Jean Whit notes that the authors of this piece about Wall Street Journal number crunching about sovereign debt default and bond ratings, WSJ Analysis: Rating Firms Not Effective at Predicting Government Defaults, got their analysis backwards. A classic case of sampling on the dependent variable or percentaging in the wrong direction: how many of the defaults had a given rating rather than how many with a given rating end up defaulting. See Jean’s comment at bottom of post.
