Wednesday, May 4, 2011

Prognostication and the Punditocracy

Paul Krugman: America's Most Accurate Pundit!
From Hamilton University in Clinton, NY comes a very interesting study entitled Are Talking Heads Blowing Hot Air? An Analysis of the Accuracy of Forecasts in the Political Media (pdf) on the accuracy of our pundit class.  The findings are fascinating, though not surprising to anyone paying attention.  Liberals are more accurate at predicting outcomes than conservatives.
Good predictors tend to be liberal and are not lawyers. More rigorous study can confirm our findings, especially the question of whether partisanship has an impact on an individual’s ability to make accurate predictions. There were nine prognosticators who were classified as “good” predictors. They were Paul Krugman, Maureen Dowd, Ed Rendell, Chuck Schumer, Nancy Pelosi, Kathleen Parker, David Brooks, Eugene Robinson, and Hank Paulson. Five of these were journalists at the time of our sample. Six are considered liberal, and the average partisanship score was 5.87 (with 5 being a perfect moderate.) Three are female and one is black. It is clear that there is a significant amount of diversity in the “good” category.

And who should you not listen to? Individuals who hold law degrees are less accurate when making predictions. Conservatives, according to our data, are also less accurate. But it is also important to keep in mind that the overwhelming majority of demographic factors have no bearing on a prognosticator’s accuracy. Gender, race, and age are all irrelevant, as are most career path choices, such as becoming a journalist. (emphasis added)

My favorite economist, and the bane of the Theoclassical Economists, Paul Krugman, scored highest of the whole bunch.

Krugman was also uncommonly accurate, only missing one prediction and hedging on one other. His powers of prognostication were impressive, but primarily confined to his field of expertise — he is, after all, a Nobel-winning economist. (emphasis added)

Worst of the bunch? The ever hacktacular Cal Thomas who's predictions proved much worse than a simple coin toss.

Although occasionally Mr. Thomas was close (predicting the Nobel Peace Prize would go to Bill Clinton after Bush left office as a political statement when many would argue it went to Obama for the same reason), more often than not his predictions were overly supportive of the Republican party (predicting a Republican president, the end of immigration law enforcement under a liberal Congress, and Palin lifting her ticket to victory). Another Republican influence in Mr. Thomas’ prognostication can be seen in his insistence that “the next terrorist attack” is “coming soon.” Cal Thomas discussed at length this perceived threat, yet none actually occurred. Mr. Thomas focused on a short time frame, yet this did not aid his prognosticating accuracy as much as his Republican support hurt it. (emphasis added)

I ran a basic correlation analysis of the relationship between the ideology score and the PVS score and found a moderate correlation between the two (0.433781038) indicating that, in general, based on this sample, the more liberal a pundit is, the more likely they are to be correct.
The Wheel of Pundits! (source data)
True statistical evidence that reality tends towards a liberal bias.  Well done Holly Donaldson, Russ Doubelday, Scott Hefferman, Evan Klondar, and Kate Tummarello!

No comments:

Post a Comment