Research

Motivating Innovation

The optimal incentive scheme for motivating innovation tolerates, or even rewards, early failure and rewards long-term success. Commitment to a long-term compensation plan, job security and timely feedback are also essential. In managerial pay, the scheme can be implemented with stock options with long vesting periods, option repricing, golden parachutes and managerial entrenchment.

Journal of Finance 66 (2011), 1823–1860.

Is Pay-for-Performance Detrimental to Innovation?

In a laboratory experiment, subjects paid under tolerance for early failure combined with reward for long-term success explored more and were more likely to discover a novel business strategy than subjects under fixed wages or standard pay-for-performance. The threat of termination reduced exploration, and golden parachutes offset that effect.

with Florian Ederer. Management Science 59 (2013), 1496–1513.

Incentives and Creativity: Evidence from the Academic Life Sciences

Investigators of the Howard Hughes Medical Institute, whose funding tolerates early failure, rewards long-term success and gives freedom to experiment, produce high-impact papers at a much higher rate than similarly accomplished NIH-funded scientists, and their research moves in more novel directions.

with Pierre Azoulay and Joshua S. Graff Zivin. RAND Journal of Economics 42 (2011), 527–554.

Incentives to Innovate and the Decision to Go Public or Private

It is optimal to go public to exploit existing ideas and to go private to explore new ones. Private firms are less transparent, so insiders can exit early if they receive bad news, which makes them more tolerant of failure; public prices react quickly to good news, which pushes insiders toward conventional projects.

with Daniel Ferreira and André C. Silva. Review of Financial Studies 27 (2014), 256–300.

Independent Boards and Innovation

Firms that move to independent boards patent more and are cited more, but in crowded, familiar areas of technology and through incremental patents in the middle of the citation distribution; the uncited and highly cited tails associated with riskier search do not change.

with Benjamin Balsmeier and Lee Fleming. Journal of Financial Economics 123 (2017), 536–557.

Shareholder Litigation and Corporate Innovation

After states adopted universal demand laws, which make it harder for shareholders to file derivative lawsuits, firms invested more in R&D and produced more patents in new technology classes, more patents built on new knowledge, and more highly cited and more valuable patents. The pressure of shareholder litigation discourages exploratory innovation.

with Chen Lin and Sibo Liu. Management Science 67 (2021), 3346–3367.

Do Technology Clusters Distort Innovation?

Inventors in larger technology clusters patent more, but the gain comes entirely from technology classes they already know: patents in new-to-inventor classes, patent creativity, forward citations and market value all fall as clusters grow. A simple explore–exploit model explains why: agglomeration makes exploitation cheaper and tilts private incentives away from exploration, opening a wedge between private and social returns.

with Benjamin Balsmeier and Lee Fleming. Working paper, 2026.

The Streetlight Effect in Data-Driven Exploration

Data on past attempts can hinder breakthrough discovery. When early data highlights an attractive but ultimately suboptimal project, agents crowd around it and free-ride on the data others generate instead of exploring. A laboratory experiment confirms the effect, and in research on the genetic origins of human diseases, diseases with early evidence of a promising genetic target are 16 percentage points less likely to yield breakthroughs than diseases where early efforts failed.

with Johannes Hoelzemann, Abhishek Nagaraj and Matteo Tranchero. Working paper, 2026.

Heterogeneous Innovation over the Business Cycle

Standard measures of innovation such as R&D spending and patent counts rise in booms, which seems to contradict Schumpeter's idea of recessions as periods of creative destruction. Measures that distinguish exploratory from exploitative patents show that firms shift toward exploration in contractions and toward exploitation in expansions, especially in cyclical industries.

with Benjamin Balsmeier and Lee Fleming. Review of Economics and Statistics 105 (2023), 1224–1236.

The Impact of Connectivity on the Production and Diffusion of Knowledge

In a social bandit model, greater connectivity speeds the diffusion of knowledge but can reduce its production, because agents who can watch others' experiments free-ride instead of experimenting themselves. Under some conditions a more connected society becomes more homogeneous and worse off.

with Farzad Pourbabaee. Working paper, 2022.

Innovation Search Strategy and Predictable Returns

Firms whose patenting leans toward exploitation of familiar technologies go on to deliver better short-term operating results than analysts expect, and their stocks are undervalued relative to exploration-focused firms even after standard risk and innovation factors. Investors appear to pay more attention to unfamiliar exploratory patents than to incremental ones.

with Tristan Fitzgerald, Benjamin Balsmeier and Lee Fleming. Management Science 67 (2021), 1109–1137.

Creating Incentives for Innovation

A summary for managers of the research on motivating innovation: organizations that want experimentation should tolerate early failure, measure performance over long horizons, provide job security and give regular, timely feedback, from the CEO down.

California Management Review 60 (2017), 18–32.

National Institutes of Health Peer Review: Challenges and Avenues for Reform

The NIH's peer-review system, built on investigator initiation and rigorous review, has long been a model for science funders, but tight budgets and a changed research ecosystem raise the question of whether post-war practices still fit. The essay examines the aging of the NIH-funded scientist population and the innovativeness of funded research, and proposes reforms along with a way to implement and evaluate them.

with Pierre Azoulay and Joshua S. Graff Zivin. Innovation Policy and the Economy 13 (2013), 1–21, University of Chicago Press.

Incentives for Innovation: Bankruptcy, Corporate Governance, and Compensation Systems

A survey of the theory and evidence on how laws, institutions and compensation systems shape the intensity and direction of innovation, focused on optimal contracting and on applications to bankruptcy law, corporate governance and compensation.

with Florian Ederer. In Handbook of Law, Innovation, and Growth, Edward Elgar Publishing, 2011.

Selecting for Innovation

Work in progress.

The Salience of Entrepreneurship: Evidence from Online Business

Using data from Taobao, the world's largest online marketplace, the paper finds that people who see a nearby store succeed are more likely to open one themselves; entrants motivated this way sell less and exit more often, a pattern consistent with salience rather than rational learning.

with Yi Huang, Chen Lin and Sibo Liu. Management Science, forthcoming.

Experimentation and the Returns to Entrepreneurship

Studies concluding that entrepreneurs earn less and bear more risk than comparable employees rely on cross-sectional data and ignore the option value of experimenting with an idea. In longitudinal data, entrepreneurial spells are short, people quit after bad results and are not penalized on returning to salaried work, and lifetime returns to entrepreneurship are more attractive than previously found.

Review of Financial Studies 29 (2016), 2319–2340.

Recommendations with Feedback

When a recommender's choices affect which product ends up being best, recommendations become self-fulfilling: the recommender can make almost any firm the winner. This can leave too little incentive to gather information, a bias toward the status quo, and avoidance of risky innovations; payments from firms can help, while competition among recommenders and payments from consumers do not.

with Ganesh Iyer. Review of Financial Studies 36 (2023), 501–533.

The Impact of Connectivity on the Production and Diffusion of Knowledge

In a social bandit model, greater connectivity speeds the diffusion of knowledge but can reduce its production, because agents who can watch others' experiments free-ride instead of experimenting themselves. Under some conditions a more connected society becomes more homogeneous and worse off.

with Farzad Pourbabaee. Working paper, 2022.

Information Percolation in Segmented Markets

In dynamic over-the-counter markets where traders differ in information quality and in how often they meet counterparties, learning externalities shape the incentive to gather information. More liquid markets raise equilibrium information acquisition when gains from trade and market duration are large; in short-lived markets the opposite can happen when traders differ enough in connectivity.

with Darrell Duffie and Semyon Malamud. Journal of Economic Theory 153 (2014), 1–32.

Libertarian Paternalism, Information Production, and Financial Decision Making

When a social planner makes recommendations (libertarian paternalism), people read information into them and gather less on their own, which slows the spread of information through social learning and advice. The policy improves welfare in some settings and is suboptimal in others because of this effect on information production.

with Bruce Ian Carlin and Simon Gervais. Review of Financial Studies 26 (2013), 2204–2228.

Feedback Effects of Credit Ratings

Because a rating changes the borrower's cost of capital and thus its chance of survival, a rating agency should weigh those feedback effects, not only accuracy. Even an accurate rating policy can produce multi-notch downgrades or immediate default after small shocks, and more competition among agencies can lead to downgrades, more defaults and lower welfare.

Journal of Financial Economics 109 (2013), 535–548.

Obfuscation, Learning, and the Evolution of Investor Sophistication

In a dynamic model of retail financial markets, providers choose when to obfuscate their products as investors learn. Educational initiatives meant to help investors can induce providers to obfuscate more, leaving investors more confused and lowering welfare; obfuscation falls as competition among providers increases.

with Bruce Ian Carlin. Review of Financial Studies 24 (2011), 754–785.

The Relative Contributions of Private Information Sharing and Public Information Releases to Information Aggregation

When agents learn both from public releases and from privately meeting other agents, beliefs converge to the perfect-information limit exponentially at a rate equal to the sum of the public arrival rate and the private matching rate. Without private sharing, convergence is strictly slower than the public arrival rate alone.

with Darrell Duffie and Semyon Malamud. Journal of Economic Theory 145 (2010), 1574–1601.

Information Percolation

As agents in a large market meet in groups and reveal what they know, the cross-sectional distribution of beliefs has an explicit solution. Beliefs converge to a common posterior exponentially, at a rate equal to the mean rate at which an individual agent is matched, regardless of the size of the groups that meet.

with Darrell Duffie and Gaston Giroux. American Economic Journal: Microeconomics 2 (2010), 100–111.

Information Percolation with Equilibrium Search Dynamics

Agents choose how hard to search for others with whom to share information. In equilibrium they search maximally until their information reaches a target precision and then minimally. A tax that subsidizes search can improve information sharing and welfare, while giving agents public signals can crowd out private sharing and in some cases reduce welfare.

with Darrell Duffie and Semyon Malamud. Econometrica 77 (2009), 1513–1574.

Information Percolation in Large Markets

Introduces a simple model of the percolation of information through a large market as agents meet over time and reveal what they know, including what they learned from earlier meetings, and solves explicitly for the cross-sectional distribution of beliefs, with an example based on private auctions.

with Darrell Duffie. American Economic Review (Papers and Proceedings) 97 (2007), 203–209.

Biased Auctioneers

A neural network that predicts art auction prices from images and catalogue data reveals that auction houses' pre-sale estimates are informationally inefficient: when the algorithm's valuation exceeds the estimate, prices come in higher relative to the estimate and lots are less likely to go unsold, and auctioneers' errors are persistent enough to be predictable from their past errors.

with Mathieu Aubry, Roman Kräussl and Christophe Spaenjers. Journal of Finance 78 (2023), 795–833.

Macroeconomic Risk and Debt Overhang

Corporate debt is riskier in recessions, so the transfers from equity to debt holders that accompany investment decisions concentrate in bad times. In a calibrated dynamic capital structure model, this makes the cost of debt overhang higher when macroeconomic risk is present, and the cyclicality of a firm's assets in place and growth options shapes its investment and financing.

with Hui Chen. Review of Corporate Finance Studies 6 (2017), 1–38.

Feedback Effects of Credit Ratings

Because a rating changes the borrower's cost of capital and thus its chance of survival, a rating agency should weigh those feedback effects, not only accuracy. Even an accurate rating policy can produce multi-notch downgrades or immediate default after small shocks, and more competition among agencies can lead to downgrades, more defaults and lower welfare.

Journal of Financial Economics 109 (2013), 535–548.

Governance Through Trading and Intervention: A Theory of Multiple Blockholders

Splitting a block among several small blockholders creates free-rider problems that weaken direct intervention, but the same coordination failure strengthens governance through trading: because the blockholders cannot coordinate to limit their orders, they trade competitively and impound more information into prices, which sharpens the threat of disciplinary trading and raises managerial effort.

with Alex Edmans. Review of Financial Studies 24 (2011), 2395–2428.

Performance-Sensitive Debt

Debt whose interest rate depends on the borrower's performance leads to earlier default and lower equity value than fixed-rate debt of the same value, so the trade-off theory cannot explain it. It works instead as an inexpensive screening device, consistent with the pecking-order theory: firms that choose performance-sensitive loans are more likely to see their credit ratings improve. The paper also derives closed-form prices for step-up bonds and linear performance-sensitive debt.

with Bruno Strulovici and Alexei Tchistyi. Review of Financial Studies 23 (2010), 1819–1854.

Investment Reversibility and Agency Cost of Debt

The inefficiency that debt financing introduces into equityholders' investment decisions has an upper bound that falls with the degree of investment reversibility and reaches zero when investment is perfectly reversible.

Econometrica 76 (2008), 437–442.

Household Debt Overhang and Human Capital Investment

Human capital, unlike labor income, cannot be seized by creditors, so investment in skills is more resilient to debt overhang than hours worked. But because skills are only valuable to people who expect to keep working, the drop in labor supply caused by debt feeds back into lower investment in human capital.

with Alejandro Rivera, Hui (Grace) Wang and Han Xia. Journal of Financial Economics 172 (2025), article 104141.

Student Loans and Labor Supply Incentives

Because student loans cannot be discharged in bankruptcy, households cannot reduce their labor supply at creditors' expense, so student debt creates less debt overhang than other debt. Income-driven repayment plans, which set payments by formula regardless of the balance, act like a partial discharge and bring the overhang back.

with Alejandro Rivera, Hui (Grace) Wang and Han Xia. Working paper, 2025.

Libertarian Paternalism, Information Production, and Financial Decision Making

When a social planner makes recommendations (libertarian paternalism), people read information into them and gather less on their own, which slows the spread of information through social learning and advice. The policy improves welfare in some settings and is suboptimal in others because of this effect on information production.

with Bruce Ian Carlin and Simon Gervais. Review of Financial Studies 26 (2013), 2204–2228.