I am estimating the effect of a goverment policy on labor supply (working hours) using two-way fixed effects. The estimate does not come as expected. When I add in the equation an independent dummy variable (created based on the dependent variable), the estimate sounds more reasonable and significant. In particular, the dummy is 1 if the working hours are above the median and 0 if below the median.
I am thinking about the alternative in which I will instead include an interaction between the dummy and my main indepedent variable (treatment). In this case, I would say about the reason to include this interaction is that there could be differences in response to the treatment between 2 groups above and below the median.
I wonder whether they are statistically viable methods and whether do they make sense?
I am thinking about the alternative in which I will instead include an interaction between the dummy and my main indepedent variable (treatment). In this case, I would say about the reason to include this interaction is that there could be differences in response to the treatment between 2 groups above and below the median.
I wonder whether they are statistically viable methods and whether do they make sense?
Comment