Including State Fixed Effects for Cross Sectional Data

[Iman Haupricht]

Hello all,

I am coming across an issue with my code when I attempt to add in state fixed effects to my regression. I am using the Donald and Lang (2007) method which requires collapsing all of the coefficients on the year#treat interaction into 10 cells and then running a second regression using them as the dependent variable. Therefore, I need ALL of the coefficients to be included, which is why I include nocons. However, when I add in the state fixed effects (i.statefip), it omits one of the coefficients. Is there a way to include the state fixed effects such that all ceofficients of year#treat remain ? I am only interested in these values, not necessarily the coefficients on the other variables. This is my code:

svy: reg maternity_leave ///
i.year i.year#i.treat ///
unemp_rte i.statefip ///
i.age_cat i.race_cat i.marst_cat i.educ_cat i.yngch bornus, nocons

For information, I am replicating the method in Rossin-Slater et al. (2013) "The Effects of California’s Paid Family Leave Program on Mothers’ Leave-Taking and Subsequent Labor Market Outcomes". Unfortunately I do not have acces to the authors' code and this part is not detailed in the text.

I would appreciate any input you may have.

Thanks,

Iman

[Iman Haupricht]

I figured out my error! Thanks anyways.

[Clyde Schechter]

Given that others may encounter a similar problem in the future and would be able to learn from your solution, it would be appreciated if you posted how you solved the problem.

[Iman Haupricht]

I removed i.yngch which was an irrelevant control I had left in by accident. When using a control group including 40 states I also switched out i.statefip (individual state fixed effects) for i.region which grouped the states into 4 regions and this caused the model to not be over saturated. Thanks Clyde.