JOB MARKET PAPER
Job market seminars at the business schools of
Boston College, Boston University, Copenhagen, Georgia State University (canceled), HEC Paris, HKUST, Indian School of Business, National University of Singapore, Tulane, Universities of Florida, Georgia, and Houston, Yale University
This paper derives ex-ante standard errors of risk premium predictions from neural networks (NNs). Considering standard errors, I provide improved investment strategies and ex-post out-of-sample (OOS) statistical inferences relative to existing literature. The equal-weighted (value-weighted) confident high-low strategy that takes long-short positions exclusively on stocks that have precise risk premia earns an OOS average monthly return of 3.61% (2.21%). In contrast, the conventional high-low portfolio yields 2.52% (1.48%). Existing OOS inferences do not account for ex-ante estimation uncertainty and thus are not adequate to statistically compare the OOS returns, Sharpe ratios and mean squared errors of competing trading strategies and return prediction models (e.g., linear, NN and random forest). I develop a bootstrap procedure that delivers robust OOS inferences. The bootstrap tests reveal that large OOS return and Sharpe ratio differences between NN and benchmark linear models' traditional high-low portfolios are statistically insignificant. However, the NN-based confident high-low portfolios significantly outperform all competing strategies. Economically, standard errors reflect time-varying market uncertainty and spike after financial shocks. In the cross-section, the level and precision of risk premia are correlated, thus NN-based investments deliver more gains in the long positions.
Won Cubist Systematic Strategies Award for Outstanding Research at the WFA 2020.
Western Finance Association (WFA) Annual Meetings, 2020
European Finance Association Annual Meetings, Lisbon, Portugal, 2019
The University of Chicago, Machine Learning and New Empirical Asset Pricing, 2018
Northern Finance Association Annual Meetings, Charlevoix, Canada, 2018
SoFiE Annual Conference, Lugano, Switzerland, 2018,
European Econometric Society Annual Meetings, Cologne, Germany, 2018
This paper develops a Bayesian methodology to compare asset pricing models containing non-traded factors and principal components. Existing comparison procedures are inadequate when models include such factors due to estimation uncertainties in mimicking portfolios and return covariances. Furthermore, regressions of test assets on such factors are interdependent, rendering comparisons with recently proposed priors sensitive to subsets of the test assets. Thus, I derive novel, non-informative priors that deliver invariant inferences. Simulations suggest that my methodology outperforms existing methods in identifying true non-traded models. I find that macroeconomic factor models dominate several, recent benchmark models with traded factors and principal components.
The paper includes the following notes on priors for comparing asset pricing models.
This article provides an extensive discussion on what priors to use, when, and why, to compare asset pricing models in general. I layout rules to derive priors under three different cases. The first case relates to Barillas and Shanken(2020), and Chib, Zeng, and Zhao(2020), which involves comparing multiple models that exclusively comprise traded factors. In the second, I discuss rules to test an individual model containing non-traded factors. The third considers comparing multiple models involving non-traded factors.
(joint with Tarun Chordia)
Invited for dual submission, Review of Financial Studies, 2021
Scheduled (but canceled) to present at the ABFER Annual Meetings, 2020
We develop a big-data methodology to estimate fundamental prices and true liquidity measures, explicitly considering the rounding specification due to the minimum tick size. Evaluation of the tick size pilot (TSP), which increased the tick size for some randomly chosen stocks, requires estimating the impact of rounding. True liquidity measures capture the TSP-driven decreased inventory costs of market-makers, whereas traditional measures without the rounding adjustment cannot. We find that the TSP increases market-maker profits, but does not improve liquidity and price efficiency. This result contrasts with existing empirical studies but is consistent with recent theoretical studies that account for rounding.