AFA PhD Student Poster Session
Poster Session
Sunday, Jan. 3, 2027 7:00 AM - 6:00 PM (EST)
Monday, Jan. 4, 2027 7:00 AM - 6:00 PM (EST)
Tuesday, Jan. 5, 2027 7:00 AM - 1:00 PM (EST)
- Chair: Andrea Eisfeldt, University of California-Los Angeles
Loan Prepayment, Securitization, and the Expansion of Productive Credit
Abstract
For borrowers, prepayment reflects improved financial health and enables better terms. For lenders, it reduces expected returns, creating frictions that restrict credit to prepayment-prone borrowers. Exploiting an exogenous shock to secondary-market pricing in the SBA 7(a) program, which reduced securitization profitability without affecting borrower demand or loan quality, I show that securitization mitigates these frictions by transferring prepayment risk at origination and expanding credit to such borrowers. These borrowers are high quality: those who prepay subsequently obtain lower spreads, larger loans, and support more jobs with their subsequent borrowing. A bunching design exploiting discontinuities in SBA prepayment penalties confirms that borrowers value prepayment flexibility. When securitization contracts following the policy shock, high-securitizing lenders disproportionately cut lending to prepayment-prone borrowers, dampening job creation and growth, revealing an overlooked channel through which securitization improves credit market efficiency.Who Captures AI Deflation? Quality-Adjusted Prices, Markups, and Pass-Through in AI Services
Abstract
Between falling compute costs and the firms that use AI stands an oligopoly. I build a novel dataset linking prices, usage, and capabilities to GPU rental costs for over 200 models competing on OpenRouter, a centralized AI marketplace. Structural demand estimates recover provider markups: pass-through of cost declines falls with market power, and the premium segment passes through nothing, retaining the declines as margin. Quality-adjusted prices still fell 78 percent over 20242026, twice the nominal price decline, cutting U.S. PPI by a cumulative 160 basis points through the production network. Entry of more capable models, not price cuts, drives consumer surplus gains.Cognitive Embeddings
Abstract
This paper develops a theory-grounded AI model that mimics analysts' cognitive process of selecting, interpreting, and forecasting from corporate disclosures. The model uses analyst reports to identify economically relevant information in earnings calls and transforms the resulting attention-weighted call representations into horizon-specific cognitive embeddings. These embeddings are behaviorally disciplined cognitive-state proxies: model-implied representations of call content constrained by the analyst's own writing and validated against realized outcomes. On analyst-disjoint test sets, the model explains the term structure of EPS forecasts, from short-to medium-and long-term horizons, as well as long-term growth forecasts. The recovered attention measures also explain forecast errors, linking differences in analysts' written interpretations of common disclosures to differences in subsequent forecast performance. Analysts make different forecasts, and systematic forecast errors, because they attend to different parts of the same public disclosure. More broadly, the framework provides a general empirical approach for recovering model-implied information selection from behavioral text. While the paper focuses on analysts and earnings calls, the same architecture can be applied to other expectation-formation environments, including macroeconomic forecasting, retail trading, managerial communication, and investor-relations settings.The Impact of Spoofing on Bitcoin Market Microstructure
Abstract
This paper investigates spoofing in Bitcoin order books on the Coinbase platform. Using high-frequency data, we show that order-book imbalances predict Bitcoin returns at both minute- and hourly-level horizons. We introduce a novel metric to quantify spoofing intensity and document its pronounced effect on Bitcoin prices: higher bid spoofing intensity is associated with positive returns, whereas greater ask spoofing intensity is associated with negative returns. Spoofing activity intensifies during periods of elevated trading volume, enabling spoofers to extract larger profits. Our estimates indicate that, after controlling for other factors, a one-unit increase in spoofing volume corresponds to a 27-basis-point increase in bid spoofing profit and a 55-basis-point increase in ask spoofing profit. Finally, we demonstrate that spoofing widens bid-ask spreads and deteriorates overall market quality. These findings underscore the pervasive impact of spoofing on cryptocurrency market microstructure.Revisit Fractional Shares: The Real Effect on Market Quality
Abstract
This paper examines the real effect of fractional trading on market quality. Fractional trades constitute 23% of one-share trades and 4% of odd-lots for large stocks, yet are generally liquidity-driven and uninformed. Exploiting the introduction of fractional trading in 2019 and a new reporting policy in 2026 as two quasi-natural experiments, we find that fractional trading widened the quoted spreads and harmed price efficiency. To reconcile the uninformedness of fractional trading and widened quoted spreads, we find evidence that the opaque disclosure of fractional trading elevated market makers' inventory risks and retail brokers "cream skim"" fractional traders by routing the whole-share part of fractional trading to wholesalers with worse execution quality. We further document that fractional trading distorts optimal split outcomes by widening realized quoted spreads ex post."The Value of Supplier Relationships: Measuring Supply-Chain Capital as an Intangible Asset
Abstract
Supplier relationships are a quantitatively important yet largely unmeasured form of intangible capital. I measure their value from revealed preference: how firms choose to keep or replace their foreign suppliers when a tariff shock raises the cost of staying. The approach embeds costly supplier search in a dynamic discrete choice model of a firm's sourcing decisions, estimated on Turkish customs microdata, with the April 2020 Turkish import-tariff shock, which raised duties by up to 50 percentage points on thousands of products from specific origins, providing the identifying variation. In the model, a firm's relationship with a foreign origin carries relationship-specific match capital that accumulates with continued use; severing the relationship destroys this capital and forces costly re-formation through search. The differential response of relationship termination to the tariff across tenure, with young relationships breaking readily while mature ones resist, identifies the accumulated capital. I estimate that a two-year supplier relationship is worth close to a full year of its import value, of which about 94 percent is the incumbent match premium (the value of holding a known, matured match relative to a fresh draw), and only about 6 percent the one-time cost of finding and qualifying a replacement supplier. This magnitude places supplier relationships alongside brand and organizational capital as a first-order intangible asset, and implies that tariffs and other supply-chain disruptions destroy relationship capital well beyond their direct cost.When ChatGPT Stops Talking: GenAI, Retail Trading Coordination, and Market Liquidity
Abstract
Using 14 unexpected ChatGPT outages, we examine whether GenAI coordinates retail trading and how it affects trading informativeness. When ChatGPT is available, retail order flow is more correlated, and a larger share of information-related order flow comoves across stocks. This common component shows little evidence of return predictability, while it appears to substitute for the informative individual component. The informativeness effect of ChatGPT is concentrated in that individual component and runs in opposite directions: it makes retail selling less informative while making retail buying more informative. This asymmetry suggests that the informational effect of such a tool may depend on how it interacts with investors' existing information sets. Such ChatGPT-related coordination also leads to higher liquidity commonality, implying greater exposure to common liquidity shocks.Does Internal Finance Matter? Evidence from Severance Payment Reforms and Italian SMEs
Abstract
This paper identifies the causal effect of internal finance on firm investment and financial structure by exploiting a large and plausibly exogenous cash flow shock to Italian small and medium-sized enterprises (SMEs). The shock originates from pension reforms mandating the outflow of the Trattamento di Fine Rapporto (TFR), a severance payment scheme historically retained on firms balance sheets and used as a source of internal funding. A size-contingent provision required firms with more than 50 employees-based on predetermined employment-to divert newly accrued TFR flows to external funds, reducing cash flows by about 6.3 percent of the annual wage bill. Using a Difference-in-Regression-Discontinuities design and administrative firm-level data, we find no evidence that firms replace the lost internal funds with external finance. Instead, relative to firms just below the threshold, treated firms reduce both long-term bank borrowing and fixed assets. These findings suggest that internal and external finance act as complements rather than substitutes for SMEs.Bridging Two Markets: Dealer Intermediation across Securities Lending and Repo
Abstract
Repo and securities-lending markets trade the same sovereign bonds through separate participant universes, bridged by dealers who borrow bonds from beneficial owners and redeploy them as repo collateral. I model the lending fee as the dealers repo deployment value: the bonds repo specialness, scaled by its redeployment intensity and split across a lending and a repo bargain. Lending fees therefore pass through repo specialness, most strongly for repo-central bonds when lending utilization is high. In matched securities-lending and repo data for euro-area sovereigns (2018 to 2024), the pass-through concentrates there and is muted elsewhere; bonds physically cross into repo, and fees, specialness, and repo volume comove. Auction reopenings confirm specialness is a supply-driven repo price. A lending fee is, in effect, a repo price.Beyond the Numbers. Professional Forecasters Narratives about Inflation and Stock Market Performance
Abstract
Professional forecasters' heterogeneous narratives about how inflation affects stock markets in 2023 rationalize their high disagreement regarding quantitative expectations for 12-month-ahead inflation and stock returns in December 2022. Professional forecasters causally update their return expectations in heterogeneous directions in response to information about the inflation outlook, depending on the narrative they entertain. Moreover, the narratives also affect their asset allocation in a hypothetical portfolio-choice experiment. Providing common signals does not necessarily lead to convergence of beliefs if agents subscribe to different narratives.Fedspeak, LLM-Derived Signals, and High-Frequency Trading
Abstract
Financial markets are often assumed to incorporate Federal Reserve policy decisions instantaneously. Yet it remains unclear whether markets fully absorb policy relevant information at the moment of release, or whether information continues to be revealed as the policy narrative unfolds across sequential communications. We construct a novel measure of policy communication shocks, the Communication Divergence Signal (CDS), which captures semantic, tonal, and framing shifts across sequential FOMC communications. Using large language models and high-frequency event-study methods, we show that communication divergence generates significant and persistent movements in intraday asset prices and abnormal trading volume. Abnormal trading volume is measured relative to non-announcement benchmarks at the same intraday horizon, isolating excess trading activity associated with information arrival. We further document systematic responses in daily open interest, consistent with belief updating and position reallocation rather than transitory liquidity provision. These findings highlight the role of communication sequencing and framing in shaping price discovery and trading dynamics across financial markets.Financial Disclosure by Government Contractors and Government Investment: Evidence from U.S. Research Universities
Abstract
Using the implementation of Governmental Accounting Standards Board Statements No. 34 and 35 (GASB 34/35) across U.S. public colleges and universities, I examine how financial disclosure by research universities affects the allocation of federal research funding. I find that GASB 34/35 reduces federal research funding to public universities relative to private ones. The cross-sectional analyses suggest that enhanced monitoring and government learning are mechanisms through which GASB 34/35 affects funding decisions. Specifically, the decline is more pronounced among institutions with greater pre-reform administrative and academic complexity, and federal funding to public universities becomes more responsive to prior private gifts and state funding. My results further show that, after GASB 34/35, public universities experience a decline in publications relative to private universities, while both patent and publication output per unit of research expenditure increase. Additionally, I document a decline in faculty compensation across academic ranks at public universities. Overall, these results show that GASB affects government investment in the higher education sector.Risk on the Run: Public Pensions, Migration Risk and Pivot to Alternatives
Abstract
How does local debt sustain with a mobile tax-base? This paper provides new evidence on how pension underfunding shapes inter-state migration. Exploiting plausibly exogenous variation in funding gaps driven by relative portfolio performance, I find that widening pension deficits significantly increase net out-migration of households and income flows from underfunded to better-funded states. This mobility erodes the tax base, raising the effective cost of funding pensions and constraining governments ability to close shortfalls through higher contributions. I further show that states facing higher migration risk - those that face larger out-migration when local taxes rise systematically under-contribute relative to actuarially required levels and shift pension assets toward opaque, higher-yielding alternatives such as private equity and hedge funds. This pattern reveals a novel mechanism by which household mobility exacerbates both fiscal fragility and risk-taking incentives. To interpret these findings, I develop a spatial model of tax-setting and pension investment under endogenous migration. The model highlights a migration-funding-risk feedback loop: governments that fail to internalize out-migration overestimate tax capacity, under-contribute, and induce fiduciaries to pursue riskier portfolios.When Price Stops Clearing: Valuation Disagreement and Housing Illiquidity
Abstract
When a housing market turns, transactions collapse while prices barely move. I trace that quantity adjustment to the individual home. Using 567,638 completed Beijing resales and a cohort of 16,108 listings observed while still for sale, I compare homes within a community and month that differ in how hard they are to value, measured without any information from the homes own price. The hardest-to-value home takes 16 to 27 percent longer to sell, is 4 to 7 percentage points less likely to sell within thirteen weeks, and, among homes observed for sale, is 3 to 8 points less likely to sell at all. It sells for less, at a larger and more dispersed negotiated discount, and earns no return premium at resale. Price stops working where value is unclear: the elasticity of selling time to the asking price falls five-fold from the clearest submarkets to the noisiest, as a search model with risk-averse, disagreeing buyers predicts. In the textbook no-short-sale asset, disagreement should raise the price; here it is borne as illiquidity.Shedding Light on Bias: Consumer Complaint Disclosure and Racial Equity in Financial Services
Abstract
This study investigates how the disclosure of consumer complaint narratives in 2015 by the Consumer Financial Protection Bureau affects racial disparities in financial services. Using triple-difference estimation, we show that minority consumers receive better treatment from financial institutions under CFPB supervision after the disclosure. These improvements manifest as higher deposit rates and lower fees in savings markets, and reduced rates for auto loans and credit cards in lending markets. Financial institutions receiving discriminatory complaints face intensified local competition and deposit outflows. Our evidence underscores the broad impact of service quality disclosure in narrowing racial disparities across savings and lending markets.AI Exposure, Financing Frictions, and Long-Term Labor Commitments: Evidence from Green Card Sponsorships
Abstract
Do advances in artificial intelligence (AI) change firms willingness to make long-term commitments to talent? We use employment-based green-card sponsorships for skilled workers as a direct measure of durable human-capital investment. We find that a one-standard-deviation increase in occupational exposure to AI is associated with roughly a 10% decline in sponsorships for the affected occupations, consistent with firms exercising the option to wait under technological uncertainty. The contraction concentrates in entry-level roles and is strongest among financially constrained firms. By contrast, R&D-intensive and AI-innovating employers keep overall commitment levels roughly unchanged by shifting sponsorships toward managerial and soft-skill-intensive positions. Taken together, these findings show how technology shocks-measured as differential task-level exposure to AI across occupations-interact with capital-market frictions and innovation capacity to shape both the level and the mix of long-term human-capital investment.What is a Central Bank Worth
Abstract
About 5.5% of GDPinmoderntimes, muchless than prior estimates for the present value of seigniorage. I adapt the methodology of Jiang, Lustig, Van Nieuwerburgh & Xiaolan (2024) to develop an asset pricing model that prices the Bank of England (BoE), excluding quantitative easing. I leverage the fact that the BoE was publicly traded until 1946, using the observed price to assess the accuracy of the model. The models relatively modest errors stem from the strong mean reversion embedded in the VAR that forecasts the BoEs dividends. Having areliable estimate for the value of the central bank since the end of WWII, I assess its contribution to the governments fiscal capacity. On average the central bank has accounted for 8.5% of the market value of gross government debt. This share increases with inflation, peaking at 21.3% in 1975. Note for AFA Poster session: Ongoing work extends the benchmark exercise by pricing the payoffs from QE and applying the framework to more than 40 central banks, for all of them I have digitized profitability data since their establishment.Understanding Risk: Lessons from Mutual Fund Disclosures
Abstract
Do mutual fund disclosures reveal risks that matter for asset prices? I use risk disclosures to show how the corporate finance of asset management shapes asset pricing. Disclosed market, size, and value/growth risks align with return-based factor exposures in sign and magnitude, establishing that the disclosures are credible and informative. I then construct disclosure-sorted return spreads to measure intermediary-level risks, including redemption pressure, concentration, mandates, management, and operations. These risks are not spanned by standard factors and are priced in fund returns. Redemption and concentration also price individual stocks, showing that organizational frictions in asset management have asset-pricing consequences.Who do regulators care about? Recovering welfare weights from bank merger decisions
Abstract
How much do banking regulators care about borrowers and depositors relative to bankers and public insurance costs? We combine novel data on bank merger applications with a structural model of merger review to quantify regulators welfare weights on different agents. First, we show that regulators rarely deny applications; when they do, denials target risky banks. Second, we show that regulators do not deny mergers that raise anti-competitive concerns; instead, they impose branch divestitures. Estimating our merger model on filed and potential applications (to address selection concerns), we find that, normalizing borrower welfare to 1, regulators assign relative weights of 0.47 to FDIC losses, 0.02 to depositors, and approximately zero to bank profits. The results are consistent with a model of regulators acting in the public interest, rather than primarily protecting the interests of the regulated (Stigler, 1971).Buy or Build? How AI Shapes Corporate Acquisitions and Innovation
Abstract
We construct a novel classification of firms' AI exposure to show that, following a shock to the returns of AI capabilities, acquirers with greater AI exposure significantly increase acquisitions of small, young, and AI-producing targets, while acquisitions of AI-adjacent targets decline. The response is strongest among firms least able to build AI in-house. These acquisitions also raise innovation, with gains extending beyond AI into acquirers core domains---a distinctive pattern relative to previous technology shocks. Our results indicate a reallocation rather than an expansion in deal-making, showing how firms use acquisitions to adjust to AI-driven technological change.Product Life Cycle Heterogeneity in the Transmission of Uncertainty
Abstract
I show that firms at different stages of the product life cycle respond differently to increases in uncertainty, and that this heterogeneity matters for understanding the aggregate implications of uncertainty for the economy. In response to higher uncertainty, early-stage firms increase investment, leading to greater innovation output (e.g., product patents). In contrast, late-stage firms reduce both investment and innovation. Using highly granular product-level data and a novel identification strategy that exploits firms' product portfolio differential exposure to aggregate volatility shocks, I find that when uncertainty rises, early-stage firms increase product entry, shifting their portfolios toward newer products, whereas late-stage firms scale back portfolio adjustments. A real-options model with product life-cycle heterogeneity rationalizes these patterns: uncertainty boosts investment, innovation, and product experimentation when firms hold valuable growth options but discourages such activity when projects are mature and less scalable. A structural estimation of the model shows that early-stage expansion offsets roughly one fifth of the investment contraction among late-stage firms. My findings reconcile mixed results in the literature and provide the first evidence that a firm's product life-cycle stage plays a central role in shaping how economic uncertainty affects corporate investment and product markets.When AI Reads the 10-Ks: The Effects of ChatGPT on Trading and Price Dynamics
Abstract
This study examines how ChatGPTs release reshaped the U.S. equity market by democratizing information processing. We devise a metric, CSA, to capture ChatGPTs ability to interpret financial disclosures. Using a difference-in-differences design, we show that high-CSA firms exhibit significant declines in informed trading, adverse selection, and high-frequency volatility. We also find increased order imbalances primarily driven by retail investors, signaling their synchronized reactions to public information. These suggest that generative AI narrows cognitive gaps between retail and institutional investors, improving market quality without accessing private information. However, heterogeneity persists across sectors, highlighting regulatory implications for disclosure standards in AI-augmented environments.Smart Contracts, Dumb Money: Open Source Lemons
Abstract
The most transparent asset market in history mints more new assets in a day than its largest auditor has examined in a lifetime. We show that attention rather than disclosure governs verification. Investors turn public code into noisy quality signals, and scarce attention admits only some projects into a consideration set where audits and funding can occur. Complexity keeps a project from competing for attention on quality, so in hot markets sound but complex projects fall out of the queue and their financing collapses. A constrained planner in this trap gains nothing from subsidizing entry or cheaper audits; only interventions that make projects processable reach the binding margin, and the trap destroys roughly a third of market surplus at the 2024 calibration. Around the January 2024 pump.fun launch, audit adoption among newly listed complex Solana protocols fell twenty-one percentage points against Ethereum-side peers while venture composition held still, so verification deteriorated before funding, as the model predictsDo Policymakers Respond to Science? Evidence from Climate Policy Stringency
Abstract
We study whether policymakers respond to climate science by examining how projected future temperatures affect climate policy stringency. Using temperature projections from 110 coupled oceanatmosphere general circulation models, we show that regions facing larger projected temperature increases adopt more stringent climate policies. Our identification strategy exploits within-region changes in temperature projections over time that arise from plausibly exogenous improvements in computing power and advances in climate modeling. The positive effect of projected future temperatures on policy stringency is stronger in regions where projections exhibit lower dispersion across models and where public belief in climate change is higher. At the individual level, policymakers with stronger beliefs in climate change are more likely to incorporate climate projections into their voting behavior, whereas age, gender, and education appear to be less relevant.Decentralized Autonomous Organizations and the Limits of Shareholder Democracy
Abstract
Using Decentralized Autonomous Organizations (DAOs) as a laboratory, I study the trade-off firms face when adopting shareholder democracy. The benefit is that shareholders' preferences can be aggregated in decisions otherwise delegated to management, which I show shareholders value. The cost stems from the shareholders disengaging amid frequent governance solicitation, amplifying large shareholders' voting power. Vote-level evidence suggests these large shareholders use this voting power to unilaterally impose outcomes. The first time they do so signals a breakdown in small shareholders preferences aggregation, reduces token prices and subsequent minority turnout. I rationalize these empirical findings in a model where large shareholders' voting behavior signals their type. Structurally estimating the model suggests that large blockholders extract private benefits when they sway votes.How Does Relaxing Capital Constraints AffectMarket Liquidity? Evidence from Flash Loans
Abstract
We study how sudden expansions in capital constraint affect market liquidity. Exploiting the token-level introduction of flash-loan eligibility as a natural experiment, we show that treated pools experience a persistent increase in depth and total value locked without deterioration in slippage or volatility. Trading activity rises and becomesmore concentrated, while liquidity provision becomes more dispersed and dynamic. To understand the mechanism, we showthat flash loans reshape market liquidity through two forces: by relaxing short-termfunding constraints for arbitrageurs and by strengthening price discovery.Nobody Knows, or Nobody Agrees? Disentangling Uncertainty and Disagreement via Prediction Markets
Abstract
How beliefs about future eventsand the disagreement among themshape asset prices is a central question in finance, but existing empirical measures of investor beliefs are in- direct, low-frequency, or aggregated. We address this using the universe of transaction- level data from Polymarket, a money-backed prediction market, over June 2024 through March 2026. From the same transaction tape, in the common metric of probabilities, we construct two high-frequency primitives: a price-entropy measure of consensus un- certainty and a dollar-volume-weighted measure of investor disagreement. We estab- lish three results. First, the prediction market and the U.S. equity market are informationally integrated in both directions: stocks lead Polymarket within the trading day, while overnight Polymarket innovations predict the next mornings equity open. Second, disagreement positively predicts next-day stock returnsopposite in sign to analyst- forecast dispersionconsistent with compensation for unspanned macro state risk rather than mispricing. Third, consensus uncertainty predicts lower near-term equity volume and realised volatility, with the volatility released as uncertainty resolves, so it governs the timing of repricing rather than its level. A parsimonious heterogeneous-beliefs model rationalises the full pattern: disagreement is priced, while uncertainty is traded.The Sound of Silence: Policy Signals and Risk Premia
Abstract
How does the Federal Reserve signal shifts in the stance of policy through silence? This paper investigates the asset pricing implications of what the Federal Reserve "stops"" saying-so called no longer said. Exploiting the dynamic sequential structure of these statementsForeign Import Competition and Process Innovation
Abstract
I examine how Chinese import competition changes the type of innovation U.S. firms pursue. Using an instrumental variable strategy, I find that rising imports from China lead U.S. firms to shift their innovative efforts toward process innovation, which focuses on cost reduction, rather than the creation of innovative products. This shift is stronger when product differentiation is difficult, when firms operate in ex-ante riskier environments, and for more labor-intensive firms. In contrast, the effect is weaker for firms that already have alternative ways to reduce costs. I also show that process innovation buffers firms against the adverse effects of import competition. Firms with higher process innovation experience smaller declines in profitability and in the growth of sales, assets, employment, and capital when import penetration rises. When Chinese import competition is high, announcement returns at patent grant dates are higher for process innovation-related patents, especially in industries in which product differentiation is limited. Overall, these findings provide new insights on firms innovation-related responses to foreign competitive pressure.The Making of Kings: Venture Capital Allocation and Competitive Deterrence
Abstract
This paper studies the competitive effects of venture capital allocation through Kingmaking deals, abnormally large early VC stage deals made to young startups. Using investor-side capital supply shocks to isolate supply-induced variation in Kingmaking, we find that Kingmaking causally deters close rivals by reducing their subsequent financing. The decline spans a broad set of investors, and suggestive evidence indicates that the investors best matched to close rivals withdraw most strongly. These effects extend beyond financing. Kings subsequently perform better, while more exposed rivals experience worse lifecycle and innovation outcomes. Markets with more Kingmaking finance a narrower set of entrepreneurial ideas, become more concentrated, and produce less innovation. Taken together, our findings show that a specific form of VC allocation can reshape competition among startups and distort winner selection and broader market outcomes. More fundamentally, because capital can itself shape subsequent competition, our findings raise the possibility that VC allocation may reflect not only information about firm quality, but also the strategic value of shaping competition and increasing the probability that the backed startup becomes a winner. In this sense, VCs may seek not only to identify winners, but also to make winners.Political Donations and Public Pension External Manager Selection: Evidence from Pay-to-Play
Abstract
We investigate pay-to-play in public pension fund management: whether private equity and mutual fund companies donate to state officials' campaigns to obtain public pension external manager contracts. Using a comprehensive panel of U.S. public pension funds from 2000 to 2022, we find that firms making political donations are approximately 3.3-6.5 percentage points more likely to become state pension external managers, after controlling for firm and state-year fixed effects. Propensity score matching confirms this finding across multiple matching specifications. Using the SECs pay-to-play rule (Rule 206(4)-5, effective 2011) as a quasi-natural experiment, we find that the SEC rule eliminated the pay-to-play relationship in high-corruption states (triple-differences). On the intensive margin, donations are associated with higher management fee ratios and greater externalized asset values. Finally, pensions with independent investment boards exhibit a significantly weaker association between donations and external manager selection, suggesting that governance quality mitigates pay-to-play incentives.Pricing Climate Risk in Mortgage Markets: Lender Responses and Regional Redistribution
Abstract
The U.S. mortgage market features substantial geographic heterogeneity in local climate risk, while Government-Sponsored Enterprises (GSEs) fund a large share of mortgages at nationally uniform securitization fees. Uniform pricing is consistent with the GSEs mandate to support broad access to affordable mortgage credit, but it generates cross-regional monetary transfers. Local risk pricing improves risk-based fairness, but may impose substantial welfare losses on borrowers in exposed markets. This paper studies how lenders shape this trade-off in the case of sea-level-rise (SLR) risk. Exploiting the distinction between GSE-eligible conforming loans and jumbo loans that are primarily retained on lenders balance sheets, I document that lenders respond to differences in mortgage-origination costs along two margins: they adjust mortgage interest rates and reallocate funding between balance-sheet retention and GSE securitization. I then estimate a structural model of mortgage demand, lender competition, and endogenous funding choice. Under uniform pricing, monetary transfers flow from inland to high-risk coastal borrowers. These transfers are concentrated among a small group of exposed borrowers, are regressive across the income distribution, and are larger in more competitive markets. A counterfactual in which the GSEs price local SLR risk eliminates these transfers, but reduces borrower surplus substantially in high-risk coastal areas, while national aggregate effects remain small. Commercial banks partly mitigate the local adjustment by retaining more conforming loans on their balance sheets. The results quantify a central policy tension between regional fairness and local welfare under the GSEs affordability mandate.Pre-FOMC Uncertainty Accumulation: Evidence from 0DTE Options
Abstract
Using 0DTE SPX options, I show that the option-implied uncertainty of the prospec- tive shock accumulates rather than resolves before scheduled FOMC announcements. The implied uncertainty monotonically rises from 10:00 ET to the 14:00 ET release and largely collapses at the announcement. Both the level and the accumulation of the implied uncertainty strongly forecast realized volatility in any window after the announcement time, but not prior to it. The implied skewness is uniformly nega- tive, varies little, and does not predict returns. Immediately after the introduction of Wednesday-expiring 0DTE options, the pre-announcement drift documented in the literature disappears, indicating that the drift reflected compensation for previously unobservable event uncertainty. The same accumulation-and-release pattern appears around FOMC Minutes and the closing auction, scaled to their information content. The 0DTE option market prices any regularly anticipated concentration of price discovery in the same accumulating fashion, whether macroeconomic or microstructural.Belief Distortions, Asset Prices, and Unemployment Fluctuations
Abstract
This paper studies the dynamics of asset prices and unemployment when expectations deviate from an objective benchmark. Using machine learning forecasts as a benchmark for objective beliefs, I quantify distortions in survey forecasts of corporate cash flows. Survey forecasts overreact to cash flow news, while machine forecasts do not. These belief distortions explain over 60% of the variation in hiring at both the aggregate and firm levels. Following idiosyncratic shocks to their cash flows, firms with distorted beliefs adjust their hiring excessively relative to firms with objective beliefs, while their stock returns initially overshoot and subsequently reverse. A search model in which firms learn about cash flows with fading memory reproduces not only the overreaction in beliefs but also the volatility in asset valuations and unemployment. Distorted beliefs that raise asset valuations also raise the value firms attach to new hires, leading stock prices and hiring to move together.MEASURING FIRM-LEVEL CURRENCY EXPOSURE
Abstract
We develop DoEx, a firm-quarter measure of currency exposure, built from how intensively managers and analysts discuss the currency environmentexchange-rate levels, volatility, and the dollars directionon quarterly earnings conference calls for 8,294 U.S.-listed firms over 20072024. DoEx tracks every major dollar episode of the period and rises with firms real international footprints, and carries information well beyond alternative currency-exposure proxies, including foreign-sales ratios and return-based FX betas. Firms with higher exposure experience equity-price declines when the dollar appreciates, and they act on this exposure by using more FX derivatives and tilting their borrowing toward non-USD-denominated debt.Flow Hedging and Revealed Beliefs
Abstract
This paper measures how the flow risk created by delegated management distorts what portfolios reveal about beliefs. Active funds earn fees on assets under management, and the flows that move those assets share a common component. Managers hedge this flow risk through their holdings, and the hedge is state-dependent: it strengthens with the conditional variance of the common flow factor. In US fund holdings, the rotation away from high-flow-beta stocks appears within fund and quarter as lagged flow variance rises, responds to flow risk rather than to aggregate volatility, and is an order of magnitude weaker in index funds facing the same flows. Because the hedge responds to flow risk rather than to expected returns, reading holdings as beliefs misattributes it as pessimism, and the misattribution scales with flow variance: naive holdings-implied expectations look most bearish when flow risk peaks. A structural framework recovers beliefs from holdings net of the estimated hedge; a first pass puts the misattribution near 0.7 percent per year. Part of what holdings reveal as pessimism is hedging.Private Equity Add-ons, Low Rates, and Local Market Concentration
Abstract
Private equity (PE) add-on acquisitions have become central to U.S. buyout activity, but their competitive effects are hard to measure because targets are often private, small, and local. I build a comprehensive U.S. buy-and-build panel linking 35,118 add-ons, 8,256 platforms, and 2,292 sponsors over 1990-2024, and study their effect on local market outcomes. I find that, first, compared to traditional M&A, PE add-ons are more serial, horizontal, and local. Second, while the effect on national level concentration is muted, PE buy-and-build contribute significantly to higher local market concentration, especially in non-tradable markets and under favorable financial conditions. Third, a counterfactual analysis attributes 1.37 percentage points (34.7 percent) of the 19902024 local CR4 rise to PE buy-and-build; the non-tradable contribution is 2.84 percentage points (30.1 percent). Finally, local outcomes show more layoffs, but not broad wage declines or firm and establishment losses.On Social Media Exposure and Bank Deposit Sensitivity to Fundamentals
Abstract
I examine how social media exposure affects the sensitivity of bank deposits, especially uninsured deposits, to bank fundamentals. Using YouTube data matched to 70 U.S. commercial banks from 2017-2021, I construct a bank-quarter measure of social media exposure based on views per depositor and merge it with regulatory Call Reports. I then study how deposit growth responds to ROA across banks with different levels of online visibility. Social media exposure strongly amplifies the flow-performance sensitivity of uninsured deposits: the interaction between lagged ROA and exposure is positive and statistically significant, and the implied slope of uninsured deposit growth with respect to ROA increases by about 4.95 percentage points between the 10th and 90th percentiles of exposure. In contrast, insured deposits remain largely insensitive to fundamentals and to social media. The results suggest that social media is an independent information channel that strengthens market discipline in uninsured funding without materially affecting insured, sleepy deposits.The Value of Insider Selling: Trading Flexibility and Managerial Incentive Design
Abstract
I study whether restrictions on insider selling affect managerial incentive design by exploiting the SEC's 2022 amendment to Rule 10b5-1. Using a difference-in-differences design, I compare firms with greater pre-amendment reliance on short-cooling-off plans with firms relying on longer-cooling-off plans. Following the amendment, the share of compensation awarded as options declines, as do managers' wealth sensitivity to stock-price volatility (vega) and pay-performance sensitivity (delta), while total compensation does not increase. These reductions in option awards, vega, and delta are concentrated among firms that relied most heavily on option-based compensation before the amendment. The findings are consistent with insider selling flexibility complementing option-based incentives and suggest that restrictions on managers' ability to monetize equity can affect the design of executive compensation.Employee Peer Pay Inequality and Innovation: Evidence from Salary History Bans
Abstract
We distinguish horizontal (peer-level) employee pay dispersion from vertical pay dispersion across hierarchical levels and examine its impact on innovation. Using the staggered adoption of state-level salary history bans, we show that the laws significantly reduce horizontal pay dispersion but not vertical dispersion. Treated firms experience an 11-16% increase in patent value. The effects are stronger in firms where innovation is more sensitive to horizontal inequality, and policy exposure and hiring are greater. Post-treatment, firms experience greater employee retention, inventor productivity, and team collaboration and performance. Our findings suggest that mitigating pay inequality among employee peers improves innovation.Passive Flows, Active Woes: Passive Investing and the Decline of Active Mutual Fund Alpha
Abstract
This paper studies whether the secular shift toward passive investing has affected active fund performance through flow-induced demand effects. Using U.S. equity fund data (19842024), I document a decline in active fund performance after 2010, with average annual four-factor alpha falling by around one percentage point and the previously positive relation between Active Share and performance reversing. These patterns contrast with theories predicting improved performance as the active sector shrinks. A flow-driven framework shows that capital reallocations toward passive funds generate asymmetric price pressure, penalizing funds active tilts. Flow-induced demand significantly impacts fund returns, with the effects of passive flows persisting over multiple years. Controlling for flow-induced trading can explain the negative Active Shareperformance relationship, suggesting that underperformance reflects structural demand headwinds rather than declining manager skill. Higher-frequency tests exploiting plausibly exogenous, beginning-of-month passive flows provide additional evidence.Lendable Inventory Concentration, Borrow Fragility, and Equity Pricing Anomalies
Abstract
Exploiting equity anomalies requires more than paying the borrowing fee: arbitrageurs must also maintain access to borrowed shares until mispricing corrects. This paper argues that fees capture the current cost of shorting overpriced stocks but not the stability of borrow access. Using IHS Markit securities-lending data, I show that concentrated lendable inventory is associated with greater borrow fragility: more volatile borrowing conditions, more severe fee and utilization tail events, and shorter loan tenure. Stocks with more concentrated inventory incorporate negative earnings news more slowly and exhibit larger fee-adjusted anomaly return spreads, driven primarily by the short leg. The evidence suggests that, beyond borrowing costs, borrow fragility arising from concentrated lendable supply is a distinct dimension of short-sale constraints that helps explain persistent anomaly-related mispricing.Price Caps and Nonprice Adjustments in Home Insurance
Abstract
I study home insurers' responses to pricing constraints. I combine manually collected California rate and product-design changes filings with insurer-product-ZIP-code underwriting records. Proposed rates systematically update to reflect past exposure-related costs but diverge from actuarial fundamentals and bunch at a 7% regulatory threshold. Using this bunching as an instrument, I find that pricing-constrained insurers restrict policy forms, revise underwriting rules, reduce exposures, and withdraw from wildfire-prone areas. Consistent with supply-side dynamics, the insurer of last resort expands where more insurers are price-constrained. Adjustments shift from price to nonprice margins to reduce exposure to inadequately priced risks.Betting on the State: Sports Gambling Legalization and Municipal Borrowing Costs
Abstract
I study how sports betting legalization affects the borrowing costs of U.S. state governments. Exploiting the staggered adoption of online sportsbooks in a stacked difference-in-differences design with secondary market bond-month observations, I find that legalization lowers state issuers' yield spreads by 6 to 10 basis points, roughly 15 percent of the sample mean. This effect is robust across three alternative spread measures, holds under alternative control group definitions, and on a propensity-score-matched sample of bonds. Cross-sectional evidence points to fiscal capacity rather than sentiment: spreads fall about twice as much for general obligation bonds as for revenue bonds, the decline increases monotonically with maturity, and it is concentrated in bonds rated below AA and in states with poorly funded pensions, indicating that investors price legalization as a permanent expansion of the general tax base. The reduction scales with the size of each state's betting market and with the sports betting tax revenue the state actually collects. State tax revenue rises after legalization without an offsetting increase in state spending on health, police, welfare, or hospitals over the sample horizon. Bond investors thus reward a policy that other research shows harms household balance sheets, a tension that any welfare assessment of sin-tax revenue has to confront.Private Money, Public Debt: Quarter-End Price Pressure in Treasury Bills
Abstract
This paper shows that quarter-end bank window dressing creates a substitution channel from private to public money-like assets. Around regulatory reporting dates, foreign banks reduce borrowing from U.S. money market funds; constrained to hold short-term, high-quality assets, the funds redeploy the displaced cash into Treasury bills. As a result, short-maturity U.S. Treasury yields fall by 3-9 basis points at quarter-ends. Using a network shift-share design that exploits predetermined fund-bank relationships and cross-sectional variation in Treasury bill holdings across funds, we causally estimate that money market fund demand inflows equal to 1 percent of Treasury bill market value lower yield spreads by 4.9 basis points. We use this estimate to quantify the fiscal consequences of bill-heavy debt issuance under alternative Treasury financing scenarios.Salient Cues of Economic Transitions in Analyst Forecasting: Evidence from the Electric Vehicle Era
Abstract
How do financial intermediaries recognize shifting business models and investor priorities? In the context of the green transition, we show that salient cues of broad economic trends shape financial analysts attention and effort. Analysts in high-salience areas revise green firm forecasts more frequently, show greater climate-related engagement during earnings calls, and increase opportunity-oriented climate language in reports. Their resulting increased forecast accuracy for green firms reflects not a net gain in forecasting ability, but a reallocation of limited attention and effort. Markets respond more to their green-firm forecast revisions, and the information environment of green firms improves under their coverage.Technological Change and the Information Content of Labor Force Restructuring
Abstract
How do new technologies affect firms' labor force restructuring decisions? I develop a model in which layoffs are strategic choices under asymmetric information about the motive for restructuring, and show that a technological regime shift reduces the informational discount associated with layoff announcements by inducing more high-type firms to enter the layoff pool. I test this mechanism using the introduction of ChatGPT as a GenAI regime shift, and find that firms with greater AI capital are 55 to 64% more likely to conduct layoffs and earn announcement returns that are 5 to 10 percentage points higher. Consistent with the information channel, these effects are attenuated among opaque firms and are expected to be absent in placebo tests using private firms. Finally, firms with greater AI capital do not disproportionately displace AI-exposed workers, indicating that these layoffs are not explained by canonical task substitution.Value Is in the Eye of the Beholder: Strategic Marking in Private Credit
Abstract
I study whether loan valuations reported by private credit funds reflect fund incentives. Comparing valuations assigned by different funds to the same loan at the same point in time, I find that funds report higher loan valuations precisely when their contemporaneous realized performance is weaker. Additional analyses of cross-sectional patterns show that this behavior is consistent with strategic marking that smooths reported returns. The effect is strongest in settings where valuation discretion is plausibly greater and where deviations matter most for reported performance and net asset value. The effect intensifies when funds have stronger incentives to relax binding leverage constraints through higher reported net asset values. The results cannot be explained by other alternative channels such as asymmetric information, portfolio selection, and delayed updating of marks. These findings highlight an agency-based component of valuation in private credit and have implications for the reliability of reported net asset values and for the monitoring of financial stability risks in stress episodes.Patents' Long-term Potentials
Abstract
This paper proposes a measure of long-horizon patent value, Potential, that integrates a patent's structural marginality, inventor team composition, and the post-grant trajectory of technology salience. At the patent level, high-Potential patents exhibit accelerating forward citations over a 20-year horizon and a higher probability of being traded in the patent market. At the firm level, Potential forecasts higher profitability, lower idiosyncratic volatility, and reduced financing constraints. We further show that Potential supplements KPSS (Kogan et al., 2017) by capturing the long-horizon component of patent value: within both low- and high-KPSS groups, high-Potential patents generate stronger long-run citation performance than low-Potential patents. Consistent with the value-discovery role, a higher share of high-Potential patents in a firms recent patent portfolio predicts stronger long-run firm outcomes.What Are Intangibles Worth? Production versus Market Power
Abstract
Intangibles create value for firms by raising productivity and increasing market power. The first channel boosts growth, while the second could weaken competition. This paper quantifies both channels with a dynamic model in which firms invest in intangible capital to raise production and market power. I combine the model with data to decompose intangible value for each U.S. public firm. By 2024, intangibles derive more value from market power than from production, contributing 28% and 21%, respectively, to aggregate market capitalization. Intangible value rises as market-power value grows within firms and firms with greater production-related intangible value gain market share. Over the firm life cycle, it shifts from production toward market power. Incorporating the value of intangible market power restores 60% of the decline in the value premium and explains firms incentives to invest in intangibles.Does Discretized Information Induce Misperception? Evidence from a Sports Betting Market
Abstract
Does the format of information influence how markets process it? In theory, transformations of existing information, such as discretization into ratings, should not alter beliefs. Yet under cognitive constraints, discretized information can distort perceptions even when it conveys no new information. Using a betting market, we develop a novel test that cleanly distinguishes these competing views by isolating the cognitive channel from coordination, regulatory, and outcome-driven explanations, a challenge unaddressed by prior research. We find that discretization induces misperception, particularly under low economic stakes or high reliance on heuristics. Our findings have implications for how regulators and intermediaries disclose information.Customer Capital and Financing Policies
Abstract
Recent research shows that firms invest heavily in building long-term relationships with their customers-referred to as customer capital. This paper investigates how customer capital shapes firms' financing policies. Empirically, I find that customer capital is positively associated with cash holdings and negatively associated with leverage and equity issuance. I rationalize these findings using a dynamic investment and financing model in which firms invest in customer capital due to product-market frictions. Customer capital is valuable but not pledgeable as collateral. Hence, under financing frictions, firms optimally hold more cash and limit leverage to preserve debt capacity and avoid costly equity issuance. Quantitatively, product-market frictions are central to explaining these financing choices: in their absence, firms' average net leverage would be about 25%, compared to -3% in the calibrated model. These findings underscore the importance of considering product-market frictions in explaining financing policies and may help rationalize several puzzles, such as the low-leverage puzzle and the fact that firms hold large cash reserves.Education and Investment Skill: Evidence from Mutual Fund Managers
Abstract
Does higher-quality education make better mutual fund managers? Using course syllabi from 702 managers' colleges, I measure two dimensions of education quality: Relevance, how closely coursework matches investment-management knowledge and skills, and Timeliness, how up to date the ideas and methods taught were. A one-standard-deviation increase in either predicts 80 basis points higher annual net returns. Prestige and selectivity do not explain the results. The two measures capture distinct investment skills: in rich information environments with high analyst coverage, Relevance helps managers use existing information more effectively. In thin information environments with low analyst coverage, Timeliness helps them develop their own analysis.Labor Market Information and Product Market Spillovers: Evidence from Employee Reviews
Abstract
We study whether labor-market information spills over into the product market and influences consumer purchasing behavior. Because job seekers frequently consult Glassdoor to evaluate potential employers and are also consumers themselves, employee-generated information encountered during job search may subsequently affect their purchasing decisions. We exploit peer firms' earnings announcements as shocks to job-search activity, using variation in peer announcement timing to identify periods when employee-generated information is more likely to be salient to local consumers. We find that negative employee reviews reduce product sales more strongly during periods of heightened exposure to peer firms' earnings announcements, whereas positive reviews have no comparable effect. The effect is stronger when peer firms have more active employee review activity on Glassdoor and when local job-posting intensity is higher, consistent with greater job-search attention amplifying the impact of employee-generated information. Our findings identify a novel informational channel connecting labor and product markets.Does Legal Marriage Change Financial Behaviors? Quasi-Experimental Evidence from Taiwan
Abstract
This paper studies whether entering legal marriage changes household financial behavior. The empirical challenge is that marriage timing is strongly selected: couples who marry differ in preferences, resources, expectations, and risk tolerance, so a simple comparison of married and unmarried individuals confounds the institutional content of marriage with selection into it. We address this challenge using Taiwan's 2019 legalization of same-sex marriage, which abruptly extended access to the legal marriage contract to couples previously excluded from the institution while leaving the option set of opposite-sex couples unchanged. We link this policy shock to confidential Taiwanese administrative data covering household registration, income, and individual asset records, which allow us to measure portfolios directly rather than through self-reports. Using opposite-sex married individuals matched on sex, age, birthplace, pre-reform assets, and pre-reform wage income as controls, we estimate difference-in-differences and event-study models around the 2019 reform. We find that legal marriage increases stock market participation by about 1.55 percentage points, but reduces the equity share of liquid financial wealth (by about 2.45 percentage points) and of total measured assets (by about 2.40 percentage points) among individuals who were already stock market participants before the shock. Both responses are robust to alternative control-group constructions, weighting, and inference, and the event-study paths display no differential pre-trends. The findings suggest that marriage affects financial behavior differently across margins: it relaxes entry barriers into risky asset markets while inducing portfolio rebalancing toward lower equity exposure among existing participants. The pattern is consistent with marriage shifting financial decision-making from an individual to a household problem, in which risk sharing lowers the cost of entry while household-level responsibility tempers concentrated equity risk.The Propagation of Arbitrage Constraints: Evidence from Settlement Mismatch
Abstract
This paper studies whether arbitrage frictions transmit across otherwise distinct trades. I exploit the settlement mismatch created by the 2024 US transition to T+1, while many foreign securities held by US-listed cross-border ETFs continued to settle on T+2. Relative to non-US ETFs tracking the same indices, primary-market arbitrage, measured by flow-mispricing responsiveness, declines among directly exposed cross-border ETFs. The decline appears primarily in creations and generally grows with portfolio exposure to T+2-settled securities. Exposed ETFs also tilt toward T+1-settled securities, experience higher tracking error, and become less liquid. Domestic ETFs face no direct settlement mismatch, yet their flow-mispricing responsiveness declines more when their predetermined reliance on authorized participants (APs) active in cross-border ETFs is greater. The decline associated with this indirect exposure is smaller when AP balance-sheet capacity is greater and larger when poor netting coincides with high global exposure. These findings suggest that market segmentation need not contain a local arbitrage friction when distinct trades draw on shared intermediary capacity.Cross Market Pricing in Prediction Markets A Comparison with Derivatives
Abstract
We study pricing efficiency in decentralized prediction markets by comparing Polymarket prices with option-implied prices derived from derivatives. We analyze nearly 5,000 Polymarket contracts written on Bitcoin (BTC) and Ethereum (ETH), two assets for which prediction-market contracts and liquid derivatives markets coexist. We find that Polymarket prices closely track option-implied prices, but systematic deviations remain: they are concentrated in tail and barrier contracts, vary with sentiment, volatility, demand, and market frictions, and imply more probability mass on extreme price movements. To guide our empirical investigation, we introduce a parsimonious limits-to-arbitrage model that formalizes equilibrium behavior in decentralized prediction markets.Barbarian or Vanguard? : Private Equity Buyout and Physician Opioid Prescription
Abstract
This paper studies how private equity (PE) buyouts of physician groups affect physicians opioid prescribing behaviour. Using a difference-in-differences approach, we find that physicians in PE-owned physician groups decrease their opioid prescription rates after the buyout. We further identify regulatory and litigation risks as the primary mechanism: the reduction is more pronounced in states that adopt more stringent opioid monitoring programs and after a key False Claims Act precedent increased litigation risk for PE-backed healthcare investments. Moreover, we document heterogeneous responses across PE firms: PE firms with fewer financial constraints and stronger corporate social responsibility (CSR) preferences are associated with larger reductions in opioid prescribing intensity.Debt as Information Shield: Evidence from Pharmaceutical Startups
Abstract
This paper examines the role of proprietary information in determining startup capital structure. I exploit the 2014 FDA Final Rule, which mandates public disclosure of clinical trial results, as an exogenous shock to the proprietary information environment of pharmaceutical startups. I find that this mandatory disclosure shock is associated with 1.5 percentage point decline in venture debt usage, representing a 36% drop from the pre-shock mean. This flight from debt is driven by the proprietary costs of disclosure: a one-standard-deviation increase in competitor technology similarity amplifies the drop by 1.8 percentage points. Within these competitive markets, startups developing novel drugs drive this shift away from debt. These findings indicate that the information sensitivity of a financial instrument and the competitive structure of the product market jointly determine the startup capital structure, extending the classic "disclosure-financing trade-off"" to the choice of financial instrument."Alternative Investments: The Consultant Effect
Abstract
This paper studies the role of investment consultants in institutional investors private equity manager selection and performance. I show that consultant networks causally affect manager selection: an investors probability of investing with a given GP rises from 1 in 13 to 1 in 9 when that GP is in the consultants network. In explaining heterogeneity in investor performance, consultant fixed effects account for 77% of the variation and increase the models R2 from 0.051 to 0.131. Finally, I document systematic differences in performance between investors with internal teams and those that engage consultants: investors with internal teams invest in funds with IRRs that are, on average, 70 to 80 bps higher than those of investors that engage consultants.Bundling Credit Card and Electricity Bills
Abstract
This paper studies how partnerships between financial institutions and non-financial firms can mitigate information and enforcement frictions in consumer lending. I examine Credito Facil Codensa (CFC), a financial inclusion program jointly operated by Scotiabank and Enel in Colombia. The partnership combines the bank's lending technology with two complementary assets: information on customers' electricity payment behavior and a billing infrastructure through which loan repayments are bundled with electricity bills. Using unique administrative data on credit applications, electricity payment records, and repayment outcomes, I evaluate these two mechanisms. First, I show that utility payment and consumption information substantially improves credit risk prediction, particularly for applicants without prior credit histories. Second, I find that permanently separating loan repayments from the electricity bill increases delinquency and default, indicating that bundled repayment strengthens repayment performance. Together, these findings show that partnerships with non-financial firms can expand consumer credit while maintaining portfolio quality by improving both screening and repayment.Altruistic Capital: Who Owns It and Who Earns Its Returns?
Abstract
We measure a firms stock of altruistic capital as the share of its employees whose professional profiles disclose volunteering, a person-level trait that we treat as fixed, so that all firm-year variation in the measure reflects the composition of the workforce. The stock is a stable firm attribute: firm fixed effects account for 72% of its variation, and it is largely orthogonal to ESG ratings and to managerial language on earnings calls. Firms with more altruistic workforces surface more corporate misconduct, and that relation sharpens precisely when the 2011 DoddFrank whistleblower bounty raises the private return to reporting, our central piece of causal evidence. Yet shareholders capture none of this value: altruistic capital earns no abnormal return, carries no valuation or profitability premium, and provides no protection in a crisis. The resolution is ownership. The asset is embodied in mobile workers who quit at higher rates, accept no wage discount, and abandon the firm precisely when it falls into distress. It is, in this sense, fair-weather capital: real, valuable, and owned by workers rather than by the firms that employ them.Cash Now, Current Later: Energy Investment in the AI Era
Abstract
A long-standing question in accounting is whether long-term contracts should be recognized at completion or as work progresses, and whether that choice changes firm behavior. I study an accounting rule in the utility sector that governs asset recognition and related cash flows. Construction Work in Progress (CWIP) accounting moves the recovery of an asset's cost from after construction to during it, letting a utility bill customers for an asset while it is still being built. Using state-level CWIP adoptions in a staggered difference-in-differences design, I find that utilities respond by building more, starting construction sooner, and completing it faster. However, they also abandon more projects midway, run finished capacity below comparable plants, and keep older units in service longer, a pattern more consistent with over-investment than with investment at the right level. The response is strongest where the regulatory regime is laxer and where the pre-adoption wait for recovery is longest. Credit terms do not change, consistent with a shift in investment payoffs rather than an eased financing constraint. Operational outcomes also deteriorate, with higher outages and lower customer satisfaction. Overall, shifting cash recovery forward raises the volume and speed of investment but loosens the discipline governing which projects finish and how well they run. The results provide insight into the current wave of CWIP adoptions serving AI-driven electricity demand.The Misuse of Knowledge
Abstract
Small information imperfections can fundamentally distort financial market outcomes. Building on psychological foundations, I develop a behavioral model in which investors universally believe in the Efficient Market Hypothesis, termed the "Believer Expectations Equilibrium"" (BEE). The model provides a unified explanation for three central asset pricing anomalies: inelastic asset demandCorporate Earnings Calls and Analyst Beliefs
Abstract
I study how linguistic features of corporate disclosure shape earnings expectations. I develop a methodology that uses large language models to generate counterfactual versions of earnings call transcripts that differ in language while holding underlying quantitative content fixed, creating a form of counterfactual variation that is impossible to obtain experimentally. I first show that textual features from earnings calls improve out-of-sample predictions of analyst forecasts and realized earnings above and beyond a rich set of quantitative fundamentals. I then intensify each transcript along six linguistic features: Forward Guidance, Jargon, Macro Focus, Uncertainty, Sentiment, and Confidence and trace how these shifts move predicted analyst expectations. The results show that analysts move their forecasts in the right direction but weight linguistic dimensions unevenly: they place excessive weight on optimistic and macro-focused language and do not fully incorporate information embedded in language emphasizing risks and uncertainty. Linguistic shifts also change which fundamentals drive predictions: optimistic language amplifies the role of recent earnings surprises, while risk-laden language diminishes it, evidence that presentation shapes not just the level of analyst beliefs but the information they attend to.Collateral Dynamics: Hurdles to Intangible Investment
Abstract
This paper quantifies how collateral distorts the composition of firms' investment. Using administrative firm data and loan-level information for private Austrian firms, I document that (i) secured credit is pervasive across the leverage distribution, (ii) unsecured loan rates rise with leverage while secured rates remain flat, and (iii) higher leverage is associated with lower intangible but not lower tangible investment. To rationalize these patterns I estimate a dynamic model of tangible and intangible investment in which debt is risky and priced against expected recovery. Tangible assets raise creditor recovery and thereby lower the rates on risky debt, creating an incentive to overinvest in tangibles. Making intangible capital pledgeable would raise its share of total investment by three percentage points among highly levered firms, and by less than one point on average. The estimates imply that low-productivity borrowers over-accumulate tangibles to obtain cheaper financing, slowing their adoption of intangibles.Giving Life to Private (Rated) Credit
Abstract
We study how ratings inflation can undermine financial regulation and inadvertently fuel the growth of privately rated credit. We exploit the 2021 Risk-Based Capital reform for U.S. life insurers, which aimed to curb reaching-for-yield through its treatment of credit ratings. Following the reform, more exposed insurers - especially those with tighter capital constraints - shifted toward privately rated bonds. These bonds exhibit within-issuer ratings inflation and offer higher yields within rating categories, consistent with greater underlying risk behind similar regulatory labels. Accounting for this inflation substantially attenuates the reforms apparent improvement in portfolio risk. Despite targeting ratings rather than market structure, the reform indirectly increased demand for private bonds. Consistent with this demand shift, firms more connected to exposed life insurers increased their private debt issuance.Text-Based Connected Stocks for IPO Firms
Abstract
The IPO firms capital structure decisions, along with its short- and long-term IPO returns, are crucial for investors decision making. Therefore, I study the extent to which an IPO firms capital structure and returns are explained by text-based similar public firms. I identify public comparables from semantic similarity between issuers S-1/F-1 business descriptions and seasoned firms 10-K disclosures. Connected firms leverage predicts IPO leverage beyond industry averages. Their pre-IPO returns predict higher first-day returns but lower one- and two-year abnormal returns; offer prices incorporate only part of this signal during bookbuilding. Connected firms idiosyncratic volatility predicts post-listing issuer volatility and pricing uncertainty. Economically similar public firms therefore contain information relevant to the financing and pricing of firms entering public markets.Taxing Internal Capital Markets and the Repatriation of Risk
Abstract
Taxing internal capital markets can curb profit shifting, but it can also disrupt valuable risk sharing within multinational firms. We study this trade-off in the U.S. property-casualty insurance industry using the 2017 Base Erosion and Anti-Abuse Tax, which raised the cost of cross-border affiliate reinsurance. To separate the motives behind these internal transfers, we exploit the 953(d) election, which allows a foreign affiliate to be taxed as a U.S. corporation while remaining offshore, decoupling tax status from location. We estimate a policy-induced substitution ratio: for each dollar the tax drove out of the non-electing channel, about a fifth returned through a fully taxed 953(d) affiliate. This substitution was concentrated among groups with high exposure to catastrophes that most require global diversification. The contraction in internal risk pooling pushed insurance risk back onto U.S. balance sheets, raising domestic retained losses. However, the reported capital ratios remained unchanged because a decrease in the risk charge on recoverables offset the higher charge from greater risk retention. We document how this mechanical offset creates a regulatory measurement blind spot, as the credit risk charge tracks the raw volume of recoverables rather than capturing the true economic shift in repatriated risk.Estimating Asset Demand: Granularity and Measurement Error
Abstract
Previous applications of structural demand systems to asset pricing often imply puzzlingly inelastic, or even upward-sloping, demand curves. I show that these pathologies arise mechanically from the granular-portfolio condition: the logit demand link weights each positions contribution to identification by its squared portfolio share, so the effective in formation is governed by portfolio concentration, not the raw number of holdings. Standard practice aggregating heterogeneous mandates into a single observed portfolio therefore does not increase statistical power, but instead induces convexity. Linear specifications applied to this convex object distort price coefficients and overstate latent demand. A granular estimator yields well-behaved, downward-sloping demand curves without imposing sign restrictions; its implied aggregate price elasticities are two to four times larger than previous estimates.Reaching for Habitats: A Stabilizing Channel of Monetary Policy Transmission
Abstract
This paper shows a broad class of institutional investors hold persistent duration habitats and actively rebalance to maintain them. Monetary shocks reprice long bonds more than short ones, pushing portfolio durations off target and triggering trades against the repricing. This forms a stabilizing channel of monetary policy transmission, Reaching for Habitats, opposite to the amplification channels emphasized in the literature. Using a comprehensive panel of institutional bond holdings, I show that investor types from mutual funds and insurers to pensions and variable annuities actively offset most of their price-induced duration drift within a few quarters, both in Treasuries and corporate bonds. I construct a bond-level measure of exposure to duration-rebalancing demand and find that around FOMC announcements, bonds with higher exposure to strong rebalancers exhibit dampened price reactions to monetary surprises. The gap grows with maturity, builds for two weeks, and closes by the next meeting. A long-short portfolio on this exposure earns an annualized alpha above 3% when held for two weeks after each FOMC meeting. A preferred-habitat model with a soft penalty on duration deviations explains the channel and decomposes the rebalancing into a habitat-anchoring and a mean-variance margin. Estimated fund by fund, habitat anchoring accounts for nearly 90% of the rebalancing. The calibrated model matches the empirical regression on the scale of the long-end dampening: per 100 basis points of monetary shock, the channel dampens the ten-year yield response by about 6(5) basis points in Treasuries (corporates).Inheritance Taxation and Capital Misallocation
Abstract
Inheritance taxation creates a liquidity obligation when wealth is transferred to the next generation. However, entrepreneurial wealth takes the form of illiquid shares in private firms. We study how this liquidity mismatch distorts capital allocation using Norways 2014 abolition of its inheritance and gift tax. Linking the universe of Norwegian private limited liability companies to their majority owners via a shareholder registry, we exploit whether the owner had children as a shifter of their firms inheritance tax exposure in a difference-in-differences design. We document two margins of misallocation. First, the tax induced a within-firm precautionary liquidity distortion: owners anticipating succession-related obligations tilted firm portfolios toward liquid assets, crowding out productive investment, particularly where firms were financially constrained. Second, across firms, the tax fell disproportionately on large and tangible-asset-intensive firms, which faced higher effective rates because of valuation differences; its removal reallocated investment toward higher-return firms and away from lower-return ones.Rivals or Partners? Private Credit and the Reorganization of Banking Intermediation
Abstract
I study how the entry of private credit reshapes bank functions in credit intermediation, using novel hand-collected data on private credit investments from SEC filings and a regulatory reform that relaxes leverage constraints for private credit lenders. As private credit expands into the small- and middle-market corporate lending segment, small banks defend their core markets by lowering loan rates, loosening underwriting standards, and expanding credit to small businesses. In contrast, large banks retreat from direct lending to firms and move upstream to supply capital to private credit funds, enabling them to capture higher returns on capital through favorable risk-weight treatment. Banks with weaker balance sheets have stronger incentives to engage in this regulatory arbitrage. My findings show that private credit both competes with and complements banks, and becomes another channel through which large banks gain exposure to segments that small banks traditionally serve.A Preferred-Habitat Model of Term Structure With Dealer Market Power
Abstract
The repo market finances approximately 70% of U.S. Treasury holdings. Recent evidence documents substantial dealer market power in this market. This paper examines how this market power shapes the term structure of interest rates and monetary policy transmission. I develop a preferred-habitat model in which dealers monopolistically compete for short-term funding to finance carry trades in long-term bonds. Market power in the funding market enables dealers to strategically restrict their carry positions and extract higher spreads between bond returns and funding costs, thereby generating elevated bond risk premia. I show analytically that this mechanism amplifies the sensitivity of risk premia to the yield curve slope and strengthens the transmission of monetary policy shocks to long-term yields. Empirically, I construct a measure of dealer market power using the Herfindahl-Hirschman Index (HHI) of repo borrowing concentration from money market funds. Consistent with the model's predictions, higher repo market concentration is positively associated with elevated term premia (both in ex-ante expectations and in ex-post realized bond returns), larger Fama-Bliss coefficients, and greater sensitivity of long-term yields to short-rate shocks. These findings reveal that dealer market power in the repo market plays a crucial role in shaping bond risk premia and the transmission of monetary policy through the yield curve.When Safe Assets Stop Looking Safe: Global Forces in Sovereign Bond Yields
Abstract
Using historical and high-frequency data for 20 advanced economies, I show that global risk-free rate models capture the secular rise in sovereign yield comovement over the postwar era, but their explanatory power breaks down at long maturities after the Global Financial Crisis. I address this challenge by decomposing sovereign yields into risk-free rates, default spreads, and convenience yields, each with global and local components. The risk-free component remains the dominant global force, while global credit risk emerges as the missing factor behind post-crisis long-rate comovement. Its role is strongest for stressed sovereigns, yet it is only weakly linked to broad U.S. financial indicators, suggesting that global sovereign credit conditions are not simply a reflection of U.S. factors. Convenience yields are also important determinants of yield variation at the country level, but not comovement across countries. Finally, I document a striking erosion in safety premia across the maturity spectrum: convenience yields fall first in the United States and then more broadly across countries after 2024.The Price of Staying Unmonitored: Strategic Incentivization in U.S. Auto Insurance
Abstract
Firms increasingly use behavioral data, yet customers often choose whether to provide them. Using a novel dataset from auto insurers regulatory filings, I study how insurers set prices to encourage customers to share driving data. Auto insurers traditionally set premiums with a limited set of variables, such as age and location, that assess risk only coarsely. Exploiting staggered state approvals of mobile telematics, I find that the transition to mobile monitoring leads insurers to raise the price of remaining unmonitored, particularly for young drivers, with larger increases among young males, and residents of densely populated areas. Repricing is stronger among insurers offering larger enrollment discounts, linking the pricing response to insurers incentives to obtain data from customers whose risks are difficult to assess.The Retail Put
Abstract
Retail investors are traditionally viewed as uninformed noise traders with little effect on aggregate prices. This paper shows that the substantial growth in retail participation has changed that view. Using high-frequency retail flow, I show that retail investors systematically buy following market declines. Their demand is highly asymmetric, increasing sharply after losses but little following comparable gains, resembling the payoff of a put option. To identify the causal effect of retail demand on prices, I exploit a regression discontinuity at the zero threshold in displayed returns. Crossing this threshold generates a discrete increase in retail buying while other market conditions vary smoothly. Instrumenting retail flow with this discontinuity, I estimate that a one-percentage-point increase in net retail purchases raises subsequent 30-minute market returns by 40 basis points, with no sign of reversal. The results suggest that households play a larger role in aggregate risk-bearing than is previously recognized, absorbing selling pressure and reducing short-term volatility.Listen To the Deal
Abstract
Managers often provide soft information that is difficult to verify independently when released. This problem is especially relevant when investors have few outside sources against which to evaluate managements claims. I introduce a new within-call measure of how an acquiring CEOs perceived vocal authority changes from prepared remarks to live analyst questioning. Using 1,734 M&A announcement-call recordings involving U.S. public acquirers, I find that CEOs become less dominant on average in Q&A. Greater dominance retention is associated with lower announcement returns for standalone private targets, where outside information is limited, but not for public targets or subsidiaries. The result comes from the within-call change, is concentrated in returns formed after the call, survives textual controls, and is more pronounced for material deals, acquirers with thinner analyst coverage, and diversifying private acquisitions. Taken together, the evidence is consistent with investors treating retained vocal dominance as an adverse credibility cue when managerial claims are difficult to verify independently.Asset Reallocation under Cap-and-Trade: Evidence from the U.S. Power Sector
Abstract
This paper examines how carbon pricing influences asset reallocation in the U.S. power sector. Carbon pricing alters the value of electricity generation assets, namely power plants, across both plants and firms. Using the staggered implementation of the Regional Greenhouse Gas Initiative and California's Cap-and-Trade Program from 2001 to 2019, I show that, in regulated states, traditional capacity sold increases by about 40.8%, while renewable capacity sold decreases by about 60.3%. This reallocation is concentrated among plants and firms with greater carbon exposure. Owners sell dirtier traditional plants, and firms with more carbon-intensive portfolios sell more traditional plants. By contrast, firms with stronger upgrading capability sell fewer plants and acquire traditional plants that subsequently exhibit lower CO2 intensity. The effects also extend beyond regulated states. Exposed multi-state firms reduce traditional capacity, expand renewable capacity, and adjust self-generation in their broader portfolios. The results show that carbon pricing affects the real economy not only by changing current emissions, but also by reallocating ownership of long-lived productive assets.From Words to Portfolios: Disentangling Narrative and Reality in Mutual Fund Differentiation
Abstract
This paper studies the difference between genuine portfolio innovation and narrative differentiation in mutual funds. Developing a transformer-based portfolio embedding model and combining it with pre-trained large language model embeddings of fund prospectuses, I construct two novel measures of mutual fund differentiation from portfolio holdings and prospectus narratives. Prospectus uniqueness commands higher fees and attracts additional capital when accompanied by an outstanding performance history, while portfolio uniqueness predicts future risk-adjusted performance among skilled managers. The results differ between retail and institutional investors: the flow response to prospectus uniqueness comes primarily from retail investors, while the higher fees it commands are borne mainly by institutional investors. Together, these findings indicate that narrative differentiation in fund disclosures operates primarily as a marketing and pricing tool, whereas real product differentiation through genuinely distinctive portfolios is what generates alpha.Putting Your Money Where Your Mouth Is: Political Polarization, Consumption, and Investment
Abstract
We show that political alignment, the match between a county's partisan composition and the party of the sitting president, shapes realized household financial behavior. Using county-level data on discretionary consumption from the BEA and on equity investment constructed from IRS dividend income, covering every county in the United States over six presidential terms, we find that counties increase both spending and equity investment when they become aligned with the party of the president. The effects are economically large relative to standard determinants of local economic activity and have intensified with the level of national affective polarization. Our estimates are robust to instrumenting alignment with a Bartik-style instrument built from pre-sample vote shares and to border discontinuity comparisons of contiguous counties, and hold at the commuting zone and metropolitan statistical area level. Taken together, these results provide direct evidence that partisan economic expectations transmit into household financial decisions at magnitudes detectable in aggregate data.Do Judges Move Markets? Ideological Shifts in the \[6pt] Supreme Court and Macroeconomic Expectations
Abstract
This paper studies how judicial ideology shapes financial markets' expectations about the macroeconomy. We use unexpected deaths of U.S. Supreme Court Justices as quasi-experimental shifts in the Court's ideological balance. High-frequency asset price responses around these events reveal how markets perceive the macroeconomic consequences of these ideological shifts. We find that a more conservative Court is associated with lower inflation expectations, lower real activity expectations and lower bond yields. A shift towards a more liberal Court induces the opposite responses. The positive comovement between inflation and activity expectations points to revisions in expected aggregate demand. The magnitude of the yield responses is comparable to estimates in the fiscal policy news literature, consistent with the interpretation that a more conservative (liberal) Court leads markets to expect tighter (looser) future fiscal policy. We validate our quasi-experimental setup with sectoral asset price dynamics, such as high-frequency responses of clean energy and health care stocks. Taken together, our findings suggest that financial markets view changes in the Supreme Court's ideological composition as a source of news about fiscal policy.Fresh Data, Stale Positions: The Staleness Trap in Market Anomalies
Abstract
I identify a "Staleness Trap"" in market anomalies: while signals are highly persistentData Ownership, Data Production, and Lending Market Competition
Abstract
When can borrower ownership of bank-produced data improve allocative efficiency in credit markets? We study an incumbent bank that strategically chooses portable-data production before a fintech with a portable-data processing advantage decides whether to enter the market. The bank's non-portable-data endowment and the fintech's analytical efficiency determine whether the bank balances improved screening against stronger competition or chooses a portable-data level that creates enough informational asymmetry to deter entry. Equilibrium portable-data production can be nonmonotonic and discontinuous in both primitives. Borrower ownership improves allocative efficiency only if it induces entry and the resulting diversification benefit outweighs the loss of precision.Geofinance: Following Sovereign Capital Home
Abstract
I study whether and when sovereign wealth fund (SWF) private investments change where recipient firms do business. This departs from the standard separation benchmark, in which financier identity should not determine a firms real operating choices. I develop a geofinance framework in which an investment relationship can induce recipient business activity toward the financiers home country. The model predicts stronger directed activity when recipients have weaker outside options and when the relationship is more valuable for home-country activity. Following SWF private investments, recipients with weaker outside options are more likely to enter SWF home countries than the home countries of non-SWF investors in the same financing round. The response is not explained by firm scale, ownership dilution or round pricing, and is stronger in sectors where SWF home countries have revealed export disadvantages. Similar business redirection appears during the 2007-2008 financial crisis: banks with larger write- downs increased syndicated lending activity toward sovereign-fund home countries on both extensive and intensive margins.A Necessary Lie: Conflict of Interest in Investor-Paid Ratings
Abstract
Investor payment is widely thought to eliminate conflicts of interest in ratings, or information sale more broadly. I show that the conflicts persist when investors have market power and are themselves delegated asset managers with agency frictions. In contrast to dispersed small investors, a large fund with a legacy position can purchase a favorable report that inflates its market value. The provider prefers to sell distorted ratings to the fund instead of dispersed investors in illiquid markets where information leakage destroys rents from broad sales. Although the bias lowers information precision, the fund demanding it can expand the access to informed trading, so tougher regulation of providers can reduce price informativeness.The Going Public Decision, Initial Public Offerings, and Product Market Dynamics: Evidence from Nielsen Retail Scanner Data
Abstract
Going public is one of the most important events in the life of private firms, especially due to the large infusion of cash on favorable terms associated with their Initial Public Offerings (IPOs) and the ability to raise further financing through subsequent sales of equity and publicly traded debt. In this paper, we use the Nielsen Retail Scanner database in conjunction with several other databases to analyze how going public affects the product market behavior of newly public firms (in terms of their product pricing, competitive strategy, advertising, and human capital accumulation), and develop implications for the accounting underperformance of these firms following their IPOs. First, we analyze how going public affects firms' product portfolios and pricing: we find that, after going public, these firms raise product prices on average, and are able to increase their overall sales and expand the geographic availability of their products, while keeping the size of their overall product portfolios roughly unchanged. Second, we analyze how firms' competitive behaviors with respect to other firms in their industry change upon going public, and study whether, after going public, firms are able to grab market share from other firms in their industry that remain private. Third, we analyze how going public affects these firms' product market advertising strategy. Fourth, we analyze whether firms are able to hire more and higher-quality employees, thus upgrading their firms' human capital after going public. Finally, we analyze whether product market performance changes after going public are an important factor driving the well-documented accounting (operating) underperformance of firms following their IPOs. We establish causality through a difference-in-differences analysis around firms' IPOs (using firms remaining private as the control group).When AI-Washing Meets Initial Public Offerings
Abstract
Valuing technological intangibles is central to IPO pricing, yet the literature lacks reliable measures of firms technological footprints before they go public. We address this challenge in the context of artificial intelligence (AI), a general-purpose technology with substantial value-creation potential but difficult-to-verify firm-level inputs. Using granular data on AI-skilled employees and AI patent filings to benchmark AI-related language in S-1 registration statements, we develop a novel framework that classifies U.S. IPOs from 2010 to 2024 as AI-washing, AI-investment, or non-AI firms. We validate this classification using independent evidence from firms post-IPO AI activities and pre-filing regulatory scrutiny. AI-washing firms receive more favorable initial market reactions but significantly underperform after listing, whereas firms with substantive AI inputs earn strong long-run returns. Cross-sectional evidence supports three complementary channels. The value premium channel shows that AI-washing firms benefit more from AI-related valuation premiums; the verification channel shows that unsupported AI-related claims are harder to detect when underlying AI capabilities are more difficult to verify; and the litigation risk channel shows that legal exposure appears to discipline such claims after listing, with suggestive evidence that higher-risk AI-washing firms subsequently scale back their AI-related disclosures. Our findings show how the difficulty of verifying emerging technologies enables opportunistic disclosure to distort IPO valuation and provide a new framework for assessing whether pre-offering technological narratives are supported by substantive investment.Technology Shocks, Human Capital, and Financial Decisions
Abstract
Broad technological change is hard for workers to interpret: a new technology can displace some tasks, complement others, and revalue skill bundles before its effects reach an individual's own career. How do workers assess their own human-capital risk while the transformation is still unfolding? I study this using AI-related layoffs at technology firms, which turn a diffuse shift into a local signal---retained workers observe which coworkers and skill bundles the firm separates from. Linking layoff announcements to worker skill histories, household records, and property transactions, I ask whether workers act on this displaced-coworker signal in a costly, illiquid housing commitment. Comparing high- and lower-similarity stayers within the same event, benchmarked against ordinary layoffs with event-year fixed effects, I find that high-similarity stayers reduce their home-purchase probability by 32% after AI-related layoffs, near zero after ordinary ones. Because stayers keep their jobs, the response reveals perceived human-capital risk before any job loss is realized.ETF Growth and the Composition of Price-Setting Capital
Abstract
Between 2010 and 2024, a large amount of capital flowed out of active mutual funds and into passive mutual funds and passive ETFs. While passive mutual funds and passive ETFs hold highly similar portfolios, we show that they have fundamentally different effects on market clearing. Using a fund-level characteristics-based demand system estimated from quarterly holdings, we find that passive ETF demand is much more elastic than passive mutual fund demand, with value-weighted demand elasticities of 0.328 and 0.092, respectively, whereas active mutual funds and active ETFs exhibit similar elasticities. Consequently, the two vehicles' importance to price formation does not line up with their importance by size. By 2024, passive mutual funds hold 41% of the assets in our sample but account for only 10% of the elasticity that clears the market. Passive ETFs, by contrast, hold just 22% of the assets but account for 32% of the elasticity. Holding total passive holdings fixed, reallocating one percentage point of passive assets from mutual funds to ETFs leaves aggregate market capitalization nearly unchanged but generates roughly $58 billion of gross cross-sectional revaluation, redistributing valuation pressure across stocks. Despite this difference in demand elasticities, passive ETF demand is at least as informative about firms' future profitability as passive mutual fund demand. Our findings suggest that the rise of passive ETFs has altered how capital flow influences equilibrium prices, not by changing aggregate passive ownership or demand informativeness, but by changing market-clearing elasticity.News Shock to Credit Supply and Business Cycles
Abstract
This paper identifies a news shock to bank credit supply, examining how economic agents respond to anticipated changes in banks lending standards. Using a structural Bayesian VAR, I find that such shock triggers a temporary economic boom, followed by a normal credit crunch when the anticipated credit tightening materializes, which is consistent with preemptive borrowing behavior. Evidence from Euro area confirms that these patterns operate similarly across advanced economies. The news shock remains distinct compared with well-established macroeconomic shocks, indicating that it is itself an important driver of business cycles. Responses vary sharply across loan categories: long-term real estate loan rises, while short-term commercial and consumer loans decline, suggesting that borrowers prefer to lock in favorable terms on longer-term debt ahead of expected tightening. Counterfactual policy analysis demonstrates that responses to the news shock are a self-stabilizing cycle without changes in policy rates, and monetary tightening during credit-news-driven booms can amplify subsequent contractions.Public Capital, Private Selection: Additionality in European Venture Capital
Abstract
Government venture capital (GovVC) is a leading instrument of European industrial policy, yet a first-order question remains open: does it finance firms that would otherwise go unfunded, or does it merely substitute for private capital? Because venture capital is a two-sided matching market, the firms a public investor backs reflect both its own preferences and its competition with private investors---two forces that reduced-form designs cannot separate. I estimate a two-sided matching model of European venture capital (1989--2024) and re-solve the equilibrium with government investors removed. Only 24.9% of GovVC-backed firms would have gone unfunded in their absence; the remaining three quarters are re-absorbed by private capital. Recovered preferences reverse the reduced-form targeting evidence: GovVC favors high-tech firms but loses them to private competition, and later-stage ventures over early stages. The additional firms exit less often than identical private-backed firms (4.88 pp.), but the shortfall vanishes once the geographic mandates are relaxed: the apparent quality gap partly reflects single-market fragmentation, not adverse selection. Around the round, these firms show higher employment but no change in profitability. Together, these results suggest an employment orientation rather than the selection of high-exit frontier firms.The Anachronistic Testing-Date Bias of Machine Learning Strategies
Abstract
We examine whether the documented outperformance of machine learning models in cross-sectional stock return prediction reflects genuine economic alpha or is partly an artefact of an anachronistic testing-date bias: the systematic backtesting of models over periods before their public availability, when investors could not yet deploy the underlying forecasting technology. We assign each of ten machine learning (ML) models a public availability date and measure the bias using value-weighted long-short portfolios, net of transaction costs, over an out-of-sample period from February 2005 to December 2022. The pattern is concentrated rather than uniform. The clearest case is the two-layer neural network, which earns 91% of its cumulative log net wealth before its architecture became public, and whose Fama-French six-factor alpha falls from 0.91% per month (t=2.80) to zero afterwards. The other neural networks decline more mildly, while two boosted-tree strategies show no strong alpha before their availability dates. Pooled estimates and continuous adoption indices built from research output, open-source code, and finance-sector hiring suggest a similar negative effect but are not statistically significant after accounting for cross-model dependence. The strongest evidence that public availability marks a shift in market behaviour comes from short selling: once a neural network model is made public, its short signal aligns more closely with subsequent movements in short interest, a pattern not observed for boosted-tree models. As a result, full-sample ML backtests can conflate implementable performance with profits earned before the relevant technology was available, especially for successful neural-network strategies.Buying the Deposit Franchise: The Inalienability of Customer Capital in Banking
Abstract
A young bank pays sixty-eight basis points more for deposits than an established neighbor in the same county, a gap that persists over a decade. This disadvantage reflects a missing local customer base. Purchasing an established branch is the natural way to acquire that base. Empirically, however, I find that a transferred branch loses about a fifth of its deposits relative to the buyers pre-existing local offices. What leaves the branch is only partly reabsorbed in the county, and that reabsorption is not proportional to market sharesrejecting canonical discrete-choice deposit demand. The unreabsorbed remainder is a permanent loss: net of buyer absorption, fifteen to twenty-two percent of the depositor base does not survive the sale. The proportional loss is flat in seller tenure, indicating that what an ownership change severs is destroyed organization capital rather than reallocated balances. Extending the Wang et al. (2022) model to incorporate inalienable customer capital, I show that a branch is worth strictly less to any acquirer than to its incumbent, generating a buyer-invariant valuation wedge that peaks when the buyer holds no local presence. Applying the calibrated model, I show that when a failed banks deposits are sold to a buyer with no local presence, 1.3 to 2.6 cents of franchise value are destroyed per deposit dollar, equal to four to seven percent of the average cost a bank failure imposes on the deposit insurance fund but recorded as zero in the official cost comparison. Forced branch sales after mergers lose customers the same way, so the buyer retains only 15.3 billion of the 23.0 billion required and delivers a third to a half less competition than the Department of Justice planned. These findings demonstrate that a deposit franchise is only partly buyable: customer capital does not transfer intact with ownership, and standard demand systems and regulatory valuations that treat the depositor base as fully transferable miss the loss.Do Information-Processing Frictions Drive the Local Bias?
Abstract
We show that information-processing costs are an important driver of investors local bias. Exploiting the public release of capable generative AI as a shock that lowered the cost of interpreting and synthesising public disclosures, we examine its effect on mutual funds allocations to geographically proximate and distant firms. Using a difference-in-differences design, we find that funds relative overweighting of local firms declines by about one-third after this shock. The effect is concentrated among firms with lengthy and complex disclosures and among discretionary and non-team-managed funds, where ex ante processing constraints are more likely to bind. Consistent with generative AI relaxing the constraints associated with analysing geographically distant firms, we further find that the stock-picking ability of distant portfolios improves relative to that of local portfolios. These findings show that advances in information-processing technologies can reshape persistent distortions in investors portfolio choices.Reaching for Guaranteed Yield: Universal-Life Insurance and the Pricing of Quasi-Sovereign Debt
Abstract
How does liability-side product deregulation reshape asset demand when financial intermediaries remain constrained by asset-side regulation? We study Chinas 2013-2016 universal-life (UL) insurance boom, during which product liberalization generated rapid growth in costly, surrenderable, deposit-like insurer liabilities. We construct provincial exposure by interacting insurers pre-boom branch networks with subsequent insurer-level growth in UL liabilities. Within province-quarter cells, greater UL exposure selectively compresses spreads on rated non-AAA local-government financing vehicle (LGFV) bonds relative to non-LGFV corporate bonds. A one-standard-deviation increase in exposure implies approximately 3.7 basis points of additional spread compression. The effect is concentrated among bonds with maturities of seven years or less, is stronger in fiscally weaker provinces, and becomes more pronounced under rating-sensitive solvency regulation. These patterns are consistent with a within-rating form of reaching for yield: insurers favor securities offering additional yield relative to perceived default risk without moving further down the formal rating scale. Greater UL exposure is also associated with entry by first-time LGFV issuers. The results show how liability-side deregulation can interact with pre-existing asset-side constraints to reshape credit pricing and market access.Reading between the Lines
Abstract
Prior research shows that underwriter affiliation leads to inflated report recommendations from sell-side analysts, but it omits the fact that underwriter affiliation grants access to issuer-specific information. I study how these forces coexist within the same report. Using Large Language Models (LLM), I extract broker-specific headline recommendations and score rating-blind report narratives using an archive of 97,425 post-IPO reports. Affiliated analysts tilt recommendations more than narratives, creating a within-report wedge. The margin is stronger near the IPO day and among lead underwriters. In joint regressions, investors react to the narrative rather than the recommendation, and the stronger response to affiliated narratives persists when other quantitative signals are included. When the narrative deteriorates, affiliated analysts are less likely to translate that deterioration into a formal downgrade, and the attenuation is stronger when a positive offer-price revision is accompanied by relatively weak prospectus revision. I conclude that underwriter affiliation protects formal recommendations while preserving value-relevant narrative information.Advertising versus Screening? Online-collected data in lending markets
Abstract
Data from borrowers online search and browsing activity can affect mortgage rates in opposite ways: better screening helps lenders assess risk and lower rates, while targeted advertising can raise markups by saving borrowers the cost of searching and capturing part of that value. To separate these effects, I compare similar borrowers served by the same data-using lender across states that restrict online data at different times. Restricting data lowers mortgage rates by 14 basis points for borrowers exposed to targeted advertising, as lower advertising markups outweigh weaker screening. Consistent with weaker screening, the probability of bankruptcy within one year rises by 2.09 percentage points. Consistent with weaker targeting, data-reliant lenders reduce their use of search advertising by 70%. A search model implies privacy can redistribute value from lenders to borrowers, while costly search and weaker screening reduce overall surplus. As mortgages move online, data-use rules become competition rules.The Watchlist Trap
Abstract
A watchlist records a form of attention prior work could not see. Where the literature measures attention in its transient forma search spike, a volume jump, a login, caught near the trade and gone within the weekthe watchlist reveals its persistent form: a self-authored, dated commitment to follow a stock, returned to over time, before any purchase. Using 28,990 Chinese investors whose every purchase is linked to a timestamped watchlist entry, I ask whether routing a purchase through this consideration stage helps. It need not. The stocks an investor watches but does not buy outperform the stocks she buys; yet among the stocks she buys, the ones she watched first (warm purchases) underperform the ones she buys on sight (cold purchases) by about one percentage point of characteristic-adjusted return within investorday pairs, and by more against never-watched purchases. The list's signal is good; the damage is in the conversion from watching to buying. I take the gap apart in two stages. First, composition: warm purchases carry more of the attention-type speculative tendencies this literature prices as costly, which accounts for roughly 45% of the gap. Second, the mechanism that produces the exposurereference-dependent conversion: the daily first-purchase hazard changes slope sharply as the price crosses the level at which the stock was added to the list and the highs and lows witnessed while it waited, identified within (stock, day) cells, so it reflects the investor's own reference point; what investors read does not carry the gap. Extending the trading-tendencies program of Han, He, and Weagley (2025) from which stock retail investors buy to when they buy it, I locate a cost that registry and execution data cannot see: not in watching, but in the reference-dependent conversion the watchlist sets up.Global Innovation Dynamics over the Past Three Decades
Abstract
We measure how knowledge is created, copied, and carried across borders, using the textual information of around 19 million patents granted across the world between 1980 and 2020. From that text we compute three measures of innovation activity: cross-country diffusion, breakthrough and imitation, at the country, technology-field and year level. Together they trace a rich picture of global innovation dynamics. These measures have an advantage over the citation-based measures, because many patent systems do not require citations at all, and applicants who do cite seldom record every source they drew on. Applying a shift-share instrument to bilateral applied tariffs, we find that trade barriers are a powerful brake on the international diffusion of technology, and that the cost falls mainly on the country raising them.When Innovation Outpaces Policy: Hyperscale Data Centers and the Local Costs of Investment Incentives
Abstract
Governments increasingly use tax incentives to attract capital-intensive projects, but their local economic returns remain uncertain. This paper asks two questions: what are the local effects of capital-intensive projects on public finance, employment, local resources, and political outcomes, and how do these effects change when technological innovation increases the scale and resource intensity of subsidized investment? I study these questions in the context of state-level data center tax incentives in the United States, exploiting staggered state adoption and geographic variation in counties suitability for data center development based on proximity to major water bodies. I find that data center subsidies are associated with increases in local government revenues, expenditures, and debt issuance, suggesting that these projects expand fiscal activity while also increasing public spending needs. Employment effects reveal an important tradeoff: overall employment declines, while employment rises in sectors more directly exposed to data center activity, consistent with counties shifting toward capital-intensive development that may crowd out more laborintensive economic activity. I also find evidence of local resource pressures, including increases in developed land prices, declines in farmland prices, and higher electricity rates when data center capacity expands rapidly. Finally, subsidy exposure is associated with greater incumbent electoral losses, suggesting that the local costs of capital-intensive development may have political consequences. Together, the results show that subsidies for capital-intensive projects can generate local fiscal gains, but they also impose employment, infrastructure, resource, and political tradeoffs that become more pronounced as technological change increases project scale.Does Teaching Entrepreneurship Produce Entrepreneurs?
Abstract
I examine whether undergraduate entrepreneurship education increases business formation by exploiting the staggered introduction of entrepreneurship programs across more than 150 U.S. universities. I construct a novel hand-collected dataset combining historical university catalogs with LinkedIn career histories for over 300,000 graduates and implement a within-university, adjacent-cohort difference-in-differences design. Exposure to entrepreneurship education reduces immediate business formation by 10.7% compared to the prior cohort, driven entirely by fewer small businesses, with no effect on growth-oriented ventures, long-run entrepreneurial entry, or longer-run business continuation. The decline is stronger at public universities and among earlier program adopters, consistent with a screening mechanism. The findings suggest entrepreneurship education has limited effectiveness as a policy tool for stimulating new business creation.Who needs a faster exchange? Evidence from agricultural futures electronic trading
Abstract
Like high-frequency traders, exchanges have invested heavily in reducing order-processing latency. Yet little is known about whether lower exchange latency improves market quality or how its benefits and costs are distributed across market participants. Using raw message data from the Chicago Mercantile Exchange (CME), we examine the causal effects of exchange latency on market quality and trading outcomes for liquidity providers and liquidity takers. To address endogeneity, we exploit cross-commodity processing congestion within the exchange as an instrument, whereby message traffic from other agricultural commodities delays the processing of a given order but is otherwise unrelated to its market quality. We find that lower exchange latency primarily benefits liquidity providers by increasing market-making revenues while reducing adverse selection costs, whereas higher exchange latency lowers the immediate trading costs borne by liquidity takers. Because all agricultural futures markets currently share a common order processing system, our findings shed light on a congestion-based externality and suggest that exchanges should account for these trade-offs in market design.Gold as a Hedge Against Anticipated Tail Inflation
Abstract
What Drives the Price of Gold? Gold's textbook inverse relationship with real interest rates holds only over 2000-2021. I show that, controlling for real rates, gold prices are also strongly associated with the right tail of household inflation beliefs, measured by the share of respondents to the Michigan Survey of Consumers expecting inflation above 10% per year over the next 5-10 years. Gold, real rates, and this tail measure are cointegrated, indicating a long-run equilibrium relationship in which gold moves first. I develop a model with heterogeneous inflation beliefs and no short sales on gold, where gold is priced by the most fearful investors and thereby reflects the right-tail belief that inflation goes out of control, while bonds are priced by the average investor. The model proposes an explanation for why real gold price nearly doubled after 2022 even as real rates rose and stayed elevated, and why gold fell immediately after the Volcker shock in the early 1980s while the Treasury market held real rates high for much longer.Trade Diversion and Government-Led Credit Provision: Evidence from China's VAT Rebate Loan Program
Abstract
This study examines how firms respond to trade disruptions and how government-facilitated credit access shapes these responses. Firms increase sales of targeted products to markets without trade barriers, with larger increases for firms with access to a government-facilitated loan program. Firms without access switch into brand-new products and retain larger workforces. The staggered rollout of the loan program eases firms' short-term financing constraints and helps them avoid the profit losses induced by locked liquidity during disruptions. However, the loan program cannot guarantee that a firm survives the disruption; it can only shape how the firm responds.Convertible Debt and Commitment
Abstract
We show that convertible debt, beyond mitigating asset substitution, plays a key role in addressing both the leverage ratchet effect and inefficient default, thus it cannot be replicated by simply issuing plain vanilla bonds and call warrants. To serve the role of a commitment device, the ability in exchanging existing claims for newly issued equity shares must be owned by debt holders, making the conversion clause an inseparable feature. Debt convertibility matters in both good and poor states, where it affects firms capital structure through endogenous conver- sion decisions and restores commitment on corporate debt policies even before any conversion occurs. The capital structure of firm is jointly determined by shareholders dynamic issuance decisions and creditors dynamic conversion decisions. Firm value under convertible debt issuance is improved ex ante due to an anticipated reduction in ex post distress cost and the restored tax benefit of debt. Moreover, issuing convertible debt induces a positive externality that benefits other types of debt holders because the deleveraging effect of conversion in the down state reduces firms credit risk. Our theory indicates that debt accumulation slows down at low leverage and active deleveraging is induced at high leverage, with highly leveraged firms design contract with lower coupon rate and conversion price.Public Learning from Decentralized Prediction Markets: Evidence from Polymarket Earnings Contracts
Abstract
Can small prediction markets inform much larger stock markets? Using Polymarket earnings contracts, we document three findings. First, trading patterns reveal information advantages not fully explained by measured skill or observed public news, consistent with private earnings information entering these markets. Profit concentration, wallet fragmentation, and accuracy patterns across auditor, city, and director networks provide further suggestive evidence. Second, evidence supports transmission of prediction-market signals to outside investors: more profitable wallets receive more profile views, consistent with attention to smart money,'' and noise-trader order flow, whose Polymarket price impact partly reverses, is positively associated with stock returns at the next opening. This association supports cross-market learning: absent such learning, noise-driven price pressure in Polymarket would be unlikely to spill over to stock prices. Third, Polymarket coverage is associated with stronger pre-announcement earnings incorporation. For covered events, the total pre-announcement earnings response coefficient is about 28% of its announcement-window counterpart. This association is stronger when signals are more visible and informative. Together, these findings suggest that prediction markets can make privately held information useful to investors beyond the platform, informing the debate over their public value.What Are the Determinants of Leverage of Nonbank Lenders?
Abstract
This paper studies the determinants of leverage among Business Development Companies (BDCs), regulated nonbank intermediaries that lend to small and medium-sized enterprises. I build a simple theoretical framework and explore information asymmetry, expected bankruptcy costs, and the supply of bank funding to BDCs as possible determinants of leverage. In the cross-section, I find that BDCs with lower expected bankruptcy costs carry higher leverage. I exploit the 2014 Leveraged Lending Guidance (LLG) as a shock to bank credit supply. I also find that BDCs in areas with greater exposure to LLG-affected banks experience a larger inflow of borrowers. Following LLG, treated BDCs increase leverage by 4.2%, accompanied by a 14% increase in borrowers, and a 32% increase in book assets. The expansion in borrowers is accompanied by a decline in the maturity of originated loans, along with a shift in portfolio composition towards senior loans, indicating stricter underwriting standards. These findings suggest a delegated monitoring role that enables BDCs to lower expected bankruptcy costs, supporting higher leverage. Consistent with this channel, BDCs that lever up draw on both bank and bond market financing, suggesting that the leverage expansion is driven by lower bankruptcy costs rather than increased loans from banks.Passive Demand and Missing IPOs
Abstract
We show that the rise in aggregate passive ownership discourages firms from going public, opposite to the effect of stock-level benchmark subsidies. A market-wide switch from active to passive funds reallocates total asset demand from small entrants toward large incumbents and reduces the active risk-bearing capacity available to absorb new listings before index inclusion. In a model of IPO timing, these forces raise small firms public cost of capital and can delay or even eliminate IPOs. Consistent with the model, industries more exposed to rising passive ownership experience larger declines in IPO activity; a Bartik-style instrument exploiting differential exposure to aggregate passive flows confirms this relationship. At the IPO-firm level, we directly test whether depriving an IPO candidate of active demand raises its cost of capital, using post-IPO cumulative abnormal returns (CARs) to measure compensation for idiosyncratic risk. CARs are persistently lower when more expert capital is available, and expert capital attenuates the increase in the cost of capital associated with intermediating anticipated passive ownership before index inclusion. Our findings provide a novel explanation for the long-run decline in U.S. IPOs.Patent Litigation Risk and Innovation Strategy
Abstract
I examine how patent litigation risk affects corporate innovation strategy. Using a Supreme Court decision that exogenously increased expected litigation risk and costs, I analyze how firms innovation strategy changes based on their ex-ante patent litigation exposure. I find that firms with higher ex-ante litigation exposure engage in more exploratory innovation and in less exploitative innovation after the decision. High exposure firms focus more on fields with lower litigation risk and reduce innovation in riskier ones. These strategic adjustments occur alongside an increase in overall R&D intensity. This change in innovation strategy comes at the cost of lower patent value and reduced market valuation. My analysis suggests that exploratory innovation serves as a precautionary response to mitigate expected patent litigation risk in ones own industry.How Insider Incentives in Bankruptcy Design Shape Firm Dynamics
Abstract
This paper examines how incumbent insider payoffs embedded in bankruptcy design shape ex-ante decisions and ex-post outcomes for financially distressed firms. I identify exogenous variation in incumbent payoffs by exploiting provisions in Indian bankruptcy law that barred incumbent managers and large shareholders from bidding to reacquire their own firm in bankruptcy auctions, together with a 2018 amendment that exempted SMEs from these restrictions. Using a triple-difference specification that compares SMEs to non-SMEs, distressed firms to healthy firms, and the pre- to post-amendment period, I document four main findings. First, permitting incumbent insiders to reacquire their firm in auction leads to higher investment ex-ante, consistent with an alleviation of debt overhang. Second, affected firms finance this investment with lower leverage, which raises profitability and revenue growth and increases the likelihood of recovering from financial distress. Third, permitting incumbent reacquisition improves ex-post firm survival rates but has no discernible effect on creditor recovery rates. Fourth, creditor recovery rates increase only in industries with lower information frictions, suggesting that insiders can extract rents from creditors in settings where outside bidders are less informed.AI's Blind Spot: The Mispricing of Breakthrough Innovation
Abstract
While AI models efficiently process standard technological disclosures, we show they systematically fail to recognize radical innovations, with severe consequences for asset pricing. Analyzing U.S. utility patents granted since 1976, we find a non-monotonic, reversed U-shaped relationship between machine-predicted patent grant probabilities (via XGBoost and LLM embeddings) and true economic quality. Crucially, AI models fail in the extreme upper tail, systematically assigning low quality scores to the most transformative breakthroughs. This AI underrecognition reflects genuine economic content rather than model noise: while machine-predicted scores fail to forecast firm performance, the breakthrough innovations missed by AI strongly predict future ROA improvements and higher acquisition target probability. Consequently, this mismatch creates severe market frictions, generating substantial long-term post-patent stock return drift (over 60 to 900 trading days) for firms holding AIunderrecognized breakthroughs, whereas average-scoring patents show no drift. Examining the ex-post shock of Generative AI (LLMs), we show that advanced models reduce information asymmetry for moderate-quality patentserasing their return driftyet remain blind to true breakthroughs. As a result, post-GPT return drift disappears for moderate innovations but dramatically intensifies for the unmapped radical breakthroughs. Our findings demonstrate both the power and limits of AI in financial markets, showing that machine learning progress can inadvertently deepen mispricing for creative disruptions.Codifying the Firm: Organization Capital over the Firm Life Cycle
Abstract
Firms accumulate firm-specific knowledge over the life cycle, but differ in whether it remains worker-owned as firm-specific human capital or becomes firm-owned as organization capital, and therefore in who captures the associated rents. Building on Atkeson and Kehoe (2005), I embed an endogenous codification choice in the firm life cycle. Codification transforms worker-owned knowledge into organization capital that is transferable with the firm, preserves firm-specific capabilities following key-talent departures, and shifts match-specific quasi-rents from workers to owners. I construct two firm-level measures of codification by combining firms' occupational composition with O*NET task information classified using an LLM. Technological codifiability captures how readily a firm's knowledge can be codified, while codification progress captures how much of that knowledge has actually been codified. I show that codification progress rises with firm age and is associated with larger managerial spans. More codified firms provide lower compensation and fewer career opportunities, consistent with reduced rent sharing between workers and owners. Codified firms are also more profitable and less volatile. The framework provides a rent-sharing perspective on artificial intelligence: technologies that expand technological codifiability or lower the cost of codification can shift firm-specific knowledge from workers to firms, changing not only organizational productivity but also the ownership and division of its rents.Public Markets for Claims in Litigation
Abstract
We study how public pricing in litigation finance markets affects plaintiffs' financing decisions and settlement bargaining. A plaintiff sells shares of case proceeds via an intermediary platform whose price reveals case quality. Settlement is modeled as a one-sided private information game in which the defendant makes offers. We show that plaintiffs optimally oversell their claims relative to efficient risk sharing. Reduced risk exposure enables credible commitment to tougher settlement demands, and bargaining gains exceed the per-share price discount. More informative prices temper this overselling, as information substitutes for issuance in strengthening the plaintiff's bargaining position. The welfare implications of public prices are nevertheless ambiguous. Better information improves the allocation of risk but sends more disputes to trial. Endogenizing the mass of investors further reinforces overselling when equilibrium issuance remains above the parity of its market. Larger issues attract deeper markets, and the plaintiff, anticipating the platform's response, sells beyond even his fixed-market optimum.Accidental Landlords: Behavioral Lock-In and Rental Market Spillovers
Abstract
Loss-averse homeowners facing nominal losses are reluctant to sell. A failed sale, however, need not keep them in place: they can list the property for rent and move, becoming accidental landlords. Using property-level data linking sale listings, rental listings, and transaction records in the U.K., I show that paper losses relative to the original purchase price strongly predict failed-sale-to-rent conversions. Rental conversion is sensitive to losses but not to gains, with the change in slope at the original purchase price. The channel operates through the failure of the attempted sale: paper losses reduce the probability of sale, while a roughly constant fraction of failed sales spills over into the rental market. Among properties that convert, owners facing paper losses set higher asking rents. At the aggregate level, regions with higher loss shares experience lower sales volume and more rental conversions. A one-standard-deviation increase in the loss share generates additional accidental-landlord entries equal to about 7% of total annual buy-to-let purchases. Reference dependence in the ownership market therefore affects how the housing stock is allocated between the ownership and rental markets.Levered Safety: Financial Investors in U.S. Public-Private Partnerships
Abstract
Public-private partnerships (P3s) are the primary contractual form through which private capital owns and finances U.S. public infrastructure. Yet, because no standard database decomposes their capital stacks, project-level evidence on their financing remains scarce. I introduce a hand-collected dataset detailing the characteristics, capital stacks, and private sponsors of 1,815 U.S. infrastructure projects. This granular data allows me to test whether private financial sponsors act as risk-bearers or simply intermediate safe, government-backed cash flows. I find that risk allocation depends fundamentally on consortium control. When a financial investor leads, projects are 36 percentage points more likely to utilize availability payments, shifting demand risk back to the public sector. Conversely, when a financial sponsor invests alongside a lead construction firm, the contract resembles the construction firm's profile and carries greater demand risk. Furthermore, financial control substitutes operational risk for financial risk: leverage averages 0.87 in financial-led, availability-payment projects, compared to 0.67-0.71 in all other configurations. Finally, financial capital consistently increases capital stack complexity---adding roughly 0.4 instruments per project---so financing expertise comes with participation, while the contract that removes demand risk appears only where financial capital leads. Although contract and capital structure are jointly determined, this systematic cross-sectional sorting holds robustly across sectors, project sizes, and contract forms.Competing Signals: How Retail Investors Allocate Attention Between Macro and Micro News
Abstract
Retail investors face competing signals when a firm's earnings and major macroeconomic news arrive on the same day. Using granular brokerage data from Singapore, I show that on days with major foreign macro news, the post-earnings surge in retail trading of the announcing firm is about 30% lower than on days without macro news. This effect disappears when macro news contains no surprise, scales with the magnitude of the macro surprise, and weakens when the macroeconomic and earnings signals are directionally aligned. The same investors who withdraw from the announcing firm simultaneously trade more in the assets the macro signal is informative about. Among investors who still trade the announcing firm on macro release days, disagreement falls and order flow becomes more directional. These retained trades outperform the same investors' non-macro-day trades by approximately 2.5 percentage points per annum, a differential that reflects better market timing rather than stock selection.When Nonbanks Dampen and When They Amplify: A Yield-Curve View of the Nonbank Lending Channel
Abstract
We show that the nonbank lending channel is yield-curve dependent. Using German credit register data and a within-firm identification, we find that nonbanks dampen monetary tightening after short-end surprises but amplify it after medium- and long-end surprises. Financial service institutions drive this sign reversal: they expand through a bank capital reallocation channel at the short end but contract at longer segments. Insurers partly offset this pattern, expanding after medium- and long-end surprises as duration-driven solvency gains support their lending capacity. The aggregate nonbank response therefore depends both on the yield-curve segment affected and on the composition of the nonbank sector.The Real Effects of Payment Anonymity: Evidence from the Crypto Ecosystem
Abstract
Does the design of payment systems cause real welfare consequences? Despite the rapid growth of cryptocurrency, the welfare effects of anonymous digital payments remain empirically unknown. I study the staggered rollout of over 30,000 Bitcoin ATMs across U.S. counties, kiosks that convert cash to cryptocurrency with minimal identification, as a laboratory for estimating these consequences. BTM entry increases total crime by 6.9 percent, with effects concentrated in offenses that exploit anonymity, irreversibility, and cash-to-crypto conversion, while crimes unrelated to payment design barely respond. Tracing the causal chain further, BTM entry fuels local drug markets and raises overdose mortality. Crime-induced losses erode household balance sheets, raising credit delinquency in affected neighborhoods. These harms are regressive. Minority and Hispanic communities, young adults, and areas with weaker financial infrastructure absorb the largest increases in victimization and the steepest declines in financial health, simultaneously widening pre-existing inequality in safety, public health, and credit access. To scale these local effects to the aggregate level, I first use on-chain forensic analysis to trace BTM-originated funds along their downstream paths, identifying the routing intermediaries each transaction passes through and the share that terminates at illicit endpoints. I then build an informational model in which anonymity at each stage weakens transaction traceability and raises social harm. Calibrated to these on-chain flow shares, the model translates the local causal estimates into aggregate welfare statements, yielding a national-level measure of the cost generated by anonymous payment infrastructure.Credit Access and Retirement Savings
Abstract
Retirement accounts are illiquid, so a household's access to credit should shape how much it commits to retirement saving by providing liquidity outside the account. Using administrative data linking retirement contribution records to credit bureau files, I examine how credit access affects both retirement contributions and portfolio allocation. I exploit the removal of a bankruptcy flag from credit reports as an exogenous expansion in credit access. A difference-in-differences estimates show that annual retirement contributions increase by $124, approximately 7 percent of the pre-removal mean. This effect is concentrated among higher-earning, credit-constrained households. By contrast, households below the median income primarily use the additional credit to finance consumption rather than saving. Greater credit access also shifts self-directed investors portfolios toward equities and away from bonds by 1.2 to 2.6 percentage points. In contrast, investors who passively invest through target-date funds show no change in portfolio allocation.Political Shifts, Mortgage Drifts: Evidence from U.S. Special Elections
Abstract
This paper investigates how partisan turnover in U.S. congressional special elections affects mortgage lending. Using HMDA data from 1990-2023 and party-switching special elections as shocks to local political control, I show that banks aligned with the incoming party expand credit in the affected constituency: origination volume rises 9.6% at the state level and 14.5% at the district level in the election year. The credit expansion is unlikely to be driven by local demand conditions or investment opportunities. This lending response suggests a quid pro quo channel: aligned banks face lower enforcement intensity and capture more municipal-bond underwriting business after turnover. Mortgage lending also tilts toward conventional, minority, and investor loans, indicating higher risk exposure.Large Language Models, Price Discovery, and Synchronized Trading
Abstract
Large language models can improve investors understanding of corporate disclosures, but reliance on common models can also make their errors more correlated. I study this trade-off using the public launch of ChatGPT and two pre-launch properties of corporate filings: readability and uncertainty. After the launch, order flow becomes more one-sided in hard-to-read firms relative to comparable firms, in both retail and non-retail trading. Price discovery also improves: relative to other firms, weekly return reversal in hard-to-read firms falls by about half, earnings news enters prices faster, and the price impact of order flow becomes more persistent. Non-retail order flow also comoves more across hard-to-read stocks. Greater disclosure uncertainty predicts less synchronized trading relative to other firms, with no corresponding improvement in price discovery. To rationalize these patterns, I develop a Kyle-style model in which delegation improves decoding but introduces shared error. Because trading on information held by others is less profitable, delegation is a strategic substitute and crowding limits adoption. The model explains how greater synchronization can coexist with better price discovery.Future Proves Past: Detecting Hidden Fraud under Imperfect Enforcement Labels
Abstract
SEC enforcement labels are selective and incomplete proxies for the latent fraud they are meant to measure. This mismatch creates two problems: a model trained on them may learn the enforcement process rather than the fraud process, and accuracy measured against them cannot reveal whether a model detects the fraud enforcement misses. We address both with a future-proves-past design. Because accounting distortions eventually reverse, future accounting outcomes modestly refine the training signal into soft labels; we train a convolutional neural network on multi-year accounting panels with these labels and validate it entirely outside the enforcement set. The model preserves SEC-labeled detection performance. Out of sample over 20002019, flagged firms complete seasoned equity offerings at nearly three times the base rate and earn large negative post-issuance abnormal returns, a long-short portfolio earns significant factor-adjusted alpha, and externally screened populations (post-scrutiny Arthur Andersen clients and bank-monitored borrowers) are systematically avoided. An identically specified network trained on the enforcement label, and benchmarks designed to predict it, show none of this profile: the training target drives the difference. The results show that selective enforcement data can be repurposed, at modest cost, to reach fraud that enforcement itself never surfaces, and they caution that detection models should be judged by validation outside the very label on which they were trained.A Macro-Finance Model of Capital Reallocation and Misallocation
Abstract
I quantitatively study the efficiency in capital allocation and financial markets jointly in a general equilibrium model with heterogeneous investors. Leveraged intermediaries allocate wealth across sectors that differ in productivity and risk exposure, and face state-dependent margin constraints. I find that efficiency in financial markets and allocative efficiency in capital are mutually reinforcing. Three forces drive this result. First, efficient capital allocation in the presence of sectoral dispersion requires strong financial balance sheets and efficient risk sharing, and reflects risk-adjusted costs of capital. Second, binding constraints during downturns distort risk sharing among investors and raise risk premia, setting the stage for capital misallocation. Third, the resulting inefficiency in capital composition further affects valuations and growth through risk sharing among investors in constrained states. The mechanisms thus create a feedback loop between efficiency in capital allocation and efficiency in financial markets.The market value of technology spillovers
Abstract
Does the market price technological spillovers? If so, what determines the spillover value of innovation? I construct a patent-level measure of innovations spillover value from peer firms equity reactions to U.S. patent grants from 1926 to 2023. Around patent grants, peer firms in the patenting firms industries earn higher three-day market-adjusted returns than firms outside the affected industries. Returns also increase with firms technological similarity to the newly granted patents. The absolute magnitude of spillover value predicts forward citations. Private patent value is positively associated with spillover value for industry peers. Peer firms' spillover exposure predicts long-run output and employment growth conditional. Spillover effects on growth are stronger in more research intensive industries. Higher spillover exposure is associated with greater investment and acquisition expenditures, as well as increased inventive activity. Spillover pass-through varies with innovation characteristics, industry conditions, and the information environment. It is larger for breakthrough patents, process patents, and technologies in rapidly evolving fields, and weaker in industries with greater market power, tighter financial constraints and larger exposure to data. Measured pass-through is also stronger in larger industries and for patents issued by more actively traded firms, consistent with investor attention facilitating the recognition of external innovation opportunities in prices.Cleaner Credit, Closed Doors: Medical Debt Reporting Bans and Access to Hospital Care
Abstract
Medical debt appeared on roughly one in six credit reports, and policymakers are increasingly restricting its reporting to credit bureaus to protect indebted households. I study the demand- and supply-side effects of removing medical debt from credit reports, using staggered state-level policy changes and administrative data. On the demand side, household delinquencies increase, raising hospitals' uncollectible medical bills by 8.6%. On the supply side, hospitals restrict access to care and, in less-competitive markets, raise prices by 3.6%. Access falls along two dimensions: hospitals admit fewer patients with less generous insurance and cut essential but unprofitable services. Repeat emergency room visits subsequently increase, consistent with deteriorating patient health. These effects are concentrated among low-income households and among financially fragile and non-profit hospitals. Together, the results suggest that medical debt credit protections trigger supply-side adjustments, with the burden falling heaviest on the very households they aim to protect.Belief Distortions and Lending Cyclicality: Evidence from Industry-Specialized Banks
Abstract
This paper studies how industry specialization affects banks belief formation and lending over the credit cycle. I show that when a specialized banks preferred industry experiences a stock market runup, the bank expands lending and reduces loan loss provisions without tightening loan terms, despite the fact that such runups typically precede weaker industry performance. Loans originated under these conditions have a 62 basis point higher default rate than comparable loans made by nonspecialized banks to the same industry. Consistent with a diagnostic expectations framework, earnings call sentiment analysis reveals that specialized banks express heightened optimism toward their preferred industries during booms. The results highlight a downside of expertise in financial intermediation: specialization can amplify credit cycles by distorting beliefs.Debt Relief and the Reallocation of Household Labor and Capital
Abstract
In developing economies, household labor and capital remain concentrated in low-productivity occupations despite the recent dramatic expansion of credit. I show that legacy household debt can itself constrain occupational reallocation, and that relieving it activates the reallocation of labor and capital across activities that credit expansion alone has not. Studying two large 2014 Indian farm-loan waivers that forgave farming debt equal to approximately 4.5% of state GDP, and exploiting quasi-random variation in eligibility across otherwise similar borrowers, I document three sets of results. First, debt relief does not restore farm borrowing. Instead, beneficiary households reduce their reliance on distant wage employment, a margin associated with debt-driven temporary migration. Second, they redirect borrowing from farm credit toward personal and small-business credit that finances entry into local non-farm businesses. Third, districts more exposed to relief exhibit greater business formation and higher nighttime-light activity, consistent with these responses aggregating. The results identify legacy household debt as a financial friction on the reallocation of labor and capital across activities, and show that a balance-sheet intervention can activate this reallocation and support structural transformation.Do Cash Flows Predict Loan Losses Earlier Than Accruals?
Abstract
The debate over when loan losses should be recognized is among the most durable in banking. Yet it has relied almost entirely on accrual-based measures that depend on the creditors judgment. Using commercial mortgage-backed securities, where borrower payments are observable at the loan level, we compare borrower payment shortfalls with the creditors delinquency classification as predictors of loss recognition. We find that the accrual classification predicts write-downs over the following year, while borrower payment shortfalls remain predictive at longer horizons, with evidence extending into the fourth year. Prior payment shortfalls also predict recognition after a loan has returned to current status. Our findings show that accruals sharpen information near recognition without fully preserving the information in underlying cash flows, and point to borrower-level payment information as a potentially useful cash-flow disclosure for financial institutions.When Firms Adopt AI Matters: Diffusion and the Value of AI Deployment
Abstract
Does the value of adopting a general-purpose technology depend on how far it has diffused among competitors? A parsimonious real-options model formalizes two opposing forces: diffusion raises the competitive cost of remaining behind while compressing the scarcity value of firm-specific implementation capability. I identify the first strict public disclosure of AI deployment in 187,191 earnings and investor-call transcripts for 4,313 U.S. public firms from 2011 to 2024 and locate each adopter on a predetermined industry diffusion curve. Firms disclosing when peer deployment remains scarce revalue earlier: the first-mover-minus-follower difference in raw book-assets q is 0.45 in the first two post-disclosure years but narrows to 0.19 and becomes statistically indistinguishable from zero in years four and five. Continuous-diffusion and peer-catch-up tests yield the same ordering. The early differential survives capitalization of internally generated intangibles. Deployment shifts investment toward intangible capital, but operating returns do not show first-mover outperformance within the observed horizon. Competitive timing is therefore a first-order determinant of the market value associated with technology deployment.Why Do Venture Capitalists Bet Young? Conflict between VC and Founder Investment Horizons
Abstract
This paper documents a mismatch between venture capital (VC) funding and entrepreneurial success across founder ages. VC dollars flow disproportionately to young founders, even though the founders who build the fastest-growing firms are typically middle-aged. I find that funds with a shorter remaining investment horizon are more likely to back young founders, and that young-founder companies reach M&A exits sooner. The age tilt becomes more pronounced after a court ruling that weakened investors' ability to force an exit through governance, suggesting that VCs prioritize exit-aligned founders. I then show that the exit speed advantage is concentrated among founders with strong outside options and that, in a within-startup design, young founders' post-exit labor market outcomes are better than those of their more mature co-founders. Overall, the findings highlight founder outside options as an important determinant of exit timing and show how horizon constraints may distort VC investment decisions.JEL Classifications
- G0 - General