Sunday, October 28, 2012

Krugman's Utter Nonsense (but then I repeat myself)

NYT scribbler Paul Krugman wrote in his blog that anyone who questions the methodology used by NYT writer and statistical guru Nate Silver is an "attack on objectivity."   Well, at least Krugman is consistent; consistently wrong.

Nate Silver is a sports nut and a math whiz.   In the 2008 election he had worked for the Daily Kos and developed a set of statistical models which had remarkable accuracy. He missed on state on the electoral map and called all 35 senate races in that year.   This year, he is pretty clear that he agrees with the modeling (a lot less sophisticated) on Intrade which suggests that the president has a 63.2% chance of winning re-election.   Silver's current odds suggest a larger chance for the President (74.6%). Both forecasts (one based on money "invested" by political junkies and the other by Silver's methodology) show that the President seems to have regained some support after the first debate.
But Krugman yammers "Yet the right — and we’re not talking about the fringe here, we’re talking about mainstream commentators and publications — has been screaming “bias”! They know, just know, that Nate must be cooking the books. How do they know this? Well, his results look good for Obama, so it must be a cheat. Never mind the fact that Nate tells us all exactly how he does it, and that he hasn’t changed the formula at all." Except like most of what Krugman writes, most of his claims are untrue.  Silver has explained the factors he uses but not the specific weightings he uses in each of the states. 

Krugman seems to have been reacting to an article on National Review Online by Josh Jordan titled Nate Silver's Flawed Model.   Jordan makes two comments in the article.  First, he suggests that to come up with his forecasts Silver weights his results, in a way that seems to favor Obama.   The methodology which Silver developed involves some subjectivity - how much weight does one assign to various polls in various states.   Those could be done to come up with a result or because of his own judgments.    Jordan then goes on to suggest that Silver is openly rooting for an Obama victory.  Indeed, in 2008 the Obama campaign shared private polling data with Silver.

I read the Jordan article and thought he raised some interesting questions about Silver's process.  Obviously (and I think reasonably) Silver has not let others peek under the hood.   So the formulas in his algorithm are unknown.   Ultimately, we will know in about ten days whether Silver was a flash in the pan or whether he developed an interesting set of formulas that project election results.  

From my perspective there are two things that are pretty clear in this election cycle.   First, polling is flawed at this point.   A smaller proportion of voters have land lines so getting responses is tougher than it once was.   At the same time we have seen a proliferation of polls - some of which have questionable reliability.  Not all polls are equal.    

Second, 2012 is not 2008.  The President has had a devil of a time getting his approval numbers to get near 50%.  (Silver's macro formulas for 2008 and this year rely on the president's favorability ratings (below 50%), economic growth (less than 2%) and then a whole bunch of other stats including polls and historical voting patterns.    In 2008, there was never much of a question who would win.   That is not true in this year.   At the same time there is that annoying fact every president who won re-election raised his votes in the second election.  If Obama wins, that is unlikely to happen.  No GOP candidate has ever won the Presidency without winning Ohio and at least the RCP averages on the state suggest a small Obama lead.   Silver could be correct about the outcome this year, or he could be wrong, but I will guarantee you that whatever the result, there will be some discussion about whether one can construct a reliable predictive model on elections that will last over time.

I have a friend who is an economist who uses economic data to project presidential elections.  He confided to me a month or so ago that while his model has been accurate in the last ten elections, and that this year projects a Romney victory, he had less confidence in the results this year.

Krugman is unwilling to admit that some prognosticators could have flaws in their models.  He, among all people should understand the frailty of any kind of prognosticators - even former Enron consultants.

No comments: