Fixed Effects and Random Effects

[Sidra Ilyas]

Hi Dear,

Can you please tell me whether Fixed Effects estimations with xtreg, fe and Random Effects estimations with xtreg, re produce long run or short run estimates?

Thank you.

[Carlo Lazzaro]

Sidra:
-xtreg- (no matter its specification) is devised for short panel dataset (N>T) with a continuous regressand.

[Sidra Ilyas]

Hi Carlo,

Thanks for your kind response. I actually meant something else. For example, GMM is believed to produce short run estimates of the coefficients i.e. the change in independent variables causes the "beta" change in dependent variable in the short run, where "beta" is just a coefficient estimate.

In this line, xtreg FE and RE are short run or long run estimates of the coefficients? In terms of the interpretation of coefficient estimates?

Thanks again.

[Phil Bromiley]

There is an argument that models with lagged dv's on the rhs give short run estimates on the other parameters and without the lagged, dv models give long run estimates. With the lagged dv, you can see why a change in x(t) changes y(t) but then that y(t) changes y(t+1) etc. so the effect of x can be much different in the long term than just the parameter on x. With no lagged dv, change in x just changes that one y(t). Now, that assumes y(t) doesn't later change future x's. It is also under a representation that does not allow for the possibility that x(t) might directly influence y(t) and y(t+1) etc.

[Sidra Ilyas]

Thanks Phil.