Working Paper: NBER ID: w23886
Authors: Valentin Haddad; Serhiy Kozak; Shrihari Santosh
Abstract: Across a variety of asset classes, we show that relative returns are highly predictable in the time series in and out of sample, much more so than aggregate returns. For Treasuries, slope is more predictable than level. For equities, dominant principal components of anomaly long-short strategies are more predictable than the market. For foreign exchange, a carry portfolio is more predictable than a basket of all currencies against the dollar. We show the commonly used practice to predict each individual asset is often equivalent to predicting only their first principal component, the index, which obscures the predictability of relative returns. Our findings highlight that focusing on important dimensions of the cross-section allows one to uncover additional economically relevant and statistically robust patterns of predictability.
Keywords: Predictability; Relative Returns; Asset Pricing
JEL Codes: F31; F65; G0; G1; G12; G17
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| specific predictors (xt) (C29) | expected returns (rit) (G11) |
| aggregation of returns into portfolios (G11) | stronger predictive patterns (C52) |
| slope of the yield curve (E43) | returns of maturity-sorted portfolios (G12) |
| principal components of long-short equity anomaly portfolios (G12) | expected returns of individual anomalies (G41) |
| focusing on relative returns (G11) | uncovering economically relevant patterns (E26) |