I need help in understanding some results that I have obtained.
I am doing some out-of-sample performance analysis for different targets of volatility in mean-variance optimization where I solely change the way the covariance matrix has been estimated. HC in the figure refers to the sample covariance matrix, whereas GS refers to the Gerber-Statistic based covariance matrix (see here) which is supposedly more robust to outliers and noise in the data.
Now, if we observe the below figures, one can see that the risk-adjusted-return (adjusted according to HC) for GS is generally lower in expansionary times while its much higher in recessionary times. Now, a plausible explanation for this is that GS works better under more volatile times influenced by noise and outliers. However, I can't find any theories that support the fact that returns are more volatile in recessionary times, compared to expansionary times. Do you have any ideas regarding this matter?