Working Papers

Consumption in Asset Returns (June 2024), with Svetlana Bryzgalova and Christian Julliard

Abstract: Using information in returns we identify the stochastic process of consumption. We find that aggregate consumption reacts over multiple quarters to innovations spanned by financial markets, and this persistent component accounts for over a quarter of consumption variation. These shocks are cross-sectionally priced, drive most of the time series variation in stocks, and a small, yet significant, share of volatility of bonds. Nevertheless, we find no support for stochastic volatility of consumption driving timevarying risk premia. Finally, an otherwise standard recursive utility model based on our estimated process explains both equity premium and risk-free rate puzzles with low risk aversion.

Macro Strikes Back: Term Structure of Risk Premia and Market Segmentation (Mar 2024), with Svetlana Bryzgalova and Christian Julliard

Abstract: We develop a unified framework to study the term structure of risk premia of nontradable factors. Our method delivers level and time variation of risk premia, uncovers their propagation mechanism, is robust to misspecification and weak identification, and allows for segmented markets. Most macroeconomic factors are weakly identified at quarterly frequency, but have increasing (unconditional) term structures with large risk premia at business cycle horizons. Moreover, the slopes of their term structures are strongly procyclical. Most macroeconomic and intermediary-based factors command similar risk premia in equity and corporate bond markets, while we find strong evidence of segmentation for other factors.

Model Uncertainty in the Cross Section (Dec 2023), with Ran Shi

Abstract: We develop a transparent Bayesian framework to measure uncertainty in asset pricing models. Our framework quantifies the tradeoff between mean-variance efficiency and parsimony for models to attain high posterior probabilities. Model uncertainty is defined as the entropy of these posterior probabilities, which is consistently interpretable even under misspecification due to omitted factors. Empirically, model uncertainty accumulates during major market events, carrying a significantly negative risk premium of approximately half the magnitude of the market. Positive shocks to model uncertainty predict persistent outflows from US equity funds and inflows to Treasury funds.

Frequency-Dependent Risks in the Factor Zoo (Feb 2023)

Internet Appendix

Abstract: I propose a novel framework to quantify frequency-dependent risks in the factor zoo. Empirically, the stochastic discount factor (SDF) composed of several low-frequency principal components captures all the risk premia in asset returns. It also explains well the cross-section of characteristic-sorted portfolios. Moreover, I decompose the low-frequency SDF into two orthogonal priced components. The first component is composed of high-frequency or traditional principal components. It is serially uncorrelated and relates to discount-rate news, intermediary factors, jump risk, and investor sentiment. The second component is persistent and captures business-cycle risks related to consumption and GDP growth.

Publications

Bayesian Solutions for the Factor Zoo: We Just Ran Two Quadrillion Models, with Svetlana Bryzgalova and Christian Julliard, Journal of Finance (2023), vol. 78(1), 487-557. 

Full replication codes (including posterior draws and usage examples, 4.56GB) 

BayesianFactorZoo R package on CRAN

Abstract: We propose a novel framework for analyzing linear asset pricing models: simple, robust, and applicable to high dimensional problems. For a (potentially misspecified) standalone model, it provides reliable price of risk estimates for both tradable and non-tradable factors, and detects those weakly identified. For competing factors and (possibly non-nested) models, the method automatically selects the best specification – if a dominant one exists – or provides a Bayesian model averaging (BMA-SDF), if there is no clear winner. We analyze 2.25 quadrillion models generated by a large set of factors, and find that the BMA-SDF outperforms existing models in- and out-of-sample.