It is a panel regression with funds and monthly observations of excess returns.
xtset Fund NYear
Array
regress ExcessReturn MktRF4 SMB4 HML4 MOM4 i.NYear i.Fund
Array
I keep time and firm fixed effects to be consistent when comparing results.
regress ExcessReturn NBERdummy MktRF4 SMB4 HML4 MOM4 MktRF4DUM_NBER SMB4DUM_NBER HML4DUM_NBER MOM4DUM_NBER i.NYear i.Fund
Alpha (Fixed-Effects Unconditional model, no dummies): -1.751 (-3.98)
Alpha (Fixed-Effects Conditional model, dummies; expansion): -1.751 (-3.81)
Alpha DUM (Fixed-Effects Conditional model, dummies; recession): -0.800 (-9.29)
ISSUE: My interpretacion here would be that alpha during expansions (control group) is -1.751 while alpha DUM indicates the marginal contribution to alpha when dummy equals "1". That's during recession periods so alpha on recession would be -2.511.
However, something does not look right to me. The results for the constant term in the unconditional model (whole sample) should not be the same as the constant term for expansions (dummy = "0") in the conditional model as this only accounts for the 70% of the sample. I would expect it to be around -1.90 or so to account for the more negative alpha during recession too.
Is my interpretation wrong? Do fixed effects play a role affecting the result in an undesirable way that I am not understanding? Do they nullify my interaction terms or dummies? Is my model not correct for doing what I want to achieve (distinguish between expansion and recession)? If so, how can I correct it?
0 Response to Odd results in fixed effects panel regression with interaction terms
Post a Comment