Panel Models - Wald Test Sensitivity

[LPD]

Hi,
I'd just like to ask if anyone has any thoughts on the following.

I have several panel models (unbalanced with random effects) based on some fairly standard econometric variables which have been built over a particular development period.

I've been interested in testing model stability on an extended sample (i.e. development plus some recent data). I've been using the Wald Test available in Eviews where the Null Hypothesis is that the model coefficients are all equal to zero when the model is re-estimated on the residuals over the extended sample. For more or less all models the test rejects the null hypothesis that all coefficients are zero (i.e. suggests that model is not capturing the relationships in the data correctly). This surprises me as in some cases I'd expect the relationship between the dependent variable and independent variables to be the same over the extended sample. Also, I don't draw the same conclusions for univariate time series models.

Could the Wald Test be unsuitable for testing the stability of Panel Models (e.g. due to the high degrees of freedom perhaps)? If so, is there an alternative test I could use?

Thanks