I have an unbalanced panel dataset of around 3000 firms for 5 years. My goal is to see if female-owned firm are less likely to survive, so I have a logit model with female-owned dummy as the key independent variable.
The cross-sectional models for year 2007 and 2009 (if year == 2007, if year == 2009) have significant female-owned variable. However, if I try to run a fixed effect logit across all 5 years then the female-owned coefficient is highly insignificant. I have included the i.year dummies in the panel model, and interact female-owned##year2007 but so far the coefficients are not significant.
So how do I interpret this sort of results? Thank you in advance for your help!