Obama Wants Statisticians!

“Politics as done by Martians.” That’s Peggy Noonan’s phrase, to describe President Obama’s call for and use of statisticians and data miners in his re-election campaign.

The hiring of such as I in politics is, she says, “high-tech and bloodless.” Well! Martians, are we! Bloodless automatons, is it! Look here, Noonan, it is you who are partly responsible for electing the Ma-Who-Would-Use-Statistics. It was you and Chris Buckley and other would-be conservative Washington insiders who openly gushed about Obama’s charm and, as spoofed by Iowahawk, the cut of his jib. Obama wouldn’t be in the position to call for statisticians if it weren’t for your ill-timed interventions.

So don’t you go disparaging statistics. Statistics can disparage itself perfectly well without outside help.

Obama’s job announcement is at Kaggle (a site for data miners). In part:

Using statistical predictive modeling, the Democratic Party’s comprehensive political database, and publicly available data, modeling analysts are charged with predicting the behavior of the American electorate. These models will be instrumental in helping the campaign determine which voters to target for turnout and persuasion efforts, where to buy advertising and how to best approach digital media.

Our Modeling Analysts will dive head-first into our massive data to solve some of our most critical online and offline challenges. We will analyze millions of interactions a day, learning from terabytes of historical data, running thousands of experiments, to inform campaign strategy and critical decisions.

Statistical models can be used to determine which voters to target for turn out. But it remains an open question how useful these models are. The problem is in verifying. If you say there is a high probability that Bob will turn up to vote, you have to go to the polls and wait for Bob to show or not. Or you ask him later whether he did. But Bob might lie to you, as people often do to pollsters.

The frightening thing about the job announcement is the “Democratic Party’s comprehensive political database.” How comprehensive is “comprehensive”? I’d be curious to see if my name is in there as a Enemy of the People.

What Obama is looking for is different than a pollster. These are folks who are used to test language that is most harmonious to certain voting segments. The Democrats are currently using the phrase “balanced approach” as a euphemism for “raise taxes” in the budget debate because some pollster found that this phrase was better than others. We don’t know what these other phrases are presumably because they tested badly.

Republicans use these same strategies, of course, proving that politicians of every stripe often say what isn’t so in an effort beguile the public.

I searched for current Republican candidates who are as aggressive as Mr Obama appears to be in his use of statistics. All I could dig up was an old article by a data guy who thought he had the key to George Bush’s re-election.

The guy is Andrew Gage. He built the same kind of “microtargeting” models the Obama camp wants to build.

When the election was over, the Republican National Committee commissioned a poll to figure out whether Gage’s suppositions about why people voted were accurate. Gage’s models predicted voters’ tendencies with 90 percent accuracy, according to Dowd, and Gage was hired to microtarget the 16 or so battleground states in the 2004 election.

Gage claims that microtargeting is akin to unraveling your “political DNA.” But ninety-percent accuracy? That is so phenomenal a number that something is not right. If the model predicts merely who will as opposed to who won’t vote, then perhaps that accuracy can be had, but I doubt it. If the model is said to predict why a person voted the way they did, then I can’t believe the ninety-percent accuracy.

Note that model accuracy was assessed by a poll. Meaning they called a small sample of the people who Gage modeled and asked them questions. How easy it is to ask, “Did you vote because of why we said you would?”

The WSJ article said that Mitt Romney used Gage for his failed campaign for the Republican nomination last time around. Evidently he didn’t hit that ninety-percent mark a second time in a row. No idea whether Romney will engage him again.

8 Comments

  1. What exceptionally good news for the many statisticians out there looking for a means to reduce student loan debt loads and/or put beans on the table! I can think of no more inspiring and fulfilling work than being a team-player of the crew enabling the greatest Harvard Law graduate ever achieve his ultimate destiny! A second term! Just what the world needs.

    Looking over the Kaggle blurb, however, I couldn’t help BUT notice the somehow
    serendipitous but hopefully prescient placement of another statistical activity. What do these people know? And when did they know it?

  2. Sean Peake

    So, OB wants data miners instead of coal miners?

  3. DAV

    Well, that clinches it. The re-up is inevitable. The Mayan model may be spot on (+/- 60 days) after all.

  4. bob

    Why not employ statisticians to predict his electoral win? There seems to be lots of money to go around.

  5. Gary

    Matt, you quickly are becoming my favorite Martian (http://en.wikipedia.org/wiki/My_Favorite_Martian).

    Who knew Ray Walston was a statistician?

    Peggy often has a sense of emerging trends and the American mindset, but sometimes she gets swayed by the inside-the-beltway mentality. However, her measure of blame for the election of our junior high school class president is small.

  6. GoIllini

    You and Ms. Noonan might be interested in reading “Franchise,” a short story by Isaac Asimov, published in 1955. (Reprinted in his collection “Robot Dreams.”) The general story line is about a time when only one person in the country is authorized to vote. All election results for the country are then obtained by computer modeling and statistical analysis of that voter’s responses.

Leave a Reply

Your email address will not be published. Required fields are marked *