Working Papers

Prompted to Start: How Generative AI is Transforming Entrepreneurship

joint with Bo Bian and Mariassunta Giannetti

We construct a novel proxy for occupation and industry exposure to GenAI based on the proportion of queries to GenAI systems involving specific tasks and show that GenAI is increasing entrepreneurial activity in more exposed industries, altering the nature of startups, and changing the characteristics of their founders. Following the diffusion of the new technology, entry into industries more exposed to GenAI tends to be more geographically dispersed, occurring in areas with less access to venture capital funding and less specialized labor. Startups also tend to be smaller in terms of employee count; their employment is concentrated in occupations that are more exposed to GenAI. They also put AI at the core of their businesses, and their founders are more likely to have gained experience in occupations that can be performed using GenAI. Finally, we observe a lower proportion of white founders and lower educational attainment among founders, suggesting that GenAI is lowering barriers to entry.

Working paper

Financing the Global Shift to Electric Mobility

joint with Bo Bian and Huan Tang

Using comprehensive auto loan data, we find that early-stage electric vehicles (EVs), compared to their internal combustion engine counterparts within the same model series, are financed with higher interest rates, lower loan-to-value ratios, and shorter durations. This financing gap is driven by lower and more volatile resale values of early-stage EVs, stemming from rapid technological advancements in EV-specific technologies. This technological obsolescence raises collateral risk, leading to tighter loan terms. Other factors, such as buyer preferences, socioeconomic traits, government incentives, and macroeconomic conditions, have minimal impact. Our findings show that technological carbon-transition risks are reflected in household finance products.

Working paper

Patent Intensity, Firm Life Cycle, and the Long-run Return and Risk Dynamics of Technological Innovators

joint with Adlai Fisher, Jiri Knesl, and Julian Vahl

We introduce patent intensity (PI), patents granted divided by market capitalization, to classify technological-innovator types starting from 1926. PI-portfolios earn large return spreads for a decade post-formation, and standard factors fail to capture these differences. We further show that traditional value, investment, and profitability premiums are much weaker or reversed among patenters versus non-patenters, explaining their general difficulty pricing innovation. Incorporating firm dynamics through expected growth or intangibles reconciles observed returns, with high-PI loadings revealing a life-cycle of persistent expected growth, increasing investment, and improving profitability. Life-cycle dynamics in risk are essential to accurately capturing technological innovators’ long-run returns.

Working paper

Relationship-Specific Investments and Firms’ Boundaries: Evidence from Textual Analysis of Patents

joint with Isil Erel, Daisy Wang, and Michael S. Weisbach

Acquisitions increase inventors’ incentives to undertake relationship-specific investments, as suggested by Hart and Moore (1990). Using inventor-level patent data for U.S. public and private M&As, we measure the complementarity between inventors’ pre-merger knowledge base and the counterparty firm’s patent portfolio, and innovation specificity based on the extent to which post-merger patents draw on technologies distinctive to the counterparty’s portfolio. High-complementarity inventors are more likely to stay with the combined firm and, after completed deals, increase innovation specificity by economically large amounts. A falsification exercise shows no such shift when inventors join via hiring without patent ownership transfers.

Working paper

Entrepreneurs' Diversification and Labor Income Risk

joint with Andrew Ellul, Marco Pagano, and Valentina Rutigliano

Entrepreneurs with more diversified portfolios of private firms provide more insurance against labor income risk: in a sample of over 524,000 Canadian firms and 858,000 owners, firms owned by such entrepreneurs offer more stable jobs and earnings to employees. In firms whose owners' portfolios are one standard deviation more diversified, the passthrough rates of foreign sales shocks to layoffs and labor earnings are 13% and 41% lower, respectively. These entrepreneurs reduce their own compensation and increase firm leverage to fund labor income insurance. Enhanced insurance is associated with better retention of valuable human capital and fewer costly terminations, potentially improving firm performance.

Working paper

Token-based Decentralized Governance, Data Economy and Platform Business Model

joint with Shiqi Zhang

We develop a model in which a founder, setting up a data-driven platform, chooses between traditional centralized governance or a decentralized blockchain-based approach featuring token-based voting and an optional token-burning mechanism. Users, who have heterogeneous privacy costs, decide whether to join and share their data to enable platform production. In decentralized platform governance, they collectively vote on data-sharing policies and may exit by burning tokens for compensation. We first characterize the optimal decentralized governance. We then demonstrate that, by reducing coordination costs via commitment, decentralized governance can outperform centralization, simultaneously enhancing founder's payoff and users' welfare. Our findings help rationalize the rise of platforms managed by DAOs.

Working paper

Owner Culture and Pay Inequality within Firms

joint with Guangli Lu and Iris Wang

Using a comprehensive dataset of employee-employer-firm owner-immigration records in 2001-2017, we examine the impact of immigrant owners' national culture on within-firm pay inequality. Firms owned by immigrants from more individualistic countries have higher pay dispersion among employees. Individualism alone explains more than half of the variation in within-firm pay inequality across owners' countries in our sample. This result is robust across various empirical methods, including difference-in-differences analysis of ownership changes. Owners' individualism is associated with their employee compensation structures: more frequent and larger performance pay components—especially for highly educated employees, quicker promotions to high-paying positions, and less pay compression. These findings highlight the prominent role of culture in shaping pay practices and elucidate broader determinants of income inequality.

Working paper

Mutual Fund Disagreement and Firm Value: Passive vs. Active Voice

joint with Iris Wang

Using mutual funds’ proxy voting behavior, we construct a novel measure of shareholder disagreement between passive and active mutual funds. To create the measure, we use mutual funds’ voting decisions to capture the difference in “approval of management” between passive and active funds. We find that the disagreement among the two groups destroys firm value when the vote outcome of a proposal is not perfectly anticipated—viable. Using Federal Open Market Committee announcements with press conferences as events that create scope for investors to interpret news differently for individual firms, we show that such value-destroying effect is causal. When proposals are not viable, the presence of disagreement increases firm value. We show evidence consistent with such disagreement being a form of shareholder engagement that is interpreted in a positive way by the financial market participants.

Working paper

The Impact of Private Money Creation: Asset Allocation and “Opacity Discount” in Bank Lending

joint with Isha Agarwal and Joyce Xuejing Guan

This paper examines how the private money creation function of financial intermediaries influences their asset composition and firms’ cost of debt. Following Dang, Gorton, Holmström, and Ordonez (2017), we hypothesize that intermediaries create money-like liabilities by backing them with assets that are safe and opaque, thereby minimizing information sensitivity. Using a quasi-natural experiment—the 2014 U.S. MMF reform—we show that changes in the moneyness of MMF liabilities significantly impact their asset allocation. Institutional MMFs reduced their portfolio weight of opaque assets—defined as holdings not disclosed by the MMF—by 40%. Furthermore, firm-level analysis of syndicated loans and bond issuances reveals that positive information acquisition cost shocks to borrowers reduce the loan-bond spread—the relative cost of firm financing using bank loans versus the public bond market—highlighting the “opacity discount” in bank lending. Instrumental variable estimation and event studies support the causal interpretation of our findings.

Working paper