Posts Tagged ‘election’

Bad Model

Friday, October 5th, 2012

Way back in late August (yes, the distant past of seven weeks ago), these chuckleheads at the University of Colorado predicted a victory for Mitt Romney in November, based on a model that has “correctly forecast every winner of the electoral race since 1980.”

There’s a bunch like these every election year; a prediction model based on various key factors like the tides or the pork belly futures or what-have-you, that has correctly predicted the outcome in “every Presidential election since ____.” And every election year several of them are wrong for the first time.

This one was especially suspect, since it used its standard bellwethers but ignored current polling (also, the prediction was announced before Romney went into full-time gaffe-generating mode). You can’t trust any prediction that suggests, as this one did, that Obama will receive fewer than 242 electoral votes. That’s the tally of states that have voted for the Democrat in all of the last five elections, like Pennsylvania, California, and Illinois. Polling in these states remains rock-solid behind Obama, even in Wisconsin, home of Romney’s mendacious running-mate, Paul Ryan. Short of the proverbial dead girl or live boy, Obama definitely will win all of these states and their electoral votes.

The only state to look at, really, is Florida. Romney simply can’t win the election if he doesn’t win Florida, so if/when the Sunshine State is announced for Obama on election night, that will mean Obama has 242 + 29 = 271 electoral votes, and the election is over. Gnash your teeth or celebrate as appropriate.

If Romney wins Florida, Obama still has several other paths to those 29 electoral votes, through other swing states such as Ohio (18 electoral votes) plus Virginia (13), or Ohio and some combination of New Hampshire (4), Colorado (9), New Mexico (5), and Nevada (6). It’s tough math for Obama without Florida, but it’s impossible for Romney without Florida.

Of course, Nevada and New Mexico aren’t really swing states anymore, thanks to their large Latino populations and the Republican Party’s mystifying decision to alienate them through regressive immigration policies, so Obama’s starting point on election night is closer to 253. Add Ohio and we’re done. Florida is just gravy.

Now, you may be thinking, “won’t Romney’s big debate win change things?” It could; debate performance can make a big difference in close elections. In 1960, Richard Nixon looked sweaty and nervous on TV across from John F. Kennedy’s youthful good looks, and that is thought to have tipped the 1960 election to JFK. And in 1980’s debates, Ronald Reagan landed a few zingers on Jimmy Carter, who looked dour and discouraged. But there are generally three debates, like this year, and in my experience the candidates usually trade wins and then parry to a tie in the third, so they’re a wash. Obama’s lead has been small but consistent all year, and at this point the “undecided” are really the “don’t cares,” so I don’t expect the debates to cause any big poll movement. Most people may think Romney won the debate, but that won’t change people’s minds about who they’ll vote for.

The Colorado professors haven’t weighed back in since their big reveal in August, but almost immediately afterward The New Republic had this takedown up on its website. That story further links to a post on the Five Thirty Eight blog explaining why election predictions based on “fundamentals” are inherently flawed.