Working Papers

Using a proprietary dataset detailing all startup applications, all internal judging scores and judges’ written comments, all signed financing contracts, and even all audio recordings of interviews and contract negotiations involving one of the largest venture capital-backed startup accelerators in the United States, we open the ‘black box’ of venture capital (VC) decision-making. We first study the entire internal VC investment selection process by examining the key determinants of judging scores from initial screening through to final portfolio firm decisions. For example, by focusing on how individual VC partners/employees evaluate the same potential portfolio firm, we provide novel evidence on the existence of significant VC judge-founder ‘homophily’ biases and detail how different judging settings (e.g., solo vs. group evaluations; availability of quantitative vs. qualitative information) can amplify or mitigate such biases. Next, we are the first to empirically document the key features of a recent innovation in startup firm financial contracting instruments (namely Simple Agreement for Future Equity (SAFE) and Keep It Simple Security (KISS) contracts) and investigate their relationship with startup firm characteristics and internal VC evaluations. We offer novel insights into the role of a salient ‘anchor’ or ‘reference point’ in setting future equity pricing terms as well as the importance of startup financial constraints in VC term sheet negotiations.

We examine the impact of positive private firm innovation shocks and private firm liquidity shocks on existing public firm peers using an extensive dynamic textual net- work. We build this textual network using a LLM to analyze millions of internet web- pages over time from 2000 to 2021 from the Internet WayBack Machine. We document that state-level shocks to the incentives of private firms to invest in R&D have positive effects on related public firms consistent with product market complementarities. Af- ter these innovation shocks to private peers, related public firms increase acquisitions of private firms and have improved profitability, sales growth, competitive positioning, andinvestmentsincludingR&D.Incontrast, state-levelliquidityshocks, whichinstead relax financing constraints of private firms located in these states, negatively impact related public firms, consistent with substitution and private firms increasing compet- itive pressure on public firms. These results indicate a new empirical framework and novel indications of where private firms are either complements or substitutes to larger public firms. ∗University of Southern California Marshall School of Business and Tuck School of Business at Dart- mouth College and National Bureau of Economic Research, respectively. Hoberg can be reached at hoberg@marshall.usc.edu. Phillips can be reached at gordon.m.phillips@tuck.dartmouth.edu. We thank the National Science Foundation for funding earlier work that laid the foundation for this project (grants #1561068 and #1937153). We thank seminar participants at Aalto University, Tulane University, the Uni- versity of Amsterdam, and Vrije University of Amsterdam. We thank Xinwei Du, Prathamesh Koranne, Deepti Poloru, and Himanshu Rawlani for excellent research assistance. All errors are the authors’ alone. Copyright ©2024 by Gerard Hoberg and Gordon Phillips. All rights reserved. Are Private and Public Firms Friends or Foes? Evidence from Millions of Webpages Abst

We extend the Berk and Green (2004) model to allow clients to derive utility from both performance-related (alpha) services and non-alpha services, such as financial plan- ning. Clients gradually learn about the value of these non-alpha services through their interactions with financial advisors. We test our model predictions using a dataset that covers 3,000 financial advisors and their 500,000 clients’ portfolios over 13 years. Consistent with our model, the total number of client investment plans, our empiri- cal proxy for non-alpha services, is the primary driver of the revenues of broker-sold funds and advisors. Investment skills and performance do not significantly affect their revenues. In line with clients’ gradual learning about the value of non-alpha services, larger advisors experience lower client exit rates and derive most of their revenues from long-term clients who steadily increase their investments and use more non-alpha ser- vices over time. Our findings help rationalize the prevalence of financial advisors with costly investment recommendations and the survival in equilibrium of underperforming broker-sold mutual funds. ∗We are grateful to Jules van Binsbergen for his valuable input during the early stage of this project, to Jennifer Huang (discussant), Alberto Manconi (discussant), and conference participants at EFA 2025, CICF 2025, FMARC 2024, as well as seminar participants at Indiana University, Nova University, London Business School, University of Notre Dame, Bocconi University, Aarhus University, BI Norwegian Business School, University of California San Diego, and University of Southern California. Alessandro Previtero is with Indiana University and NBER. Email: aleprevi@indiana.edu. Ran Xing is with Stockholm University and the Swedish House of Finance. Email: ran.xing@sbs.su.se. 1 Introduction Households rely heavily on financial advisors. In the U.S., advisors intermediate over 50% of mutual fund transactions, and 56% of households repo

We show that household linkages, formed primarily of spousal employment ties, are important in explaining asset managers' skills and their portfolio choices. Mutual fund managers with spouses that work at the executive and C-suite levels obtain monthly gross returns of up to 0.32% above asset managers with non-executive spouses. This effect is driven largely by managers' quarterly stock trades in the industries where their spouses are employed. The spouse-industry stocks bought by executive-spouse-linked managers outperform those bought by the nonexecutive-spouse-linked managers by a large and significant 4.30% in the next quarter. Symmetrically, the stocks sold by executive-spouse linked managers underperform those sold by the nonexecutive-spouse-linked managers by a significant -4.33% in the following quarter. These patterns suggest that spousal relationships facilitate fund managers' comprehension of industry-level information. The linked fund managers' spouse-industry trades predict subsequent earnings surprises and firm level news. Overall, our results highlight the importance of household links to information production in the asset management industry.

Using over 4.6 million retail investor queries from China’s investor interactive platforms, we analyze individual investors’ environmental, social, and governance (ESG) priorities. Governance overwhelmingly dominates their ESG interests, countering the assumption that retail investors prioritize environmental or social concerns. Retail investors strategically focus on firms facing significant ESG risks, indicating economically rational behavior. Notably, retail investors’ ESG priorities diverge substantially from commercial ESG ratings, particularly concerning environmental and governance dimensions. Textual analysis reveals that most ESG-related queries are financially motivated (85–90%), while ethical and altruistic motives account for 8–12%, and warm-glow motivations are minimal (1–2%). These findings highlight the distinctive governance orientation and financially driven ESG preferences of retail investors, offering important implications for ESG frameworks and behavioral finance theory.

This study examines how retail investors respond to the advent of generative AI and the ensuing capital-market consequences. Using a large language model to analyze millions of investor questions from China's Investor Interactive Platforms (IIP), we document a structural reallocation of retail investors' information demand. Following the launch of ChatGPT, firms with high prior AI engagement receive significantly more AI-related questions, while questions about non-AI topics decline. This shift in information demand is associated with higher trading volume and volatility but, paradoxically, weaker price informativeness. Our findings suggest retail investors' AI information demand introduces speculative noise and crowds out their attention to non-AI fundamentals, providing a novel perspective on how generative AI transforms financial markets.

This paper explores whether relationships with banks or individual bankers de- livergreatervaluetomunicipalborrowers. Akeyidentificationstrategyexploits the quasi-exogenous shock from the 2021 Texas underwriter ban, which barred five of the largest banks from underwriting municipal bonds in the state and triggered widespread banker departures. Using novel data on banker moves, I show that affected municipalities follow their banker at twice the rate of unaf- fected peers. Instrumenting the follow decision in an IV–DiD framework, I find that following the banker reduces yield spreads by 36 basis points, fully offset- tingtheban’s4basis-pointspreadincrease. Theseimprovementsariseviathree channels: an informational channel, where unrated issuers lacking public credit signals experience an extra 16 basis-point decline; a network channel, where institutional investors allocate $1 million more per quarter to the banker’s new bank; and a banker-quality channel, where following an MBA-educated banker generates an additional 22 basis-point spread reduction. Overall, relationship- specific human capital significantly shapes municipal underwriting outcomes.

Thisstudyinvestigatestheeffectsofpoliticalhomophilybetweenventurecapital(VC) partnersandcompanyCEOsoninvestmentdecisionsandoutcomes. Usingacompre- hensivedatasetofU.S.VCinvestmentsmatchedwithpoliticaldonationrecordsfrom 2000to2021,wefindthatpoliticalsimilarityincreasesthelikelihoodofinvestmentbut negativelyimpactsexitperformance,loweringIPOandM&Asuccessratesanddelay- ing exits. These findings support the in-group favoritism explanation. Shared parti- sanship promotes trust and collaboration but can lead to overconfidence and group- think that deteriorates exit performance. Alignment with the broader political envi- ronment(e.g., theincumbentgovernmentorlocalpoliticalpreferences)canmitigate theseeffectsbyenhancinglegitimacyandaccesstoresources. Ouranalysisofinvest- ment structure shows that politically aligned CEO-VC pairs favor early-stage, first- round, non-syndicated deals, yet experience slower follow-on financing and longer intervals between rounds. These findings offer novel insights into how ideological alignmentinfluencesventureinvestmentbehaviorandperformancewithimplications forentrepreneurs,investors,andpolicymakers. D72,G24,L14,L26,M13

We study whether partisan alignment with the president shapes housing market ex- pectations, home purchase behaviors, and aggregate housing market outcomes in the United States. Survey evidence shows that individuals report more optimistic home buying expectations when their affiliated party controls the White House. We then create a novel dataset that links individual home purchase records to voter registration records for approximately 48 million registered voters in states Florida, Georgia, North Carolina, Nevada, New York, and Ohio from 2010 to 2023. Repub- licans purchase more homes than Democrats in a given year. We find individuals whose party controls the White House purchase more homes. These partisan driven shifts in housing demand translate into aggregate housing market outcomes: po- litically aligned counties experience higher home purchase volumes, homeownership rates, and house prices but lower returns.

Thispaperinvestigateshowsalienceinfluencesdecision-makinginearthquake- prone real estate markets in Türkiye, focusing on two critical events: the 2018 revision to the national earthquake hazard map and the catastrophic 2023 earthquake that resulted in over 50,000 fatalities. Our findings indicate that while the updates to the hazard map have little effects on property val- ues, the actual occurrence of a disaster significantly reduces home prices and increases insurance uptake in high-risk but physically unaffected areas. A one-standard-deviation increase in baseline seismic risk is associated with a 4% decline in home prices after the earthquake. Additionally, the data show that areas with strong social connections to disaster-stricken regions experi- ence more pronounced declines in home sale prices, highlighting the role of personal relationships in amplifying risk perception. Overall, these results suggest that the salience of a vivid, catastrophic event is far more impact- ful in shaping economic behaviors than abstract, probabilistic information in high-risk scenarios.

Individual investors are sensitive to peer performance and particularly dislike “falling behind.” We use unique granular data on the transactions and holdings of retail investors to study portfolio adjustment in response to relative performance of their portfolios. We show that investor behavior is consistent with preferences over future wealth that are S-shaped around an external reference point provided by a salient market benchmark: if their portfolio lags the “market,” they tend to increase the risky share of their portfolio, as well as purchase riskier securities, as characterized by high market beta, idiosyncratic volatility, and positive skewness. As the salience of the market index increases, investors become more sensitive to relative performance. The effect is asymmetric, more pronounced in bull market periods, and does not reverse when individual portfolios are ahead of the market. Our evidence provides a novel perspective on the individual investors’ demand for risky assets.

Wedevelopastructuralmodelofcreditcounterpartyriskinwhichcontagionarisesfromaninter-firm production network. We then propose a parsimonious empirical approach that directly incorporates networktopologytopredictcreditspreads. Wefindthatincorporatingnetworkedgefeaturesinducesan averagecredit-spreadchangeofapproximately21.8%andyieldsanincremental𝐿2of0.56inexplaining creditspreads. Ourresultsshowthatnetwork-basedcounterpartyriskisstronglypricedandplaysafirst- order role in shaping credit spreads, particularly during periods when production networks experience severedisruption,reorganization,orrewiring. Networke!ectsareespeciallyimportantforfirmsoperating in industries that occupy intermediate positions within supply chains, rely heavily on distribution and logistics,andfacelowsubstitutabilityofinputs.