The claim came from Ed Gillespie, who was defending Romney's record in Massachusetts by pointing out that median income in Massachusetts rose by $5,500. The problem, as Bernstein pointed out and PolitiFact acknowledged in the story, comes from the fact that family income actually dropped slightly after adjusting for inflation.
Saying that income went up by thousands of dollars while it dropped in terms of real dollars is flatly misleading. Bernstein has a good foundation for a sense of outrage.
But even though Bernstein's basic point is accurate, he oversteps a bit by assigning the statement a rating of "Mostly False." That's because PolitiFact gives each of the ratings a definition. Here's the definition of "Mostly False":
MOSTLY FALSE – The statement contains an element of truth but ignores critical facts that would give a different impression.
When PolitiFact started out, the "Half True" rating appeared to fit statements like Gillespie's:
HALF TRUE – The statement is accurate but leaves out important details or takes things out of context.
PolitiFact's change of the definition of "Half True" (with zero fanfare outside of us) makes it no longer quite fit fundamentally misleading true statements:
HALF TRUE – The statement is partially accurate but leaves out important details or takes things out of context.
A more important point underlies Bernstein's criticism of PolitiFact: PolitiFact uses an unwieldy rating system. Many statements do not fit the ratings as PolitiFact defines them, particularly when PolitiFact rates two statements at a time and averages the ratings.
So Bernstein scores a glancing blow against PolitiFact. It doesn't hurt PolitiFact much but it does expose a critical weakness.