Working papers

Dissecting Racial Politicization: Long-Run Evidence from the Food Stamp Program

(with Carlos Avenancio-Leon and Troup Howard)   SSRN link    Abstract
Many public policies – such as those on immigration, welfare, and policing – consistently attract partisan political attention, often with a racial dimension. How a public policy becomes politicized along racial lines is the focus of this paper. We develop a framework in which political parties gain electoral advantage by framing policy in political terms. This shows that an ex-ante group-neutral policy can generate political polarization across different voter groups (e.g., by race), that polarization is larger for cohorts learning about the policy at its onset, and that polarization persists over time. We apply this framework to study the politicization of the Food Stamp program. Using voter roll data for the entire U.S., we show empirically that the introduction of the program increased political polarization across racial groups, that this racial polarization was larger for voters that experienced the FS rollout at its onset, and that this polarization persists today, about a half-century later. More specifically, we show that individuals of voting age at the time of the program’s rollout (1961–1975) diverge along racial lines in their likelihood of voting and registering as Republicans or Democrats, with this divergence decreasing among younger cohorts. Our design ensures that these findings are not driven by geographic or age-specific racial trends. We also explore contemporaneous effects and additional contributing factors. First, we show that access to the safety net also had short-run effects on voters’ beliefs and turnout, as well as on the ideological composition of Congress. Second, we explore the interaction between Food Stamps and contemporaneous events such as the Voting Rights Act and recessions.

Celebrity Persuasion

(with Matteo Benetton, Marina Niessner and Jan Toczynski)    Abstract
Celebrities have long leveraged their influence to shape outcomes in politics, marketing, and now in cryptocurrency markets. As investors increasingly rely on social media for financial news and investment guidance, celebrities are playing a larger role in the landscape of financial products. Using survey, market, and transaction-level data, we examine the persuasion rates of celebrity cryptocurrency endorsements on Twitter. Investors appear to treat these celebrity tweets as financial advice: controlling for crypto-related news, the probability of cryptocurrency investment by individuals on tweet days increases by 11%, with stronger effects among men, wealthier individuals, and older investors. We find that celebrities have relatively high persuasion rates and they impact equilibrium outcomes -- market trading volume in the targeted coin increases by 8% in the hour following the celebrity tweet. Finally, we show that a representative retail investor who trades following celebrity tweets makes negative returns after transaction costs.

Entity Neutering

(with Joey Engelberg, Asaf Manela, and Luka Vulicevic)   SSRN link    Abstract
Cutting-edge LLMs are trained on recent data, creating a concern about look-ahead bias. We propose a simple solution called entity neutering: using the LLM to find and remove all identifying information from text. In a sample of one million financial news articles, we verify that, after neutering, ChatGPT and other LLMs cannot recognize the firm or the time period for about 90% of the articles. Among these articles, the sentiment extracted from the raw text and the neutered text agree 90% of the time and have similar return predictability, with the difference providing an upper bound on look-ahead bias. The evidence here suggests that LLMs are able to effectively neuter text while maintaining semantic content. For look-ahead bias, LLMs can be both the problem and the solution.

Market Signals from Social Media

(with Tony Cookson, Runjing Lu, and Marina Niessner)   SSRN link    Abstract
This paper develops daily market-wide sentiment and attention indexes derived from millions of posts across major investor social media platforms. We find that sentiment extrapolates from past market-wide returns and exhibits a strong reversal. In contrast, attention predicts negative returns as a continuation of previous trends. The two indexes have distinct predictions for aggregate trading: abnormal trading rises when sentiment is low and attention is high. To identify the drivers of attention and sentiment, we use a shock to data sharing networks: We find sentiment spreads through real firm connections while attention does not. Moreover, attention rises after abnormally high trading, while sentiment rises after abnormally high returns. This extrapolative return pattern is asymmetric, primarily driven by negative market jumps. These findings provide new evidence on the daily market dynamics of sentiment and attention.

Immigration and Credit in America

(with Tony Cookson and Benedict Guttman-Kenney)   SSRN link    Abstract
Immigrants enter the U.S. with a blank credit history, regardless of age or home country experience. Motivated by this, we study the assimilation of immigrants into American consumer credit markets. We find immigrants are positively selected: immediately upon credit market entry, immigrant credit scores are 20 to 35 points higher than non-immigrants, on average. Despite their greater creditworthiness, immigrants–especially those who arrive later–are delayed in their credit access and have lower average credit card limits for up to a decade. Immigrants are also less likely to access auto loans and mortgages, a gap that persists into their forties and is unexplained by geographic fixed effects.