ESSAYS ON APPLIED CAUSAL INFERENCE IN BUSINESS SETTINGS: EVENT STUDIES, BUNCHING, AND RANDOMIZED CONTROL TRIALS

Abstract

I apply both reduced-form and structural methods for the causal estimation of parametersof interest in disparate business settings. Chapter 1 uses an event study design to examine whether rms benet from donation to United States politicians. Firms that contributed to a congressperson experienced negative returns following news of a connected politician's scandal. This eect exists for politicians representing swing districts, where scandals often foreshadow changes in party representation, but not in safe districts. Additionally, the return shock increases in the degree of policy alignment between rm and politician.Together, this suggests that contributions do not cause favoritism, but arise because rms contribute when candidates' viewpoints match their own. Chapter 2 expands structural bunching estimators to a two-dimensional setting to evaluate the eects of government interventions in small business lending. Using data from the Small Business Administration's (SBA) Express Loan Program, we estimate a tractable model of bank competition with endogenous interest rates, loan size, and take-up. We introduce a novel methodology that exploits loan bunching in the two-dimensional contract space of loan size and interest rates, utilizing a discontinuity in the SBA's interest rate cap. In concentrated markets, we nd that a modest decrease in the cap would increase borrower surplus by up to 1.5%, despite the rationing of some loans. Chapter 3 utilizes a randomized control trial in partnership with an online retailer to evaluate the consequences of price promotions. We nd that discount-eligible customers made more purchases and accounted for more revenue and prots during the six-week period of coupon eligibility. However, the control group made substantially more purchases at full price in the period following discount eligibility. The net eect was no signicant dierence in order frequency or prot from the two groups in our experiment. We conduct an incentivized survey of customers to evaluate the possible mechanisms driving this result.