Thursday, March 19, 2009

Grading PolitiFact: Christina Romer and "almost all of them"

It is almost uncanny how poorly PolitiFact plays the fact-checking game.

The latest rating from PolitiFact, penned by Alexander Lane, concerns a economic prediction from economic adviser Christina Romer. Romer backed her prediction by appealing to a survey of private forecasters:
"We know that this last week the Blue Chip Economic Indicators came out that surveys lots of private forecasters. Almost all of them are predicting a turnaround in the third quarter and positive growth in the fourth quarter."
Sounds pretty easy to check, doesn't it? Watch Lane turn it into an adventure.
The consensus, or average prediction of the panelists, was that the economy as measured by the gross domestic product would grow 0.5 percent in the third quarter of 2009 and 1.8 percent in the fourth quarter. That was not quite enough information to gauge the accuracy of Romer’s claim,
Correct so far, with the caveat that "consensus" is not nearly the same thing as an average prediction. Consensus indicates general agreement, but one can take an average from a set of numbers that do not generally agree. Lane's sentence tends to suggest otherwise.
so we asked editor Randell Moore for more detail. He said 50 economists were surveyed, with 30 predicting positive growth in the third quarter of 2009 and 45 predicting positive growth in the fourth quarter.
Those are some useful numbers. Obviously with only 50 surveyed we'll find some overlap between panelists predicting third quarter growth and those predicting fourth quarter growth.

But let's review Romer's claim:
We know that this last week the Blue Chip Economic Indicators came out that surveys lots of private forecasters. Almost all of them are predicting a turnaround in the third quarter and positive growth in the fourth quarter.
It appears that Romer is saying that "Almost all of them" are predicting two things, linked with the connector "and." First, "Almost all of them" predict a turnaround in the third quarter. If "turnaround" is equal to positive growth then we have a problem, since 30 of 50, also known as 60 percent, seems perhaps less than "almost all of them."

On the second part of the statement Romer is on better footing, since 45 of 50 predicted fourth quarter growth. It's fair to count 90 percent as "almost all of them."

Now, if Romer has said "Almost all of them are predicting a turnaround in the third quarter or positive growth in the fourth quarter" then we'd have enough information to reasonably rate her statement "True." But that is not what she said.

Somehow, Lane sees it differently:
That does indeed support Romer’s claim, since 45 out of 50 constitutes "almost all" the forecasters.
Apparently we're supposed to forget about approximately half of Romer's statement. And Lane never bothers to deal with it. After sifting through more of the panel's findings and noting that the picture isn't exactly rosy, he offers us the final conclusion:
But Romer is entitled to highlight optimistic forecasts. And it's true that almost all the panelists surveyed in Blue Chip Economic Indicators predicted a rebound by the fourth quarter of 2009. We find Romer's claim to be True.
See? Third quarter prediction all but forgotten.

With half of the statement pretty clearly false, how can a rating of "True" be justified?

That answer is easy. It cannot be justified. And that is the mystery that is PolitiFact. The project is supposedly made up of capable professional journalists. And PolitiFact stresses that its ratings are subject to peer review among other PolitiFact staffers.

So how can the results be this bad?

Grade: D

Note: One potential saving grace for Romer is that there might have been more information in the survey to support the notion of a "turnaround" as opposed to literal economic growth. That offers no help to PolitiFact, however, as we have no indication that Lane explored that avenue at all.

No comments:

Post a Comment

Please remain on topic and keep coarse language to an absolute minimum. Comments in a language other than English will be assumed off topic.