Systematic Risk, Liquidity Premia, and the Cross-Section of Expected Returns: A Unified Factor Framework
Evidence from U.S. Equity Markets
The Economic Institute
15 March 2025
Abstract
This paper develops and tests a unified multi-factor asset pricing framework that incorporates systematic market risk, size, value, liquidity, and volatility risk premia into a single coherent model for the cross-section of expected equity returns. Using comprehensive data on U.S. common stocks from the Center for Research in Security Prices (CRSP) spanning 1963 to 2023, merged with Compustat accounting data, intraday Trade and Quote (TAQ) microstructure data, and OptionMetrics implied volatility surfaces, we construct tradable factor-mimicking portfolios for each priced risk dimension and evaluate the model's ability to explain average returns on a broad array of test assets. Our liquidity factor, constructed from a composite measure that synthesizes the Amihud (2002) illiquidity ratio, the Pastor and Stambaugh (2003) traded liquidity innovation, and a bid-ask spread measure derived from TAQ data, earns an unconditional premium of 0.41 percent per month (t-statistic = 3.87) and exhibits low correlation with existing Fama and French (2015) factors. Our volatility risk premium factor, constructed as the return differential between portfolios sorted on the spread between option-implied and realized volatility, commands a premium of 0.34 percent per month (t-statistic = 3.21) and captures compensation for bearing aggregate uncertainty risk that is not subsumed by the market or value factors.
In formal Fama and MacBeth (1973) cross-sectional regressions on 25 size and book-to-market portfolios, 25 size and momentum portfolios, 10 industry portfolios, and 25 size and illiquidity portfolios, the unified five-factor model (Market, SMB, HML, LIQ, VOL) produces a cross-sectional R-squared of 78 percent and a mean absolute pricing error of 0.07 percent per month, compared with 54 percent and 0.14 percent for the CAPM, 68 percent and 0.11 percent for the Fama-French three-factor model, and 73 percent and 0.09 percent for the Fama-French five-factor model. The GRS test of Gibbons, Ross, and Shanken (1989) fails to reject the null that all pricing errors are jointly zero at conventional significance levels (p-value = 0.127) for our model, whereas it decisively rejects the CAPM (p < 0.001), the three-factor model (p = 0.003), and the five-factor model (p = 0.031). Conditional analyses reveal that the liquidity factor premium is strongly countercyclical, tripling during NBER-dated recessions and periods of elevated financial stress, while the volatility risk premium is highest at intermediate horizons of three to six months. These findings support a unified risk-based interpretation in which investors demand compensation for exposure to systematic illiquidity and aggregate volatility shocks, and suggest that the apparent anomalies associated with liquidity and low-volatility strategies reflect rational equilibrium pricing of these risk dimensions rather than behavioural mispricing.
1. Introduction
The central question in empirical asset pricing is why different securities earn different average returns. The Capital Asset Pricing Model of Sharpe (1964) and Lintner (1965) offered an elegant answer: expected returns are determined solely by covariance with the aggregate market portfolio, and the market beta is the sole priced risk factor. Yet half a century of empirical research has demonstrated that the CAPM provides an incomplete description of the cross-section of expected returns. Size, value, momentum, profitability, investment, liquidity, and volatility have each been shown to predict returns in ways that the market factor alone cannot explain. The proliferation of factors, sometimes termed the "factor zoo" (Cochrane, 2011; Harvey, Liu, and Zhu, 2016), has generated both intellectual progress and considerable confusion about which risk dimensions are genuinely priced and which represent redundant repackaging of common underlying sources of systematic risk.
This paper contributes to the literature by proposing and rigorously testing a parsimonious, unified factor model that augments the traditional market, size, and value factors with two additional factors designed to capture compensation for bearing systematic liquidity and volatility risk. Our approach is motivated by a substantial theoretical literature demonstrating that both illiquidity and aggregate volatility represent state variables that investors care about and for which they rationally demand compensation. Acharya and Pedersen (2005) develop an equilibrium model in which commonality in liquidity generates a liquidity-adjusted CAPM; Brunnermeier and Pedersen (2009) show that funding liquidity and market liquidity can interact to generate liquidity spirals during crises; and Ang, Hodrick, Xing, and Zhang (2006) demonstrate empirically that exposure to aggregate volatility innovations carries a negative price of risk, implying that assets with high sensitivity to