Andrew McGuinness of the blog AnomalyUK has pointed me to a study by Alesina, Glaeser and Sacerdote that may have been the basis for Paul Krugman's claim that racial diversity (plus racial animosity) explains America's smaller welfare state in comparison to Europe. I haven't read the study. But given that the Scandanavian countries have the least racial diversity and the most generous welfare states, and the US the reverse, a correlation across OECD countries does seem plausible.

Andrew rightly notes that "all this is quite orthogonal to the argument over the effectiveness of state welfare systems in reducing poverty."

If there is a correlation across OECD countries, it raises the question of reverse causation. A country with an expensive welfare state can't afford to just let the whole world in. I experienced this effect firsthand some years ago when I was offered a job in Australia, but the immigration authorities barred entry of my kids (who need special education). A country with an expensive welfare state will impose immigration policies that tend to maintain its ethnic homogeneity. Which are the racist countries now? If we consider the racially disparate impact of immigration barriers against the world's poor, the causal story becomes quite different from Krugman's: America can afford to be less racialst in its immigration policies (e.g. against Latinos) because it has a less expensive welfare state.

Testing the correlation across US states would remove the immigration-restriction factor, because mobility across states is unrestricted. But that also means that we need a time-series test, because if the (disproportionately black and Latino) poor have migrated to states with larger welfare benefits, a cross-section will spuriously suggest (contra Krugman) that a whiter state population causes lower welfare benefits.

ADDENDUM: Having now perused the Alesina-Glaeser-Sacerdote paper, it's obvious that Krugman was drawing heavily on that paper -- so heavily that it's shameful that he didn't mention it in his column.

2nd ADDENDUM: Frank Stephenson has pointed me to a paper by Erzo Luttmer [Journal of Political Economy, 2001, vol. 109, no. 3], which seems to reinforce AGS. Its abstract states that "levels of welfare benefits are relatively low in racially heterogeneous states."