Research

Prices and Price Limits (with Jonathan Brogaard)

Abstract: This paper studies the effects of price limits implemented by the Securities and Exchange Commission (SEC) after the May 2010 ‘Flash Crash.’ The security-level price limits halt trading after a security’s price experiences a sudden and large movement. The difference-in-difference design exploits the staggered introduction of the limits to address omitted variable concerns. The data show that price limits reduce the frequency and severity of extreme price movements, but induce price underreaction. The results are consistent with Subrahmanyam’s (1997) theory that price limits cause informed traders to be less aggressive.

Information Acquisition and Short Selling (with Kaitao Lin)

Abstract: Sophisticated traders, like short sellers, may deter information acquisition by less sophisticated traders. We measure information acquisition as the contribution of earnings announcements to volatility—when earnings announcements account for a greater share of volatility, then less of the information is priced prior to announcements. We use the SEC’s short sale pilot program and find that information acquisition falls when short selling is easier and the effect reverses after the pilot. We find symmetric results for good and bad news, consistent with information acquisition as opposed to a direct effect of short-selling activity. Further, we study the mechanism by combining the pilot program with the discontinuity in institutional ownership around the Russell 1000/2000 threshold. Pilot (but not control) stocks suffer particularly large losses in information acquisition where institutional ownership increases around the threshold, consistent with a complementarity between lendable shares and easier short selling rules. Our results suggest that short sellers, while informed themselves, can decrease price efficiency by deterring the information acquisition of others.

Mortgage Restructurings and Regulatory Arbitrage: Evidence from the JOBS Act

Abstract: Banks may be unwilling to restructure troubled mortgages after a crisis because such restructurings immediately reduce earnings and regulatory capital. To test this, I use a provision of the Jumpstart Our Business Startups (JOBS) Act of 2012 that allows some banks to deregister from the Securities and Exchange Commission (SEC). In the first stage, I show that banks deregister primarily to reduce accounting oversight. Further, deregistration leads to lower audit fees and provisions for loan losses, which helps earnings. Finally, deregistration causes banks to double their troubled mortgage restructurings. These results suggest that regulatory constraints may hamper restructurings.

The Heterogeneous Response to Homebuyer Credits (with Jonathan Brogaard)

Abstract: In an attempt to stimulate the housing markets from 2008-2010, the US instituted tax credits of up to $8,000 for homebuyers. We use within-locale variation to document substantial heterogeneity in the effects of these credits. In particular, the market for three-bedroom homes experiences an increase in sales of 1.3% relative the market for four-bedroom homes; conversely, the market for four-bedroom homes experiences an increase in price of about $2,000 relative prices for three bedroom homes. Both effects reverse after the credit expires, and are lower bounds on the total effect of the credits. Finally, the marginal buyers can be slightly unprepared—they are more likely to miss a payment, and are less likely to refinance when rates fall. Our results suggest that the relative elasticities of supply and demand are very different within the same metro area, with participants in expensive markets less likely to change their housing plan.

The Enduring Effects of Interest Rates at Mortgage Origination (with Don Carmichael and Dimuthu Ratnadiwakara) – Semifinalist for best paper in the financial institutions category at the 2017 FMA meeting

Abstract: Modest differences in the interest rate at home purchase can have long-lasting effects on mortgagors. We use within-year variation in average interest rates at loan origination to instrument for contracted mortgage rates. For homeowners with negative equity, a 50bp increase in the national rate at origination leads to an increase in defaults of 10-20% of the sample average, but the instrument is not correlated with worse credit. Consistent with liquidity constraints, the magnitude is constant across many different levels of negative equity. During the boom, lower interest rates result in increased consumption of non-durables and services, while total expenditure is unchanged.

An Introduction to Nonprofits’ Electronic 990 Filings

The Internal Revenue System (IRS) has relatively recently begun disclosing all electronic nonprofit 990 filings in a machine-readable format. This paper provides an overview of the data. I first describe the nontrivial process of organizing these electronic filings into a useable flat-file database; the database has analogs to Compustat, Boardex, and Execucomp. Second, I tabulate the electronic filers along key dimensions: the company’s subsection code, its organizational structure (e.g. operating nonprofits vs. private foundations), and its mission. Third, I evaluate potential selection problems in the data, and propose solutions. Finally, I outline a research agenda that can be studied with this new data. Of particular interest are 38 million person-year-firm compensation records, and 12 million grant records.

Technology Undermining Education: The Case of Chegg

Chegg is an education technology company with a market capitalization greater than traditional textbook publishers. Its main product, Chegg Study, allows students to obtain homework solutions from ``expert” contractors. I track the effects of Chegg over a five-year period for a large finance course of upperclassmen. I find that 25% of students—including 15% of high-scoring students—use Chegg blindly, copying obviously wrong answers. To study the causal effect of Chegg, I use variation in takedown requests for copyright violations as a quasi-experiment. These takedown requests remove answers from the website, making Chegg less tempting. I find that high-scoring students do particularly well on subsequent exams after the temptation to shirk on homework is reduced.