Cross-equation test in Panel Analysis

[Pryanka Boerio]

HI,
I'm trying to compare coefficient from different regression models with panel data.
In a normal regression I would use cross-equation test, that allows to test for the difference between two regression coefficients across independent samples. The test answers the question: “does b1 = b2?”; where b1 reflects the effect of explanatory variable within group 1 and b2 is the effect of the same variable within group 2.
code:

regr Y X1 X2 if X3 == 0

est store zero

regr Y X1 X2 if X3 == 1
est store one

suest zero one

test [zero_Y=one_Y]: X1

If I do the same thing after "xtreg, fe" (fixed effect) the output is:

Constraint 1 dropped

chi2( 0) = .
Prob > chi2 = .

Is there a way to do cross-equation test using "xtreg, fe"?
thank you

[Clyde Schechter]

I don't know how you are even getting to test [zero_Y = one_Y]: X1, as -suest- simply does not support -xtreg, fe- and the -suest- command itself should break with an error message to that effect.

If you want to do a cross model comparison like this with -xtreg, fe-, you have to do it using interaction terms.

xtreg Y i.X3##c.(X1 X2), fe

The contrast between the coefficient of X1 when X3 == 0 and when X3 == 1 will appear in the regression output in the row for 1.X3#c.X1.

Note that the above code assumes X1 and X2 are continuous variables. If either is not, change the code accordingly.