UC Santa Cruz

This dissertation studies issues related to public economics, lending markets, real estate, and wealth inequality. The first chapter examines the distortionary effects of federal mortgage repurchasing on lending patterns. Fannie Mae and Freddie Mac are restricted by law to purchasing loans with origination balances below a county-specific "conforming loan limit" (CLL). I examine the behavior of borrowers near the temporarily increased CLLs after the global financial crisis to understand the impact of mortgage repurchasing on lending patterns. I find a sharp bunching of the fraction of loans originated at the loan limit. Borrowers with bad credit histories disproportionately select into the program by offering a large enough down payment to ensure the loan size falls just below the cutoff. The default rate over the medium run of mortgages barely below the CLL is 2.17 percent higher than those barely above the CLL. This impact is largely driven by buyers with a previous home loan. Finally, to understand the distributional impacts of the federal home loan program, I examine differential sorting across the CLL by demographic and income characteristics. I find no evidence that racial minorities or people from "poorer zip codes" comprise a disproportionate share of those who manipulate their loan size to take advantage of the program.

The second chapter provides a test between a negative and positive selection model at a micro-level. The Stiglitz-Weiss model and the de Meza-Webb model make opposing predictions of the correlation between interest rate and default in the credit market. I employ loan-level data from a peer-to-peer online lending marketplace, Prosper, to investigate whether lowering interest rates improves or worsens the mix of applicants and repayment. My empirical result is generally consistent with the prediction of the de Meza-Webb model. I find that even after controlling for all observable characteristics, the pool of borrowers is still affected by some selections on unobservable information. The default rate statistically significantly increases as the interest rate drops for lower-rated borrowers.

The third chapter is a joint project with Nirvikar Singh and Anirban Sanyal. This chapter analyzes and quantifies how differences in the wealth levels of Black and White Americans relate to socioeconomic characteristics, including education, occupation, asset portfolio structures, inheritance and financial literacy, using data from the 2016 Survey of Consumer Finances. Some combination of inheritance, education, and occupation is significantly related to differences in wealth levels across races. However, education, homeownership, business ownership, and financial literacy are not, by themselves, pathways even to reducing wealth gaps, let alone eliminating them. Much of the wealth gap is related to unmeasured structural or systemic factors, rather than measured characteristics: this is estimated by a decomposition of overall wealth differences into those associated with characteristics or endowments and those with differential impacts across groups. Some of the empirical approaches in the estimates are relatively novel in the context of quantifying individual and systemic contributors to the racial wealth gap. Additionally, quantile regressions, which allow for different impacts of characteristics at different portions of the wealth distribution, enable some inferences about the role of class vs. race. The results reinforce the view that race matters for the wealth gap even after accounting for class.