James Clive Mathews asks why we have the small countries represented. He points out, correctly, that a simple random sample of the EU as a whole might easily leave out the small countries. And he asks, if we are really presenting this as a scientific sample of Europe how can we adjust the sample to ensure representation of the small countries?

The answer turns on how we represent the small countries. By employing stratified random sampling (rather than simple random sampling) we can ensure representation of the small countries and, in theory, actually reduce sampling error. If we have separate strata that are mutually exclusive (in this case separate countries) and we randomly sample from each, we can actually produce a more representative sample of the overall population.

While there are some modest political compromises built into our pegging the stratification to the proportions in the EU Parliament's allocation, the project has generated good data to evaluate its representativeness, as well as to evaluate the opinion changes and what motivated them.

Given the centrality of the nation-state to the whole agenda (what issues should be left to nation-states and what issues might in some way be decided at an EU level), we needed to represent the individual states as well as the EU as a whole. In this case, every nation state was represented and the whole sample of 362 can be compared to the 3,500 (see details).