Working Papers

Why do small businesses exploit business opportunities better in some areas than others?In a sample of 1.2 million consumer-facing establishments, social capital predicts the take-up of risk-free loans controlling for nearby bank branches, income, and education. One standard deviation in the social capital metric accounts for 20 percent of the variation in take-up rates across zip codes, surpassing the impact of a bank branch within 1000 yards. Strong social capital benefits larger, low-growth stores in less-dynamic areas, whereas bank branches help smaller, fast-growing stores in more dynamic areas. Virtual connections have the greatest effect in already advantaged locations.

Using plausibly exogenous variation in regional Twitter adoption in the United States, we show that a 10% increase in social media usage causes a 2.5% rise in stock ownership. Consistent with lowering the costs of acquiring information, Twitter has larger effects in counties with low pre-existing stock market knowledge, improves knowledge about asset returns, and leads to a decline in the number of financial advisors. Social media also boosts interest in volatile "meme stocks" favored by retail investors. Our findings highlight the unique influence of social media on household portfolio decisions, distinct from other modern information technologies.

We identify a new dimension of cross-firm linkages by exploring the social connectedness between firms' geographical locations. Industry peers located in regions with strong social ties tend to adopt similar strategies and exhibit strong co-movements in both fundamentals and returns. However, this information is not immediately reflected in stock prices and can be exploited using information contained in social peer returns (SPFRET). The predictability of SPFRET lasts for up to a year and forecasts future earnings surprises, analysts' forecast errors, and returns around earnings announcements. The effect is particularly strong for low-visibility firms and those located outside of industry clusters.

We introduce a new measure, Partisan Portfolio Disagreement (PPD), which captures the extent to which Democratic- and Republican-leaning investors hold consistently different equity portfolios. We demonstrate how the long-term evolution of PPD can be estimated from county-level portfolio and political differences. Our findings show that PPD among wealthy U.S. households more than doubled between 2001 and 2019. By 2019, an average of 20% of Democrats' and Republicans' portfolios is subject to partisan disagreement. Leveraging the staggered county-level entry of Sinclair, a conservative television network, we provide causal evidence that political disagreement drives portfolio disagreement.

Can retail investors on social media platforms effectively target hedge fund short positions? We show that the disclosure of hedge fund short positions triggers activity on WallStreetBets, which in turn precipitates price increases for heavily shorted stocks. In line with a causal interpretation, the effect of social media activity on stock returns (i) emerges when short sales are publicly disclosed but not when they are settled, (ii) strengthens when social media posts reflect coordination attempts against hedge funds, and (iii) attenuates during trading restrictions imposed by Robinhood. The resultant price appreciations hurt hedge funds, which respond by shorting less aggressively, leading to prolonged overpricing.

We study the spillover of corporate diversity culture among interacting firms. Utilizingdata from the U.S. Census, we find that through labor market competition, the opening ofa high-profile plant by a more (less) pro-diversity firm leads to a decrease (increase) in thelocal gender wage gap. Furthermore, changes in the local employer’s gender wage gap canlead to broader changes in the diversity culture of their parent companies. But a culturalspillover occurs only when the leaders of a company are receptive to the new entrant’sculture and an internal learning platform is in place.

Using account-level social and trading data, we document that after observing peers win randomized IPO allotments, connected investors shift their purchases toward lottery-like stocks. This finding reveals a cross-domain social transmission mechanism that stimulates the formation of behavioral biases, where investors mistakenly extrapolate peers' IPO lottery wins to their own prospects in lottery-like stock trading through the representativeness heuristic. This effect is amplified when IPO wins are more salient and when affected investors trade more frequently or have recently experienced losses. Alternative mechanisms, including hot hand fallacy, social persuasion, and catching up with the Joneses, do not explain our results.

Using submission level data from the social media platform Reddit, we rely on the theoretical framework of Pedersen (2022) to examine how social media affects belief formation, price discovery, and trading dynamics. Consistent with the predictions on network belief spillover, we find that the opinions of hardheaded investors (rational or fanatic) significantly predict future opinions of naive investors, especially when these investors have larger influence. For return predictions, social media tones positively and significantly predict future returns, and more so when agents' influence is higher. Finally, for trading dynamics, higher agent tones in networks with higher agent influence increase retail flows and deter shorting flows. Short sellers' consideration of agent influence in deciding to ride or burst bubbles enhances their abilities to predict negative returns. These findings generally support Pedersen's predictions.

This paper examines how informal and formal networks shape performance in the venture capital (VC) industry. Using data on all U.S.-based VC investments from 1990 to 2009 and career histories from LinkedIn, I develop a structural framework that connects three types of networks: coinvestment ties, historical affiliations, and latent social connections. In the baseline model, VC performance is a function of peer performance, capturing network spillovers through a micro-founded production function. To address endogeneity in network formation, I extend the model using a two-step instrumental variables strategy that leverages variation in past professional and alumni ties. Finally, I introduce an endogenous network formation model in which VCs strategically choose connections based on expected peer quality, allowing for the recovery of latent social networks from equilibrium outcomes. Across specifications, better-connected VCs exhibit significantly higher exit rates. Estimates from the endogenous model suggest that a 1% increase in social connectedness raises a VC's exit rate by 0.2 percentage points, while a 1% improvement in peer performance leads to a 0.74 percentage point increase in connection intensity. These findings highlight the economic value of informal relationships and offer new empirical tools for measuring network effects in private capital markets.

Sophisticated investors exert more effort at human-intensive tasks in the age of AI. I hypothesize that AI reduces costs of collecting machine-based information, thereby facilitating human- interaction-based information acquisition. Using an event study design and an IV approach, I find that hedge funds increase earnings call participation—along both the extensive and intensive margins—after adopting machine downloads of SEC filings. Post-automation call attendance is associated with higher fund returns and profitable stock trades. Overall, this study identifies a novel AI-productivity mechanism: by substituting for human effort at automation-prone tasks, AI complements high-skilled workers without directly augmenting them at interaction-based tasks.

As retail trading has expanded over the past few decades, concerns have grown that noisy social-media signals dominate the information environment, leaving it unclear whether analyst recommendations still matter or can improve it. I examine the effect of a stock’s first analyst recommendation on transaction costs using Rule 605 data across wholesalers and exchanges. After the first coverage, effective spreads decline on both venues, with the largest reductions at the leading wholesalers. The leading wholesalersalsocedemarketshare,particularlyforlargerorders,astheydonotprovide a greater reduction in effective spread compared to others, and rivals catch up on price improvement. Thus, credible information production strengthens venue competition and delivers measurably better retail execution.

Using voter registration data of loan officers originating residential mortgages in coastal areas, I analyze whether climate change partisanship is reflected in mortgage lending. I find that Democratic loan officers charge higher loan spreads for mortgages on properties exposed to sea level rise (SLR) than Republican loan officers. The results hold after controlling for loan officer fixed factors. Partisan sorting is more pronounced for properties outside FEMA-designated flood zones, and for loan officers located nearer the coast or in communities with fewer climate change believers. These findings highlight how political ideology shapes the pricing of climate risks in mortgages.