Alpha Accounting
I present a decomposition based on the principle that an equity portfolio’s CAPM alpha is the value-weighted average of the CAPM alphas of its underlying dividend strips, each corresponding to a specific cash-flow maturity. This decomposition shows that the long-short CAPM alpha of an equity anomaly can be broken down into a term reflecting cash-flow duration differences and two additional terms arising from differences in maturity-specific CAPM alphas. By measuring their relative contributions, this decomposition provides insight into whether cash-flow duration acts as the main driver of CAPM alphas, thereby simplifying the understanding of equity anomalies. To apply it empirically, I construct a new and comprehensive dataset of synthetic dividend strips. This dataset addresses the limitations of dividend futures by spanning a longer sample period, covering a broader set of firms, and being available for all maturities. My results suggest that differences in maturity-specific CAPM alphas explain 92% of the CAPM alphas of major equity anomalies, while cash-flow duration accounts for only 8%. Hence, cash-flow duration plays a more modest role as a unifying driver of anomalies than previously thought, offering limited help in taming the "anomaly zoo."
Global Equity Yields
joint with Jens Kvaerner
We use the model of Giglio, Kelly, and Kozak (2023) to construct a panel of global equity yields. We revisit stylized facts about equity yields, primarily based on US data, and provide several new results. On old facts, we study the dynamics of global equity yields, their slopes, and the relative contribution of risk premium and growth expectations in explaining variation in yields. On new facts, we study co-movements in risk premia and growth expectations across markets and Fama French portfolios, estimate the term-structure of the global equity risk premium, and link yields to changes in exchange rates and future macroeconomic outcomes.
Long-Run Factor Returns (draft coming soon)
joint with Jens Kvaerner and Stig Lundeby
We construct a new dataset of U.K. equities spanning 1860 to 2019 to study the long-run behavior of equity factor returns and its relevance for portfolio choice. We find strong evidence of autocorrelation at multi-year horizons. These findings reject the assumption that factor returns are i.i.d. and motivate horizon-specific allocations. To evaluate optimal portfolios, we introduce an “imagined distribution” that regularizes the empirical return distribution by allowing for future return realizations outside the historical sample. This approach improves stability and reduces sensitivity to in-sample overfitting. Out-of-sample tests in both the U.K. and U.S. show that long-horizon investors are willing to pay a meaningful portion of their initial wealth to access horizon-specific strategies.