Method to compare 2 models of different years (not by a dummy variable)

Question

I'm a student in economics having trouble with a research in SPSS (BINARY LOGISTIC REGRESSION).

I'm using 2 different datsasets with different samples, same variables, each dataset represents a different year (economic recession year VS non-economic recession year)(SPSS).

I want to investigate the effect of my IV on my DepV in each dataset + I want to compare each individual IV for the economic recession year to see if there was a significance greater/ smaller effect as compared to the non-economic recession year.

PROBLEM: The database where i got my data from (NSSBF) uses in the 2 years different definitions in some variables. I have to keep this in mind when interpreting the results. (example: 1993: > 50% in hands of family = family firm , 2003: 100% in hands of family = family firm). If I copy the 2nd dataset into the first and make a dummy variable for year (to do an interaction effect with my IV), I get the two different definitions mixed in the IV?. So then my interaction is correct but not my single effect of family firms (without comparison of the years).

Can someone please help me with this? You have to excuse me for my limited knowledge in statistics, but I'm trying for days to figure this one out. Thank you!

Answer 1

Definition differences are the bane of analysis. Sometimes, though, data sources provide values under old and new definitions or retroactively redefine the older data for consistency. So I would check that possibility first.

If that fails, you could look at the summary characteristics of the old and new data to see if you can detect changes that seem to be due mainly to the definition change. If the distributions look about the same, you might proceed provisionally on the assumption that the definition changes don't matter, but there will always be doubt. Of course, if the distributions do look quite different, then you are be pretty sure that you can't make a comparision.