I am currently analyzing the IVOL puzzle and I find in a single sort that the higher the IVOL the lower the return. Additionally I do a double sort, first sorting on skew and then on IVOL. I find that after controlling for slew the IVOL puzzle disappears. After that I do another double sort, first sorting on return of the previous month and then on IVOL. Here also, the IVOL puzzle disappears.
Now I am wondering if it is the lottery preferences (investors seeking highly skewed stocks) that explain the IVOL puzzle or the return reversal?
Would it be now correct to do a Fama macbeth Regression ret ~ IVOL + RET(t-1)? Because this show me that only RET(t-1) has a significantly negative influence on returns.