices to becoming net arbitrageurs. While such programs have increased household incomes in Kenya (Burke et al., 2019)and Tanzania (Channa et al, 2022), their theory of change relies on grain prices rising, which is a highly uncertain proposition across sub-Saharan Africa. During our study period, prices of maize and other major crops stayed flat. While we find that the loans induced households to store more crops later into the season, we do not find significant effects on sales or overall welfare. While this is an example of the downside risk of PHLs being realized, we also use a simple model of intertemporal arbitrage to show how ex ante risk can have ambiguous effects on the demand for PHLs, depending on whether households are more vulnerable in states with high vs. low prices.

Chapter 3, based in part on work coauthored with Ethan Ligon, focuses on production function estimation when input choices are distorted. These estimators, which are used to estimate the production function in Chapter 1, extend the canonical approach in industrial organization (Ackerberg et al., 2015; Gandhi et al., 2020), to risk-averse producers facing imperfect markets. In particular, they proxy for unobserved productivity by inverting the demand function for a flexible input from the setting with profit-maximizing firms in competitive markets to risk-averse households, possibly facing distorted input markets. The method involves combining consumption and production data to model input demands as a function of unobserved productivity and a stochastic discount factor, which includes the covariance between production shocks and consumption at harvest. Essentially, the consumption side of the household's problem provides information to help us identify the production side. Three main specifications are considered: the canonical Cobb-Douglas with Hicks-neutral shocks, a heteroskedastic generalization of Cobb-Douglas that allows for differentially risky inputs, and a dynamic multi-stage Cobb-Douglas featuring sequential shocks. The differences across specifications show the importance of accounting for risk, both overall and input- and stage-specific, to consistently estimate production functions and draw inferences about efficiency and misallocation.

Together, these three chapters show how better understanding households' tradeoffs between productivity and consumption smoothing can improve policies to address both micro-level food insecurity and macro-level productivity.