The main constraint on which countries to include will be whether data are available. Assuming you have located a data source that has a large number of countries you might want to take different approaches at different stages of the analysis. Starting with a large number of countries can be useful in identifying the overall pattern and then you can narrow down the number of countries using the most polarised or interesting countries to illustrate the findings. Limiting the number of countries has benefits in making the analysis and presentation of results much more manageable. Don't underestimate the difficulty of displaying data for a large number of countries in a way that is easily accessible to the reader. A handful of countries can sometimes serve to demonstrate the points more usefully than including 30 or 40 countries. For example, if all the Nordic countries or all the newer members of the EU are show a similar story then it may be worth only showing a subset in your graphics to illustrate the point rather than including all of them.

If you have a fairly well defined hypothesis about how different countries will feature, you may wish to start with a subset countries while you do the initial analysis and then add in more countries to test the hypothesis in a further set of countries and then make a final selection for the purpose of commenting or presentation of the results.

It is also important to remember that finding no major difference across countries is as valid a research finding as finding systematic differences, particularly if it might intuitively be expected that they would differ.

Aside from issues of substantive interest such as social or political attitudes many international surveys such as the European Social Survey offer the opportunity to analyse the effect of different research methods such as differences in mode.