Published Papers

In the dynamic landscape of the digital economy, social trading platforms are experiencing rapid growth. Our study delves into the impact of changes in social audience size-measured by the number of followers-on traders' performance and behaviors. Utilizing data from a company-led randomized field experiment conducted on a prominent cryptocurrency social trading platform, we unearth intriguing findings. Traders garnering increased social audience size exhibit tendencies to trade more frequently, utilize higher leverage, and, surprisingly, attain poorer performance. Notably, these adverse effects intensify among traders who previously excelled, suggesting a link to overconfidence. Interestingly, our research uncovers a reference point effect associated with social audience size. Removing accumulated social audience size does not alleviate the negative consequences, instead, they persist. Additionally, we observe an extra adverse effect when traders experience a reduction in the digit magnitude of follower counts, supporting our hypothesis about social audience size serving as a reference point. Our study carries significant implications for the design of social trading platforms. It serves as a crucial reminder for both traders and platform managers to carefully navigate the interplay between social audience size dynamics and trading decisions.

We examine how indirect connections (i.e., friends of friends), an important yet understudied feature of social networks, may affect bank loan contracts. Using a sample of bank loans issued by U.S. public firms, we find that indirect networks built on board interlocks significantly reduce loan spreads. However, bank monitoring and loan quality are negatively affected, suggesting that indirect networks may give rise to favoritism treatment by banks. A novel set of difference-in-difference tests exploiting changes in higher-order network structures provides the network foundation and lends support to a causal interpretation of our findings. Overall, our results suggest that indirect connections within social networks, specifically through board interlocks, can have significant economic impacts on bank loan contracts.

We show that the social capital embedded in employees' networks contributes to firm performance. Using novel, individual-level network data, we measure a firm's social capital derived from employees' connections with external stakeholders. Our directed network data allow for differentiating those connections that know the employee and those that the employee knows. Results show that firms with more employee social capital perform better; the positive effect stems primarily from employees being known by others. We provide causal evidence exploiting the enactment of a government regulation that imparted a negative shock to networking with specific sectors and provide evidence on the mechanisms.

How do venture capital investors react to social movements, including those that relate to historical underrepresentation in funding? We use image and name algorithms combined with clerical review to classify race for 150,000 founders and 30,000 investors. These data allow us to assess the impact of George Floyd's murder on VC funding of Black entrepreneurs and identify which VCs were most responsive. Although VCs responded swiftly, investment in Black-founded startups reverted to prior levels within two years. This temporary reaction was concentrated among those who had never previously invested in any Black entrepreneur. Moreover, the investors who responded were less likely to invest in more than one Black-founded startup and were less inclined to engage deeply by taking a board seat. Finally, it appears that the best Black entrepreneurs may have anticipated this "token" response because they did not match with investors who had no experience funding Black startups.

Tipping is a complex phenomenon cutting across various stakeholders-firms, customers, and workers. Analyzing the long-run impact of policies related to tipping is therefore challenging. To facilitate such analysis, we develop a modeling framework in which a tipping norm forms endogenously in a market consisting of a firm that offers service to potential customers. Customers choose whether to consume the service or not and, if yes, how much to tip the server afterward. With tipping, customers show appreciation to the server by sharing a fraction of their surplus but also undergo social pressure to comply with the prevailing norm. This tipping norm is shown to evolve endogenously through a dynamic process of sequential market adjustments over time: the average tip in each period determines the tipping norm for the following period, causing the firm to adapt the price and customers to adapt their tips accordingly. Characterization of this equilibrium outcome allows us to derive qualitative results on the long-run impact of different exogenous factors on tipping: we find that the equilibrium tip-to-price ratio increases when customers are more sensitive to social pressure, their range of service valuations spreads out, or they consider the service more valuable. Building on this framework, we further investigate several economic implications of tipping pertaining to social welfare, labor cost, and service quality, thus uncovering incentives and trade-offs to which the tipping mechanism gives rise from the firm, the worker, and the customer's perspectives.

Honest and dishonest behaviors may both diffuse among the members of an organization. Knowing which of the two spreads faster is important because it impacts the extent to which managers will need to resort to other, potentially more costly solutions to curb dishonest behavior. Assessing empirically which of honest behavior or dishonest behavior spreads faster is challenging because this requires field measurements of social relationships and dishonest behavior of individual members, which poses both measurement and inference problems. We examine an original fine-grained data set from a large company that allows for identifying agents likely to be dishonest and interactions among employees while offering a natural experiment that circumvents the inference problems associated with identifying peer-to-peer diffusion. We find (1) that dishonest behavior diffuses, whereas honest behavior does not; (2) that diffusion likely operates through spreading information about opportunities for collusion; and (3) that policies that screen on dishonesty at hiring may be efficient to curb dishonest behavior in environments with high turnover.

We investigate peer effects associated with a user's churn and subsequent retention of other users in the network. We use data from an upscale gym with information on exogenous churns of users that allows us to identify peer effects. We estimate a structural model of users' plan choice and usage decisions that allows for bidirectionality in peer effects, direct/indirect peer effects, heterogeneity in the degree of centrality of users, and dynamics in peer effects. Our main results are as follows. (i) Unlike the churn of a low- or medium-centrality user, the churn of a high-centrality user has a strong negative cascade effect on the usage and subscriptions of other users in the network over time. (ii) To stop this negative cascade effect and increase the retention in the rest of the network, the firm's decision on which of the central churner's peers to target with promotions depends on the tradeoff between two forces: the individual level impact of the targeted promotion versus the positive spillover effect to the rest of the network that can increase over time because of carryover effects of usage and subsequent social multiplier effects (that stem from bidirectionality in peer effects). Thus, retention strategies focused on short-term outcomes should prioritize targeting users with low centrality, whereas strategies aimed at long-term outcomes should prioritize targeting users with high centrality.

We develop a dynamic model of naive social learning from consumer reviews. In our model, consumers decide if and what to buy based on both the products' expected quality and their idiosyncratic taste for them. Products' qualities are initially unknown and are (mis)learned from reviews. At the heart of the model lies a dynamic feedback loop between reviews, beliefs, and choices: period t reviews influence t+1 consumers' beliefs and, thus, choices; these determine the average of t+1 reviews, which, in turn, influences t+2 beliefs, choices, and reviews. We show that, in the long run (t=), reviews are systematically biased, leading some consumers astray. In particular, in both monopoly and duopoly, reviews relatively advantage lower quality and more polarizing products because these products induce stronger taste-based consumer self-selection. Thus, in stark contrast with the winner-takes-all dynamics of classic observational learning models in which consumers learn from the choices of their predecessors, social learning from opinions generates excessive choice fragmentation. Our findings have implications for interpreting the variance and number of reviews, pricing in the presence of reviews, and the short- and long-term effectiveness of fake reviews.

We present evidence that loan allocations and loan terms are closely linked to the strength of social connections between bank and borrower regions. Lending increases with social connectedness, particularly in the presence of strong screening incentives. If connectedness is high, banks discriminate between borrowers more, the terms of originated loans are more borrower friendly, and loan performance is better. Furthermore, social connectedness is associated with higher bank profitability and with lending-related increases in borrower-region GDP growth and employment. Our results suggest that lending barriers decrease with social connectedness, which helps our understanding of geographic lending patterns and regional economic disparities.

We investigate whether herding behaviour of bond funds influences corporate bond issuance. Using quarterly data from 1998 to 2018, we find that only buy herding significantly increases bond issuance activity. Further analysis supports an informational channel, suggesting that 'investigative herding' among funds enhances credit information efficiency more significantly than demand-driven price impacts. Our main identification strategy employs an instrumental variable approach, based on the share of  inexperienced managers among fund bondholders. 

Media tone constructed from 7,000,000 articles from 2,000 global media and 800 social media sites is found to be a genuine risk factor that cross-sectionally prices currencies. It can predict excess US dollar returns for up to six months and surpasses the no-change benchmark in predicting returns out of sample. Its predicted value contains information beyond those predicted by currency factors and business cycles. Evidence collaborates with the theory that Media tone increases investment returns, has pronounced predictive power for the currencies associated with hard-to-value characteristics, and its predictive power increases with media sources. Trading of rational investors, including banks, is associated with Media tone.

We examine how social norms measured by religiosity influence institutional investors in the stock lending market. We find that firms with higher average local religiosity of their block institutional investors are associated with higher utilization rates, lower supply, but not higher demand of lendable shares. Such firms have higher stock lending fees and likelihoods of becoming special. A higher short interest, when coupled with higher blockholder religiosity, is a stronger predictor of lower stock returns. Our findings suggest that the social norms of institutional investors act as an origin for limits to arbitrage that impede market efficiency through stock lending.