The authors first describe the time-varying properties of both the distribution of firm betas and the behavior of the high-beta minus low-beta portfolio. They then describe the one-step and two-step regression models that define their conditional CAPM and present the data used to construct beta-sorted quantile portfolios. The statistics of the regressions are used to show the low significance of the low-beta anomaly, which forms the basis of their conclusions. They also attempt to explain the time-varying betas of the sorted portfolios on the basis of, among other things, valuation effects, IPO timing, and capital structure decisions. These ideas are supported by regressions using the betas as dependent variables.

In the current climate of relatively high enthusiasm for low-beta-focused products, arguably insufficient thought has been given to the origins and potential persistence of the alleged anomaly. Academics have been debating the perceived challenge to the CAPM ever since such factors as value and size readily achieved a foothold in the industry mindset. The identification of additional factors that are robust predictors of firms’ betas is indispensable in helping to understand how their performance is affected by the market’s equity premium and volatility. Investors with a low-beta bias, either by design or as a byproduct of another focus (e.g., higher-quality names), should be interested in the authors’ evidence that the historical low-beta premium is spurious.

This research also informs a deeper understanding of the cyclicality of the market risk premium, describing both the time variation in market betas and the relationship between market betas and macro variables. Understanding the relationship between the macro environment and the distribution of risk is vital to any asset allocator who takes a position on the business cycle.

The CAPM relates expected individual security returns to the risk-free rate and an undiversifiable market risk premium. The CAPM cannot capture how individual security returns may vary (it assumes a covariance of zero) under different market regimes. A conditional CAPM approach allows for variation in the beta over time to be more closely considered.

The authors formulate a conditional CAPM with (1) conditioning variables that include lagged long-term and short-term betas and (2) macroeconomic variables that include dividend yields and credit spreads. They estimate the regression model using the generalized method of moments and assess the significance of findings using Newey–West corrected standard errors. Most of their conclusions rest on comparisons of the significance of both conditional and unconditional alphas, but they provide a range of regressions in support.

The authors seek to demonstrate the importance of separating portfolio performance differences due to conditioning variables and those due to beta exposures through a comparison with the Fama–French multifactor model and find that the results reinforce one another.

Because a conditional CAPM has a far greater number of parameters than an unconditional CAPM, it is not necessarily surprising that the unexplained portion of the cross section of returns is less significant. The recent clamor about the crowding in low-beta investments—see, for example, Marmer (Journal of Investing 2015)—will be borne out one way or another in the short-term performance of the factor. For investors with their eyes on a farther horizon, however, the more pertinent question concerns the existence of a premium and the attractiveness of a permanent systematic tilt. The authors’ conclusions should be read alongside the literature on higher-order moments; the skewness of stock return distributions is a complementary explanation for some of the cross section of returns explored in this research.