Dear all ,

I have performed a fixed effect regression with cluster(id)

xreg y i.FFF i.young FFF#young + Controls, fe cluster(id)

i use an interaction term with 2 dummy variables.

FFF=1 Family firm else FFF =0

young = 1 if firm age < 20years else young = 0


my dependent variable is a measure of innovation

Var | Coef | Std.Err | t | P>t

1.FFF | .0010349 | .0007549 | 1.37 | 0.171
young | .0000637 | .0007209 | 0.09 | 0.930
1.young | 0 (omitted)

FFF#young | -.0001431 | .001038 | -0.14 | 0.890
1 1

margins i.iFFF_PU##i.young

Predictive margins
Model VCE : Robust
FFF Margin Std. Err.
0 | .0045633 | .0001112
1 | .0055404 | .0003584
young
0 | .0047803 | .0002522
1 | .0048104 .0003708

iFFF_PU#young |
0 0 | .0045375 | .0003248
0 1 | .0046012 | .0004298
1 0 | .0055725 | .0006025
1 1 | .005493 | .000499

i am not sure how to interpret the interaction term.

is my assumption correct that the family firm's influence on the mean of the innovation measure decreased by -.0001431 when FFF=1 and young=1 ?


I also don't understand why the interaction term in the regression is negative, and the margins are all shown positive.

Maybe someone would have a hint or advice for me .