Comparing regression coefficients between two groups –> violation of independence?

Hi all, I have a question about the comparison of regression coefficients of two groups (using SPSS) and I think this is the correct sub forum for such issues.

Basically, I want to compare France to a bunch of other European nations combined (about seven) using nationally representative samples from the 2013 wave of the European Social Survey.

I thought the best way of doing this would be (1) to throw all of the respondents of the eight countries together to make one big sample; (2) make a dummy variable for each country to control for between-country differences in the DV; (3) proceed normally and just model the data with a regression analysis using the IVs relating to my hypotheses; (4) add France as an interaction term to check to what extent France’s regression coefficients differ from those of the other seven countries combined (in essence like this tutorial).

However, by lumping the respondents of several nationally representative samples together, I’m pretty sure I’m violating the assumption of independent observations. The next logical step would thus be to try a random slopes model. There seem to be two problems, however: (1) if I understand correctly, multilevel models look at the aggregate Level-2 variance, making it unfit to compare one specific Level-2 unit (in my case: France) to the other units; (2) the Level-2 sample, i.e. the eight countries, seems to be far too small (wiki says at least 20 groups; Maas & Hox 2005 even argue that samples under 50 groups may lead to biased Level-2 SEs).

So concretely, my first question is whether I was even correct in dismissing my first two strategies (both the model with the interaction terms and the hierarchical model); second, whether any of you see a solution to my particular problem.

I hope I have provided enough information. If not, don’t hesitate to ask for more details. As a student from the Humanities I’m self-taught in (frequentist) stats, so I apologize in advance if all of this is just too painfully obvious for those with a more rigorous statistical training. ;)

Anyways, thanks a lot for your consideration!
Re: Comparing regression coefficients between two groups –> violation of independence

Hey, a quick update for those of you who might be dealing with a comparable problem. I stumbled upon an article about the so-called "fixed level approach" that I found quite useful: The fixed effects approach as alternative to multilevel models for cross‐national analyses, by Katja Möhring.

As far as I know, the method is basically the same as the tutorial I linked to in my original post (i.e. controlling for country heterogeneity by adding dummy variables). The advantage of the article, however, is that the author makes an explicit connection to hierarchical data-sets when the L2-sample is too small (and not really random either) to apply a multilevel analysis.

See also this manual by Paul Allison.