About a month ago, I posted on the dubious finding that half of bankruptcies are medically related. I just can't seem to let this baby go.

Both this good article in Forbes and this article in the Atlanta birdcage liner cite the study that came up the 50% statistic. So, a la Don Boudreaux, I've been firing off letters to the editor.

The author of the Forbes piece sent back an interesting question: Since Medicare expenditure per enrollee varies widely across states (lowest state Hawaii=4165 per enrollee, highest New Jersey=7994), couldn't medically related bankruptcies also vary widely? This is a reasonable question--Medicare expenditure per enrollee might reflect interstate differences in healthiness and it might capture any systematic treatment differences across states. If such differences in healthiness or treatment intensity differs then there could be big differences in medical expense related bankruptcies.

So I found some data on bankruptcies and Medicare expenditure per enrollee (scroll down to the download for 2001 Medicare expenditure by state). And the result? There is essentially no correlation (0.028) in Medicare expenditure per enrollee and the number of households per bankruptcy filing. Though not statistically significant from zero, the point estimate actually has the wrong sign.