Accepted Manuscripts
Moderate Utility. with Paulo Natenzon
American Economic Review: Insights. Vol. 6, Jun. 2024.
Choice hinges on utilities and the ease of comparison. Moderate transitivity characterizes such decision-making processes.
Bayesian Contextual Choices under Imperfect Perception of Attributes.
Management Science. Vol. 70, Mar. 2023.
Contextual choices are inconsistent with the usual utility-maximization modeling. However, by considering a simple attribute perception error, the standard Bayesian choice model predicts many contextual patterns seen in lab experiments.
The Experimental Evaluation of Brand Strength and Brand Value. with Bobby J. Calder
Journal of Business Research. Vol. 115, Jul. 2020.
A simple method to assess brand strength with experiments and its financial value as an asset.
Risk Reducers in Convex Order. with Qihe Tang and Huan Zhang
Insurance: Mathematics and Economics. Vol. 70, Sep. 2016
The convex hull of random variables equal in distribution is the set of reduced risks. We study its applications to multivariate stochastic ordering, index-linked hedging strategies, and optimal reinsurance.
On the Necessity of Pairs and Triplets for the Equivalence between Rationality Axioms.
Economics E-Journal. Aug. 2012
Can we maintain the equivalence of classic rationality axioms without including all finite budget sets in the domain of a choice function? Yes, if the domain is closed under certain finite unions.
Moderate Utility. with Paulo Natenzon
American Economic Review: Insights. Vol. 6, Jun. 2024.
Choice hinges on utilities and the ease of comparison. Moderate transitivity characterizes such decision-making processes.
Bayesian Contextual Choices under Imperfect Perception of Attributes.
Management Science. Vol. 70, Mar. 2023.
Contextual choices are inconsistent with the usual utility-maximization modeling. However, by considering a simple attribute perception error, the standard Bayesian choice model predicts many contextual patterns seen in lab experiments.
The Experimental Evaluation of Brand Strength and Brand Value. with Bobby J. Calder
Journal of Business Research. Vol. 115, Jul. 2020.
A simple method to assess brand strength with experiments and its financial value as an asset.
Risk Reducers in Convex Order. with Qihe Tang and Huan Zhang
Insurance: Mathematics and Economics. Vol. 70, Sep. 2016
The convex hull of random variables equal in distribution is the set of reduced risks. We study its applications to multivariate stochastic ordering, index-linked hedging strategies, and optimal reinsurance.
On the Necessity of Pairs and Triplets for the Equivalence between Rationality Axioms.
Economics E-Journal. Aug. 2012
Can we maintain the equivalence of classic rationality axioms without including all finite budget sets in the domain of a choice function? Yes, if the domain is closed under certain finite unions.
Working Papers
No Sparsity in Asset Pricing: Evidence from a Generic Statistical Test
with Lingxiao Zhao and Guofu Zhou
R&R @ Review of Financial Studies.
We present a novel test to determine sparsity in characteristic-based factor models. Applying the test to industry and pseudo-random portfolios, we reject the null hypothesis that fewer than ten factors are sufficient to explain returns, and show that at least forty factors are needed for the various sample periods examined. We find that dense models outperform sparse ones in both pricing and investing. Testing with tree-based portfolios also indicates no sparsity. Our results suggest that most existing factor models, which have fewer than six factors, are questionable, and that future research on such low-dimensional models is unlikely to be fruitful.
Learning Source Biases: Multisource Misspecifications and Their Impact on Predictions
with Lin Hu, Matthew Kovach, and Anqi Li
R&R @ American Economic Review.
We study how a Bayesian decision maker (DM) learns about the biases of novel information sources to predict a random state. Absent frictions, the DM uses familiar sources as yardsticks to accurately discern the biases of novel sources. We derive the distortion of the DM's long-run prediction when he holds misspecified beliefs about the biases of several familiar sources. The distortion aggregates misspecifications across familiar sources independently of the number and nature of the novel sources the DM learns about. This has implications for labor market discrimination, media bias, and project finance and oversight.
Ridge Estimation of High Dimensional Two-Way Fixed Effect Regression
with Jean-Marc Robin
We study a ridge estimator for the high-dimensional two-way fixed effect regression model. We show in simulations that it performs much better than OLS, and possibly also, when the network is very sparse, than the usual bias corrections for second order moments. We then develop an asymptotic theory that helps understand why it is so. We develop concentration inequalities showing that when the ridge parameters increase as the log of the network size, the bias and the variance-covariance matrix of the vector of estimated fixed effects converge to deterministic equivalents that depend only on the expected network.
Diversified Production and Market Power: Theory and Evidence from Renewables (HBS wp)
with Michele Fioretti and Jorge Tamayo
We study the Colombian energy market, where diversified energy firms strategically substitute thermal generation for hydropower before droughts. This withinfirm substitution, due to thermal generators internalizing the drop in hydropower supply during droughts, mitigates higher market prices. We show theoretically and empirically that these virtuous spillovers exist when thermal generators have market power but are severed when their residual demands are vertical or horizontal, which attenuates a firm’s business stealing incentives. We conclude that industry consolidation can reduce prices if it promotes diversified production portfolios. Diversification can keep the green transition affordable by reducing the cost of renewable intermittencies.
Accelerating Equity: Overcoming the Gender Gap in VC Funding (link to draft.pdf)
with Chuan Chen, Michele Fioretti, and Yanrong Jia
We examine the growing gender gap in venture capital funding, focusing on accelerator programs in the U.S. We collect a unique dataset with detailed information on accelerators and startups. Using a two-stage methodology, we first estimate a matching model between startups and accelerators, and then use its output to analyze the gender gap in post-graduation outcomes through a control function approach. Our results show that female-founded startups face a significant funding disadvantage, primarily due to relocation challenges tied to family obligations. However, larger cohorts and higher-quality accelerators help reduce this gap by offering female founders better networking opportunities and mentorship.
Random Choice and Differentiation
with Paulo Natenzon
Differentiation determines the comparability of different options and can be crucial to predict how choice architecture elicits behavioral responses. To facilitate the measurement of differentiation, we develop a flexible yet tractable model of random choice in a multi-attribute setting. We show the analyst can separately identify vertical and horizontal differentiation from binary comparison data alone. We characterize the binary choice rules that arise from our model using four easily understood postulates. In multinomial choice, we show that the intersection of our model with the classic random utility framework yields random coefficients with an elliptical distribution. We provide applications to consumer demand with differentiated products and to measuring the complexity faced by an agent in individual decision-making problems.
Teaching Them How To Fish? The Makings of Business Accelerators
with Chuan Chen
One-third of early-stage venture financing currently goes to business accelerator-backed startups. We examine whether startups join accelerators to alleviate financing constraints or to improve human capital for long-term growth. Using a novel three-stage two-sided matching econometric framework, we quantify and compare explanatory powers of short-term financing prospects and long-term growth prospects for the revealed preference during the accelerator admission process. We find that short-term financing prospects explain 48% of the variations in the matching values, while variations of key five-year startup performance prospects can explain 96%.
Prior Free Bayesian Estimation through the AIC
with Werner Ploberger
In this paper we show that, in linear models with an increasing number of parameters, the estimator resulting from the maximization of Akaike's Information Criterion is asymptotically equivalent to some Bayesian estimators. The family of prior distributions which generates our estimators consists of normal distributions, defined on the space of all sequence, and is characterized by an exponential decay of the variance for the higher order components of the parameter.
Optimal Estimation when the Parameter Space is of Infinite Dimension
with Werner Ploberger.
Many classical non-parametric estimation problems can be reduced to the estimation of an infinite dimensional parameter vector. When maximum likelihood estimators do not exist, we construct simple prior distributions for the parameters that force the parameters to obey some higher order smoothness conditions. We show that under such prior distributions, certain shrunken sieve estimators are asymptotically optimal for a family of loss functions.
No Sparsity in Asset Pricing: Evidence from a Generic Statistical Test
with Lingxiao Zhao and Guofu Zhou
R&R @ Review of Financial Studies.
We present a novel test to determine sparsity in characteristic-based factor models. Applying the test to industry and pseudo-random portfolios, we reject the null hypothesis that fewer than ten factors are sufficient to explain returns, and show that at least forty factors are needed for the various sample periods examined. We find that dense models outperform sparse ones in both pricing and investing. Testing with tree-based portfolios also indicates no sparsity. Our results suggest that most existing factor models, which have fewer than six factors, are questionable, and that future research on such low-dimensional models is unlikely to be fruitful.
Learning Source Biases: Multisource Misspecifications and Their Impact on Predictions
with Lin Hu, Matthew Kovach, and Anqi Li
R&R @ American Economic Review.
We study how a Bayesian decision maker (DM) learns about the biases of novel information sources to predict a random state. Absent frictions, the DM uses familiar sources as yardsticks to accurately discern the biases of novel sources. We derive the distortion of the DM's long-run prediction when he holds misspecified beliefs about the biases of several familiar sources. The distortion aggregates misspecifications across familiar sources independently of the number and nature of the novel sources the DM learns about. This has implications for labor market discrimination, media bias, and project finance and oversight.
Ridge Estimation of High Dimensional Two-Way Fixed Effect Regression
with Jean-Marc Robin
We study a ridge estimator for the high-dimensional two-way fixed effect regression model. We show in simulations that it performs much better than OLS, and possibly also, when the network is very sparse, than the usual bias corrections for second order moments. We then develop an asymptotic theory that helps understand why it is so. We develop concentration inequalities showing that when the ridge parameters increase as the log of the network size, the bias and the variance-covariance matrix of the vector of estimated fixed effects converge to deterministic equivalents that depend only on the expected network.
Diversified Production and Market Power: Theory and Evidence from Renewables (HBS wp)
with Michele Fioretti and Jorge Tamayo
We study the Colombian energy market, where diversified energy firms strategically substitute thermal generation for hydropower before droughts. This withinfirm substitution, due to thermal generators internalizing the drop in hydropower supply during droughts, mitigates higher market prices. We show theoretically and empirically that these virtuous spillovers exist when thermal generators have market power but are severed when their residual demands are vertical or horizontal, which attenuates a firm’s business stealing incentives. We conclude that industry consolidation can reduce prices if it promotes diversified production portfolios. Diversification can keep the green transition affordable by reducing the cost of renewable intermittencies.
Accelerating Equity: Overcoming the Gender Gap in VC Funding (link to draft.pdf)
with Chuan Chen, Michele Fioretti, and Yanrong Jia
We examine the growing gender gap in venture capital funding, focusing on accelerator programs in the U.S. We collect a unique dataset with detailed information on accelerators and startups. Using a two-stage methodology, we first estimate a matching model between startups and accelerators, and then use its output to analyze the gender gap in post-graduation outcomes through a control function approach. Our results show that female-founded startups face a significant funding disadvantage, primarily due to relocation challenges tied to family obligations. However, larger cohorts and higher-quality accelerators help reduce this gap by offering female founders better networking opportunities and mentorship.
Random Choice and Differentiation
with Paulo Natenzon
Differentiation determines the comparability of different options and can be crucial to predict how choice architecture elicits behavioral responses. To facilitate the measurement of differentiation, we develop a flexible yet tractable model of random choice in a multi-attribute setting. We show the analyst can separately identify vertical and horizontal differentiation from binary comparison data alone. We characterize the binary choice rules that arise from our model using four easily understood postulates. In multinomial choice, we show that the intersection of our model with the classic random utility framework yields random coefficients with an elliptical distribution. We provide applications to consumer demand with differentiated products and to measuring the complexity faced by an agent in individual decision-making problems.
Teaching Them How To Fish? The Makings of Business Accelerators
with Chuan Chen
One-third of early-stage venture financing currently goes to business accelerator-backed startups. We examine whether startups join accelerators to alleviate financing constraints or to improve human capital for long-term growth. Using a novel three-stage two-sided matching econometric framework, we quantify and compare explanatory powers of short-term financing prospects and long-term growth prospects for the revealed preference during the accelerator admission process. We find that short-term financing prospects explain 48% of the variations in the matching values, while variations of key five-year startup performance prospects can explain 96%.
Prior Free Bayesian Estimation through the AIC
with Werner Ploberger
In this paper we show that, in linear models with an increasing number of parameters, the estimator resulting from the maximization of Akaike's Information Criterion is asymptotically equivalent to some Bayesian estimators. The family of prior distributions which generates our estimators consists of normal distributions, defined on the space of all sequence, and is characterized by an exponential decay of the variance for the higher order components of the parameter.
Optimal Estimation when the Parameter Space is of Infinite Dimension
with Werner Ploberger.
Many classical non-parametric estimation problems can be reduced to the estimation of an infinite dimensional parameter vector. When maximum likelihood estimators do not exist, we construct simple prior distributions for the parameters that force the parameters to obey some higher order smoothness conditions. We show that under such prior distributions, certain shrunken sieve estimators are asymptotically optimal for a family of loss functions.
Work in Progress
A Note on Hypothesis Testing
On the Loss of Efficiency from Inequality
with Inkee Jang.
A Note on Hypothesis Testing
On the Loss of Efficiency from Inequality
with Inkee Jang.
