Abstract

In an attempt to explain stock prices in a simple and articulate manner, researchers in asset pricing have long employed a “one model fits all prices over time” approach. In this thesis, a regime-switching linear asset pricing model, conditioned on trailing market volatility, is introduced in an attempt to map specific linear model specifications to specific macroeconomicconditions which are quantifiable and observable at t. Called the Time-Robust Iterative Adaptive LASSO (TRIAL) model, this model computes the expected return more efficiently out-of-sample than a benchmark model which does not use volatility-defined regimes, yielding a risk-adjusted mean monthly return of 1.36%, compared to the “no regime” model return of 1.12%. Compared to a seminal model in asset pricing, the Fama-French 3-Factor Model (“FF3”), the TRIAL model evinces a higher monthly return than the FF3’s return of 1.23%, while also providing lower monthly volatility than the FF3, at 6.31% and 6.97%, respectively. The TRIAL model fits its distinct regimes well, having statistically significant, unique linear specifications for each regime. The TRIAL model is asymmetric to the popular size and value factors across the regimes: it estimates big firms better than small firms and value firms better than growth firms. When four single-sort out-of-sample portfolios are created on size and value, the “big firm” portfolio yields a risk-adjusted mean monthly return of 2.38% and a Sharpe ratio of 1.27 over the 17-year testing period, well above any other benchmark model in the study.

Committee Chair

Guofu Zhou

Committee Members

Brett Green, Andreas Neuhierl,

Degree

Doctor of Business Administration (DBA)

Author's Department

Finance

Author's School

Olin Business School

Document Type

Dissertation

Date of Award

Spring 5-15-2023

Language

English (en)

Download

Share

COinS