Published Papers

We examine how increased competition stemming from an innovation in financial technology influences sell-side analyst research quality. We find that firms added to Estimize, an open platform that crowdsources short-term earnings forecasts, experience a pervasive and substantial reduction in consensus bias and a limited increase in consensus accuracy relative to matched control firms. Long-term forecasts and investment recommendations remain similarly biased, alleviating the concern that the documented reduction in bias is a response to broad economic forces. At the individual analyst level, we find that bias reduction is more pronounced among close-to-management analysts, and that more biased analysts respond by reducing their coverage of Estimize firms. The collective evidence suggests that competition from Estimize improves sell-side research quality by discouraging strategic bias.

We study whether social media can play a negative information role by impeding price discovery in the presence of highly speculative rumors. We focus on merger rumors, where most do not materialize. We find that merger rumors accompanied by greater Twitter activity elicit greater immediate market reaction even though rumor-related Twitter activity is unrelated to the probability of merger realization. The price distortion associated with tweet volume persists weeks after a rumor and reverses only after eight weeks. The price distortion is more pronounced for rumors tweeted by Twitter users with greater social influence, for target firms with low institutional ownership, and for rumors that supply more details. Our evidence suggests that social media can be a rumor mill that hinders the market's price discovery of potentially false information.

This study examines whether crowdsourced forecasts of earnings and revenues help investors unravel bias in earnings announcement news, which is commonly derived from analyst forecasts. Our results suggest that investors, on average, understand and price the predictive signals reflected in crowdsourced forecasts about the bias in analyst-based earnings and revenue surprises. Using the staggered addition of firms to the Estimize platform, we find that crowdsourced coverage is associated with reductions in the mispricing of forecast bias and declines in earnings announcement premia. We further find some evidence that managers use income-increasing accruals to meet the crowdsourced forecast benchmark and that they respond to crowdsourced coverage through increased downward earnings and revenue guidance. Overall, we conclude that user-generated content on crowdsourced financial information platforms helps investors discount biases in traditional equity research and thereby better process the news in earnings announcements.

We model visibility bias in the social transmission of consumption behavior. When consumption is more salient than nonconsumption, people perceive that others are consuming heavily, and infer that future prospects are favorable. This increases aggregate consumption in a positive feedback loop. A distinctive implication is thatdisclosure policy interventions can ameliorate undersaving. In contrast with wealth-signaling models, information asymmetry about wealth reduces overconsumption. The model predicts that saving is influenced by social connectedness, observation bi-ases, and demographic structure, and provides new insight into savings rates. These predictions are distinct from other common models of consumption distortions.

We propose a new measure of private information in decentralized markets-connections-which exploits the time variation in the number of dealers with whom a client trades in a time period. Using trade-level data for the U.K. government bond market, we show that clients perform better when having more connections as their trades predict future price movements. Time variation in market-wide connections also helps explain yield dynamics. Given our novel measure, we present two applications suggesting that (i) dealers pass on information, acquired from their informed clients, to their affiliates, and (ii) informed clients better predict the orderflow intermediated by their dealers.

This study examines whether sell-side equity analysts engage in "social learning" in which their earnings forecasts for certain firms are influenced by the forecasts and outcomes of "peer" analysts associated with other firms in their respective portfolios. We find that analyst optimism is negatively correlated with recent forecast errors, by peers, on other firms in the analyst's portfolio. An analyst is also more likely to issue "bold" forecasts when peers recently issued similar forecasts for other portfolio firms. Analysts learn more from peers with similar personal characteristics. Overall, social learning benefits analysts and improves their forecast accuracy.

Flooding is the most costly natural disaster faced by US households, yet policymakers are puzzled by the low take-up rates for flood insurance. Leveraging novel transaction-level data, this paper studies the influence of social interactions on households' insurance decisions. I show that households increase flood insurance purchases by 1-5 percent when their geographically distant friends are exposed to flooding events or to campaigns for flood insurance. These exogenous shocks to far-away friends should not affect local households' own insurance decisions except through peer effects. I provide evidence suggesting that social interactions facilitate learning through information dissemination and attention triggering.

We present a model of media coverage of corporate announcements. Firms strategically use the media to communicate corporate announcements to a group of traders who observe announcements not directly but through media reports. Journalists strategically select which announcements to report to readers. Media coverage inadvertently incentivizes firms to manipulate the underlying announcements. In equilibrium, media coverage is tilted towards less manipulated negative news. The presence of financial journalists leads to more manipulation but makes stock prices more informative on average. We provide additional predictions regarding the media's impact on the quality of firm announcements and stock prices.

I present closed-form solutions for prices, portfolios, and beliefs in a model where four types of investors trade assets over time: naive investors who learn via a social network, "fanatics" possibly spreading fake news, and rational short- and long-term investors. I show that fanatic and rational views dominate over time, and their relative importance depends on their following by influencers. Securities markets exhibit social network spillovers, large effects of influencers and thought leaders, bubbles, bursts of high volume, price momentum, fundamental momentum, and reversal. The model sheds new light on the GameStop event, historical bubbles, and asset markets more generally.

We study the influence from social interactions on equity trading. Using unique data on stock transactions, we exploit the quasi-random assignment of students to classrooms in a financial training program to identify how peer experience affects investor behavior. We find that individuals react more to peer gains than to peer losses. Students enrolled in courses where peers have positive outcomes: (i) are more likely to start trading, (ii) purchase similar stocks as their classmates, and (iii) are disproportionally attracted to stocks with extreme returns. These stocks have low subsequent returns, and new investors reacting to peer gains underperform other investors.

Do social networks help firms recruit talented managers? In our setting, firms are randomly connected to prospective young managers through former employees. Under a discrete choice model, we find networks increase the likelihood firms hire high-ability managers, while having no effect on the hiring rate of low-ability managers. Effects are greatest for nonlocal firms, strong ties, and peers living in the same neighborhood. Survey evidence suggests social networks promote recruitment by providing information about firm fundamentals to potential applicants. Our results help rationalize why the majority of managers hold prior connections to the firm.

We show that institutional investors are more likely to invest in firms from regions to which they have stronger social ties but find no evidence that these investments earn a differential return. Firms in regions with stronger social ties to locations with many institutional investors have higher valuations and liquidity. These effects are largest for small firms with little analyst coverage, suggesting that the investors' behavior is explained by their increased awareness of firms in socially proximate locations. Our results highlight that the social structure of regions affects firms' access to capital and contributes to geographic differences in economic outcomes.