Comments on Bauer and Hamilton
Nov 5 2015. Comments on Robust Bond Risk Premia by Michael Bauer and Jim Hamilton, at the 5th Conference on Fixed Income Markets, San Francisco Federal Reserve, Nov 5 2015. I look at the evidence whether macro variables help to forecast bond returns. It turns out a trend does even better, pushing the R2 up to 62 percent! That finding suggests that specification issues rather than distribution theory are the central problems. I don't find that the Bauer-Hamilton effect size distortion is big. I document measurement errors in the data, which are a good target for econometric help. I opine we need to spend less attention on one asset at at time forecasting and more attention on the factor structure of expected returns across assets, and how that structure lines up with covariances of returns with shocks. There is also a little example of perfect non-spanning in a term structure model; bond yields have an exact one factor model, but a non-spanned factor drives expected returns. Programs and data (zip)