U.S. Election Mega-analysis

Political pundits are second in line after the politicians themselves in putting spin on political poll data to make those data sound as if they support a particular interpretation. But there is another class of analysts who do not prognisticate. Instead, they simply examine the data and tell what those data show at this time.

Nate Silver of the New York (NY) Times has gotten a lot of publicity recently for his versions of this sort of work, but there are several others who are doing similar work (and to me, some are maybe even better, but let’s not argue about that right now). These people aggregate data from the polls (and, in many cases, other sources of evidence) to arrive at statistically dispassionate estimates of the situation. They don’t use hunches about momentum, ad-buys, and so forth. They follow the data.

As of this Sunday-before-the-Tuesday-of-the-only-poll-that-really-matters, the one when the voters go to the polling places and vote, there is a consensus among these geeks: Were the election to be held today, B. Obama would win more than 300 electoral college votes.

If one’s not accustomed to reading statistics—statistics? sadistics? what’s the difference—then it’s a lot easier to go with the good stories that the pundits weave. “The polls show it’s nearly a dead head!” “Too close to call.” “Candidate X is closing strongly.”

For those who have their wits about them, though, the data are pretty powerful. Here’s a list of four sensible sites that apply evidence and rational statistical analysis rather than belt-level instinct and blather:

  • http://election.princeton.edu/ Sam Wang, a cog psych prof at Princeton, uses the cleanest take on aggregating polling data;
  • http://votamatic.org/ Drew Linzer, a political science quant prof at Emory, uses a regression approach that narrows as more data become available, adding polling data along the way;
  • http://horsesass.org/ Darryl Holman, an anthropology prof at University of Washington uses Monte Carlo simulations; there are multiple bloggers on HA, so one has to look for Professor Holman’s posts;
  • http://fivethirtyeight.blogs.nytimes.com/ Mr. Silver’s widely known analysis and commentary.

The modelers probably don’t sell many cars, beer, or other products, I guess. But those pundits surely must. Gotta keep the viewers and readers coming back for more!

Behind the scenes, of course, I hope that the current candidates for the U.S. presidency (and other offices) are getting really honest analyses from their pollsters, that they are saying something akin to this: “Mr. [insert candidate’s name], you have an [nn]% chance of winning. Your best path to victory, sir, is to campaign heavily in [insert states] on these [insert issues] themes.” If the candidates are not using analyses of the quality of Mr. Wang’s, Silver’s, Holman’s, and Linzer’s, they should hustle their bustles off to their local quant shop and hire a good data analyst. Forget the guys in the back room who are blowing smoke.


1 Comment

Filed under News, Notes and comments, Politics, Science, Technology, What I'm reading

One response to “U.S. Election Mega-analysis

  1. Late Sunday over on his Princeton site, Sam Wang provided a run down on aggregators and modelers, describing the pros and cons for three of these and several other sources: http://election.princeton.edu/2012/11/04/comparisons-among-aggregators-and-modelers/. He didn’t include Professor Holman’s approach.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s