Cross-section specific coefficient

[g_mbww]

Hi all,

When doing panel data modelling we have one section where to include regressors that we demand having a cross-section specific coefficient.

If I include there a variable that has non-variability for the all time period for one of the cross-section "A", I would then expect to have no beta coefficient for that same variable (as it wouldn't be included for that retailer)

The opposite is actually happening, and I am getting a coefficient that is very high!

In mind it doesn't make sense because if there is no variablity for that cross-section then no effect could be calculated, but maybe I am doing my maths wrong.

What is happening here? If someone can point me to material in order for me to understand the computations involved I would really appreciate.


Thank you,
Guilherme