Verifying predictions

Have you ever wondered if all those prognosticators ever get anything right? Probably not.

But in case you have wondered and in case you haven't, a Hamilton College class and their public policy professor have bothered to analyze the accuracy of the predictions of 26 talking-heads.

Their results? After Paul Krugman, the most accurate pundits were Maureen Dowd, former Pennsylvania Governor Ed Rendell, U.S. Senator Chuck Schumer (D-NY), and former House Speaker Nancy Pelosi. "The Bad" list includes Thomas Friedman, Clarence Page, and Bob Herbert.

And:

<!-Quote-!>Quote:

Perhaps most importantly, being a good prognosticator seems to be a product of choices, not birth. Anyone can be good; all they need to do is avoid law school and buy into liberalism as an overarching philosophy.<!-/Quote-!>

Report.

Politics is better as a spectator sport.

Interesting. I wish they had included a larger sample of the actual predictions.

Two more "possible criticisms of the prognosticator value scoring system" that I believe they should have included:

1) The motivation of pundits is not always necessarily to make accurate predictions. They seem to have implicitly acknowledged this in some places; one might say it's irrelevant, as they are testing the accuracy of predictions, not the achievement of individual pundits' objectives, but still it is a criticism of the study that should legitimately be included as a caveat in their results. It might go a long way in explaining why lawyers did not predict as well as journalists, as the latter may benefit more from making accurate predictions.

2) The winners in the 2008 elections were largely liberals and Democrats. If the pundits, especially the politicians among them, were simply biased toward predicting successes for their side, for whatever reasons, then the liberal and Democrat prognosticators ended up looking good as predictors not because of their predictive abilities, but just because their side won. A follow up of the 2010 elections, in which conservatives and Republicans were largely the winners, would test this hypothesis. If true, it would be very detrimental to the study's validity; if false, it would go a long way toward confirming it.

Who verifies the verifiers?

Don't do that. Don't cast doubt on the honesty of the report without any ounce of justification. That's just a very ugly no-no. And the reason it is such an ugly no-no it's because the method is on the report and you can clear all your doubts with the author, so you can verify it.

Jo jo,

1. In the profile of each prognosticator, they try to explain the reasons for their scoring; in politicians it is noted they might make very wrong predictions because of their motives to promote their party and candidate. But in my view that seriously doesn't matter, because they were are looking for honest, accurate predictions, so motives are irrelevant: say the truth while pep-talking. But it is possible to be "accurate", as in Nancy Pelosi's case, because she "could dodge a question with ease, and she avoided finding herself making uncomfortable predictions." She also had the fortune of being a democrat and the democrats won the 2008 election.

2.

<!-Quote-!>Quote:

Democrats seem to be better at predicting than Republicans, but we are cautious in claiming that this is a generalizable conclusion outside of our time period. There may be underlying factors in the 2008 elections that might not occur in other time periods. It's also plausible that partisan strategies may lead to inaccurate predictions by Republicans--Republicans, unlike Democrats, may hope to face who they see as the weaker candidate from the opposing party in November. This may mean that some predictions are not really meant to be predictive--rather, they hope to shape the debate. Of course, it is certainly possible that Democrats really are better predictors<!-/Quote-!>

If you listen to Wendy, Paul Krugman is a right-winger. ;)