Nate Silver’s Rasmussen obsession

The New York Times’ statistics guru, Nate Silver, is infallible. You see, Silver doesn’t predict election winners and losers; he only gives odds of winning and losing. So even if Silver gave Obama a 99% chance of winning, an Obama loss would still be perfectly consistent with Silver’s projections. In other words, when it comes to measuring Silver’s performance, heads Silver wins, tails his critics lose.

But while his statistical methodology might give Silver cover against charges of liberal bias in his projections, it does little to explain the double standard he applies to Rasmussen’s polling. Dating back to his blogging days at FiveThirtyEight.com, Silver has been on a mission to discredit Rasmussen’s polls, which Silver views as biased in favor of Republican candidates. What’s more, Silver has allowed his Rasmussen obsession to influence his supposedly objective statistical analysis.

For example, after Rasmussen landed near the top of Silver’s 2008 pollster rating, which took into account the 2004, 2006 and 2008 elections, Silver was determined not to allow another strong finish by Rasmussen. Luckily for Silver, a reader of his alerted him to the fact that prior to launching Rasmussen Reports in 2003, Scott Rasmussen ran an obscure polling company called Portrait of America, which performed poorly in the 2000 presidential election.

Silver finally had the information he needed to torpedo Rasmussen’s ranking. Never mind that Portrait of America was a different company from Rasmussen, or that Silver’s previous pollster ratings only went back to 2004. Finding data that could lower Rasmussen’s rating was justification enough for Silver to change the underlying calculations for his entire ranking.

Whatever Silver’s rationale for not including pre-2004 results in his initial ranking, it clearly no longer applied after Silver discovered old numbers reflecting unfavorably on Rasmussen.

Silver’s most brazen demonstration of his Rasmussen obsession, however, came immediately after the 2010 midterm elections. Likely nursing his wounds from his party’s congressional losses and his own inaccurate forecasting, the usually diplomatic Silver ran an article the morning after the elections with the strongly worded headline “Rasmussen polls were biased and inaccurate.

To Silver, the most pressing issue after the 2010 midterm elections was Rasmussen’s polling performance. Never mind that Silver himself was off in his forecasting — unsurprisingly, in the Democrats’ favor — or that plenty of other reputable polling organizations, such as CNN and Gallup, performed no better than Rasmussen. No, to Silver the world needed to be immediately alerted to Rasmussen’s alleged bias.

Of course, this is a far cry from Silver’s behavior when he suspected that Research 2000, the polling company commissioned by the liberal Daily Kos blog, had serious flaws. Back in February 2010, Mr. Silver became “suspicious” about what he described as Research 2000’s “really weird results.” But instead of informing his readers about these problems, Silver reserved his suspicions for a private email to another pollster, Mark Blumenthal, as well as to Daily Kos founder Markos Moulitsas.

Only months later, when Daily Kos officially fired Research 2000 and sued the company for fraud, did Silver inform his readers of his long-time suspicion. In fact, in the very same week that Silver sent his private email to Blumenthal questioning Research 2000’s credibility, Silver published a lengthy blog post discussing a Research 2000 poll, without ever letting his readers know his real thoughts on the pollster.

This isn’t to suggest that Silver should have outright accused Research 2000 of fraud, but if his own statistical analysis indicated serious problems with Research 2000’s polls, Silver shouldn’t have hid this from his readers. I suspect that if Silver had found the same “really weird results” in Rasmussen’s polls, the probability that he would share this information with his readers would be very high — statistically speaking, that is.

These examples are just a small piece of Silver’s larger pattern of Rasmussen obsessing. In fact, a cursory review of Silver’s old FiveThirtyEight blog — which conveniently sorts its posts by topic — reveals that prior to moving to The New York Times, Silver ran four pieces on Gallup, 10 on Pew, 15 on Zogby and a whopping 29 on Rasmussen. And in case anyone is curious, these 29 posts didn’t exactly lavish praise on Rasmussen.

Perhaps all of the attention Silver devotes to Rasmussen would be justified if he were actually able to prove that Rasmussen’s performance is any worse than that of the other major polling companies. But even after all of his attempts to discredit Rasmussen, Silver’s own analysis still only demonstrates that Rasmussen is, at worst, an average pollster. And this is after Silver rigged his entire pollster ratings for the sole purpose of lowering Rasmussen’s rank.

But then again, who am I to question Silver’s motives. After all, I’ve been told he can’t be wrong.

Mendy Finkel is a corporate attorney practicing in New York. He is a graduate of Columbia Law School.