interpreting coefficients in fixed effects logit

[Charles Williams]

Hi all,

Sorry this is more of a stats question than stata-specific. I have a regression panel model that uses logit fixed effects with a level variable and a change variable for a variable of interest. By level, I mean the variable is simply the value of the variable at time t-1 and by change I mean that the second variable captures changes to the prior variable one period prior (that is, between time t-2 and t-1).

I recently had a discussant at a conference tell me that the interpretation of the level effect should be change in the variable and the change variable is the second derivative. I understand that these estimates are identified on these changes, but I can find no statement of this in any stats material online. In fact, some notes on fixed effects OLS are very specific that the interpretation of the coefficients should be the same as regular coefficients in OLS.

Anyone else seen or heard this interpretation of fixed effects estimates before?

Thanks,

Charlie

[Carlo Lazzaro]

Charles:
welcome to this forum.
As per FAQ, please note that your chances of getting helpful replies are conditional on posting what you typed and what Stata gave you back. Thanks.
Descriptions are not helpful as they would seem to the poster.

[Phil Bromiley]

Let me try to explain what I think your discussant intended. In y = b x(t) + c (x(t)-x(t-1) + e, ignoring that x(t) appears in the second term,

dy/dx= b dy/d(x(t)-x(t-1)) = c. But x(t)-x(t-1) looks like the change in x which is like a derivative, giving you what the discussant described as a second derivative. That is, c is the effect of a change in the change in x.

This over simplifies the problem, since x(t) does appear in the second term. And, almost everything gets more complex in non-linear estimators.