Working Papers
with Andrea Eisfeldt, Gregor Schubert, and Bledi Taska
Journal of Finance revise & resubmit (2nd round)
☆ Best paper award at the TCU Finance Conference 2024
☆ Media coverage: WSJ | Bloomberg | VoxEU | Barron's (Frontpage Coverage) | Financial Times | Marginal REVOLUTION
☆ Presentation slides | Data: Occupation Exposure to Generative AI
Abstract
How do recent advances in Generative AI affect firm value? We construct the first measure of firms’ workforce exposures to Generative AI and show that an “Artificial-Minus-Human” (AMH) portfolio earned 5% in the two weeks following the ChatGPT release. The labor-exposure effect is more pronounced for firms with greater data assets and is distinct from the effect of firms’ product exposures to AI. We assess whether exposed workforces are substituted or complemented by Generative AI based on whether their exposed tasks are core or supplemental. Examining firms’ labor demand and profitability following the ChatGPT release supports a labor-technology substitution channel.
with AJ Chen and Zhao Zhang
Journal of Finance revise & resubmit (2nd round)
☆ Presentation slides: Keynote at Bretton Woods Conference
Abstract
How does competition for talent affect firm growth? Using establishment-level occupational employment microdata and job postings, we measure local occupational market tightness and a firm’s corresponding talent retention pressure (TRP) from other firms’ job postings in its talent’s labor markets. TRP predicts job-to-job outflows of a firm’s talent and captures firms’ talent-related concerns. We find that higher TRP significantly reduces firms’ capital investment, increases talent turnover, and lowers productivity. A shift-share instrumental variable approach reinforces these results. Notably, TRP dampens the growth of laggard firms but not superstars, leading to limited impact on aggregate U.S. investment but increased industry concentration.
with Francesco Trebbi and Michael Simkovic
Review of Financial Studies revise & resubmit (2nd round)
☆ Presentation slides: NBER Corporate Finance Meeting
☆ Media coverage: NBER Digest | The Regulatory Review | Fortune | Cato Institute | SLUB | Starling Insights | Marginal REVOLUTION
Abstract
A key question for studying business dynamism is whether the costs of regulatory compliance fall homogeneously on small and large businesses. Using comprehensive establishment-level occupational microdata and occupation task information, we quantify a firm’s compliance costs as the share of wage bill for performing regulatory compliance tasks (RegIndex). We reveal an inverted-U relation between firms’ RegIndex and their size, with 500-employment firms facing compliance costs 40 percent higher as a share of total wages than small or large firms. We further develop a shift-share methodology to disentangle the influence of regulatory requirements and enforcement on driving firms’ compliance costs.
with AJ Chen and Gerard Hoberg
Journal of Accounting Research revise & resubmit
☆ Panel talk: Dow Jones Investor Attention Panel
Abstract
Financial media frequently report the predictions of institutional investors. Using texts of all Wall Street Journal articles from 1979-2020, we measure the intensity of institutional investors’ participation in news production (InstPred) for each industry. We show that InstPred (i) predicts more information production about firm fundamentals, (ii) boosts institutional trading on mispricing, and (iii) accelerates the correction of longer-term mispricing by up to 34% to 62%. Our results are reinforced by quasi-exogenous variation in industries' investor-WSJ connections. Overall, our study shows that crowd-sourcing institutional investor participation in news production improves the quality of the information environment in the long term.
with AJ Chen and Gerard Hoberg
☆ Selected Presentations: NBER Asset Pricing Meeting, Financial Research Association Meeting, UT Dallas Fall Finance Conference
Abstract
We postulate that our historical record has become adequately long and informative that newly arriving economic states often resemble historical states. Building on this insight, we develop a framework to predict future economic outcomes using the average of the realized outcomes that follow highly similar historical states. Using 210 million newspaper articles from 1815 to 2021, we identify historically similar months for each focal month and construct a predictor of aggregate U.S. stock returns, “SeenItRet”. SeenItRet strongly forecasts future market-wide stock returns up to two years ahead, with an annualized impact of 4–7% for a one standard deviation shift. Our framework is general and also predicts real economic outcomes, including recessions, inflation, and patenting activity. A virtue of our approach is its use of economic principles to reduce the high dimensionality of the underlying state space to an ex-ante measurable and intuitive unidimensional predictor. Our model performs better when historical states are more similar to the focal state, and it offers interpretable economic insights by highlighting the specific themes that drive its predictions.
with Michael Blank and Gregor Schubert [coming soon]
☆ Selected Presentations: OpenAI
Abstract
TBD
with Danny Qin [coming soon]
Abstract
TBD
Publications
with Matilde Bombardini and Francesco Trebbi
Annual Review of Economics, 2025, 17
Abstract
This article discusses recent methodological innovations in the area of cost and benefit assessment of government regulation, in both a prospective and retrospective sense. Much of the extant progress is presented on the front of private costs of compliance. Private benefits, social costs, and social benefits remain much less systematically organized and more arduous to quantitatively assess, mostly due to the difficulty of standardizing partial and general equilibrium counterfactuals. We offer a discussion of potential future methodological improvements in cost- benefit analysis.
with Nir Jaimovich and Nicolas Vincent
Economics Letters, 2024, 234(111437)
Abstract
Using establishments’ occupational data, we quantify the role of entrants, exiters, and incumbents in driving the decline in the share of routine occupations (R-share) in the U.S. First, entrants have a higher R-share than incumbents, casting doubt on a “creative destruction” mechanism whereby entrants drive this decline. Second, exiters have a higher R-share than their peers, supporting a “positive selection” mechanism. Finally, as incumbents age, they experience a fall in their R-share, which is not due to their size, consistent with the “technology adoption” mechanism. Quantitatively, we show that incumbents are the primary drivers of the aggregate decline in R-share.m. Quantitatively, we show that incumbents are the primary drivers of the aggregate decline in R-share.
with Mete Kilic and Louis Yang
Journal of Financial Economics, 2022, 145(3), 706-724
Internet Appendix
Abstract
Asset pricing predictions from the investment CAPM depend on the cross-sectional relation between investment and profitability. In samples of U.S. stocks featuring high cross-sectional investment-profitability correlation, both investment and profitability premiums are weak. Consistent with the conditional predictions from the investment CAPM, triple sorts on size, investment, and profitability as in Hou et al. (2015)’s q-factors resurrect the premiums in the high-correlation samples. We find similar results using cash-based profitability, consistent with the dynamic investment CAPM. Our work has important implications for constructing asset pricing factors and interpreting out-of-sample asset pricing test results, in particular the insignificance of historical investment and profitability premiums.
with Selale Tuzel
Journal of Finance, 2021, 76(6), 3347-3399
Presentation Slides | Internet Appendix | USC Tommy Talks (animation) | Data: Statement Section 179 Limits
Abstract
Do investment tax incentives improve job prospects for workers? We explore states’ adoption of a major federal tax incentive that accelerates the depreciation of equipment investments for eligible firms but not for ineligible ones. Analyzing massive establishment-level data sets on occupational employment and computer investment, we find that when states expand investment incentives, eligible firms immediately increase their equipment and skilled employees; whereas they reduce routine-task employees after a delay of up to two years. These opposing effects constitute an overall insignificant effect on the firms’ total employment and shed light on the nuances of job creation through investment incentives.
Journal of Finance, 2019, 74(4), 1793-1839
☆ WFA Cubist Systematic Strategies PhD Candidate Award for Outstanding Research 2016
☆ Best Paper Award at the USC Marshall Ph.D. Conference in Finance 2014
Presentation Slides | Internet Appendix | Data: Routine-Share for Industries 1990-2018 | Replication
Abstract
This paper studies the asset pricing implications of a firm’s opportunities to replace routine-task labor with automation. I develop a model in which firms optimally undertake such replacement when their productivity is low. Hence, firms with routine-task labor maintain a replacement option that hedges their value against unfavorable macroeconomic shocks and lowers their expected returns. Using establishment-level occupational data, I construct a measure of firms’ share of routine-task labor. Compared to their industry peers, firms with a higher share of routine-task labor (i) invest more in machines and reduce more routine-task labor during economic downturns, and (ii) have lower expected stock returns.
with Nir Jaimovich, Sergio Rebelo, and Arlene Wong
NBER Macroeconomics Annual, 2019, 34, 285-316
Discussion by Daron Acemoglu | Discussion by Jonathan Vogel | Summary of NBER Conference Discussion
Abstract
We first show that within industries, firms selling higher-quality goods rely less on routine-task workers and more on high-skilled abstract-task workers; and that higher-income consumers disproportionately purchase these goods. Based on these facts, we propose a new theory in which rising incomes increase demand for high-quality goods and high-skilled labor, thereby raising the U.S. skill premium.
with Selale Tuzel
Journal of Finance, 2017, 72(1), 325-370
☆ Best Paper Award at the SFM Annual Conference 2013
Presentation Slides | Internet Appendix | Data: Local Beta for MSAs 1986-2016
Abstract
Firm location affects firm risk through local factor prices. We find more procyclical factor prices such as wages and real estate prices in areas with more cyclical economies, namely, high “local beta” areas. While procyclical wages provide a natural hedge against aggregate shocks and reduce firm risk, procyclical prices of real estate, which are part of firm assets, increase firm risk. We confirm that firms located in higher local beta areas have lower industry-adjusted returns and conditional betas, and show that the effect is stronger among firms with low real estate holdings. A production-based equilibrium model explains these empirical findings.
with Eric Chang and Dragon Tang
Journal of Financial and Quantitative Analysis, 2015, 50(3), 597-622
Internet Appendix
Abstract
The suitability of complex financial products for household investors is an important issue in light of consumer financial protection. The U.S. Dodd–Frank Act, for instance, mandates that distributors check suitability when selling structured products to retail investors. However, little empirical evidence exists on such transactions. Using data from Hong Kong, we find that investors purchase 8% more structured products, on average, when the suitability is not checked. The effect of suitability checks is more pronounced for less financially literate investors. Moreover, investors tend to buy products with lower risk-adjusted returns when product suitability is not checked.
Selected Work in Progress
with Rawley Heimer and Allen Hu
Other Publications and Reports
Journal of Monetary Economics, Forthcoming
Abstract
This article summarizes my discussion at the Carnegie-Rochester-NYU Conference 2025 Conference on The Consequences of AI use on Society and Policy about an article titled “Is It AI or Data That Drives Market Power.”
with Francesco Trebbi
☆ Media coverage: ProMarket (Chicago Booth)
Abstract
This brief research note employs the quantitative approach developed by Trebbi, Zhang, and Simkovic (2025) to provide a descriptive overview of the main differences in costs of regulatory compliance across U.S. states for the year 2014 and over the period 2002-2014. These descriptive stylized facts can be useful in grounding extant discussion on regulatory compliance burden across different U.S. regions and over time and presents both unconditional results and results controlling for state industry composition.