Fixed Effects Regression - Binary Variable Can Only Equal 1 Once

[Daniel Brown]

Dear all,

I am working with a dataset looking at employee departures across a group of firms. The dataset is organized on a firm-inventor-year basis, where the binary dependent variable is equal to 1 if the employee leaves, and 0 if the employee stays. Once an employee leaves (Move=1), they never appear in the dataset again. I'm already including firm and year fixed effects. Am I also allowed to use inventor fixed effects? Stata lets me include them (with reghdfe), but I'm a little confused on what they're doing - if the dependent variable can only equal 1 once for a given inventor, I would have thought there was too little within inventor variation for fixed effects to work.

I'd be grateful to anyone who could clarify this for me!

Many thanks!

Dan

[Carlo Lazzaro]

Daniel:
welocme to this forum.
Please follow the FAQ on how to post more effectively. Thanks.
That said:
1) I'm guessing that you're dealing with a linear probability model;
2) as you correctly surmised, a too little within panel variation for time-invariant variables makes the -fe- estimator unhappy;
3) from your description, you can probably stop at firm and year fixed effects.