Reghdfe and constant term

[Petri Ferreira]

Hi all. I have read both the following posts regarding regdfe and the constant term (https://www.statalist.org/forums/for...fe-and-predict and https://www.statalist.org/forums/for...-effects-model). I think I understand the basic idea but I want to make sure that I interpret the constants correctly.

I have a model in which I regress Value on Book Value of Equity (an accounting construct) [Value = BVE]. I do this for Type A firms. Each manager has companies which he values and I therefore include manager fixed effects in addition to year fixed effects. I am able to back out the constant as suggested in the posts above. I therefore have a manger fixed effect (for each manager), a year fixed effect (for each year) and an overall constant.

Suppose I have the same managers that value a different type of firm (Type B firms) during the same time period. I use the same approach as above [Value = BVE] and back out the constant term. I therefore again have a manager fixed effect (for each manager), a year fixed effect (for each year) and an overall constant.

My question: Can I compare the constant term of the two models? And therefore conclude that when managers value type A firms the apply a larger (smaller) constant than to type B firms (on average of course)?

[Phil Bromiley]

You'll increase you chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

I suspect you can do what you're saying, but I would be inclined to estimate the entire model at once using factor variables to allow different parameters for the different types of firms.