My guess is that he has an extraordinary amount of faith in his macro mode, that his macro model contains an accelerationist Phillips curve, and that his macro model does not allow for sudden large declines in the non-accelerating-inflation-rate-of-unemployment.
Thus my guess is that he believes that "special factors" (the squeeze on health care costs as firms move more people in HMOs, the accelerated fall in computer prices over the past two years, the stronger dollar with declining import prices, and just plain luck) have been masking inflationary pressure over the past two years--and that we are about to see inflation start rising by one percent per year every year, unless the unemployment rate creeps up to 5.5% or higher...
I think there are two chances in three that he is right...
Brad DeLong