In your original objection you pointed out that "he claims that no two democracies have ever made war on each other. But such a claim is misleading, because it uses a) an ideological definition of 'democracy' and b) a narrow definition of 'war'."
to which I replied "What of our tight definitions of socialism (and mine of free trade) on this board? i think his wording opens eyes to the fact that we are not living in defacto democracies regardless of their names."
Hence those example are the kind of loose representative democracies (SDF and I are discussing it below) whose claim to be 'democratic' are weak at best.
It would be useful if Rummel spent more time defining these terms to support what appears to be very credible data. the thrust of his conclusion given in the summary 3d 'blanket' graph is that a significant relationship exists between the extent of a governments power over its subjects and the degree of murder by said government.
He doesnt, to my knowledge, say that if you have a big government that therefore it *will* kill millions - only that in the past such has been observed. Its very difficult to draw a causal relationship from data like this as it can only reveal a significant correlation - It then has to be backed up by actual evidence of gulags etc in order to show government as cause.