Ranking the Predictions

January 19, 2009

Georgetown Professor David Walker ranks the different forecasting models: the Iowa Electronic Market came out on top with a 0.2% error, while the RealClearPolitics average (0.3% error) was a close second.

RCP’s success speaks to a larger point about predicting presidential elections: with the abundance of polls, a simple aggregate provides a fairly accurate result.  And, even without polling, plugging a series of macroeconomic variables into a model gets close to the final margin.


What’s Tamping Obama Down

August 25, 2008

Hillary supporters have not yet flocked to the Democrats’ tent, even though they concur with the party platform. Wary of Obama’s slim resume and perceived aloofness, the group remains a roadblock to victory. In this tact, the Republicans have succeeded: changing the election into a referendum on Obama. But this convention is about undoing that change and shifting the attention to McCain’s record. The indispensable Ambinder notes about the Democratic holdouts:

It is MUCH harder to convince them to vote for Obama because they LIKE him. It is much easier to convince them to vote for Obama because they think McCain represents a continuation of President Bush’s policies. (Obama’s campaign has polling data suggesting that an unusually large number of pro-choice Democrats don’t know that McCain is pro-life.)

In other words, Obama can consolidate the party by turning it against an external foe. And, using this week to unite the party will lead to a sizable and lasting gain in the polls, not some ephemeral convention euphoria.

Let’s delve into some numbers. Obama hovers in the high-seventies among self-described Democrats, although the precise number ranges from poll to poll. In the past four elections, the Democratic nominee has garnered 89-90% of the vote on Election Day; treat that figure as an optimistic ceiling.

Assume Obama comes out of the convention with 85% of the party backing him. The result: a three to four point bounce in the polls. Right now, the RCP average tacks Obama with a narrow 1.6% lead; the convention bounce would widen that gap to a comfortable margin and potentially put the election away.

It’s worth recalling a point Alan Abramowitz made during the tail-end of the primaries:

The fact that Democratic identifiers now decisively outnumber Republican identifiers means that in order to win, Democrats only have to unite and turn out their own base.

And that’s the purpose of the convention.

What a Bounce!

June 20, 2008

Obama opens up a fifteen point lead in the latest Newsweek poll, buoyed by the shifting party sentiments:

55 percent of all voters call themselves Democrats or say they lean toward the party while just 36 percent call themselves Republicans or lean that way.

Yowzers!  Couple this with statistical models granting Obama an average of 343.8 electoral votes, and this election may turn into a landslide.

How Divided Are the Dems?

May 14, 2008

The Obama campaign released a memo today, both pre-spinning tonight’s blowout and attempting to dispel a series of myths. The chief canard:

MYTH 1: The Primary has left Democrats divided.

FACT: Democrats are united behind Barack Obama, even more so than Republicans are united behind McCain

May 12 Washington Post poll shows that Obama wins 81% of Democrats in a matchup against John McCain.

Indeed, more Republicans crossover to vote for Obama (15%) than do Democrats for McCain (13%).

Tonight’s exit polls revealed deep rifts in the Democratic party, at least in the mainly white, less educated, and relatively poor West Virginia.  However, some are erroneously extrapolating these divisions to the general election.

For the past four presidential elections, the Democrat defection rate has been about ten or eleven percent. According to recent polls, with Obama as the nominee, the Democratic crossover to McCain ranges from eleven to fifteen percent. And, this is in the midst of a heated primary. The number of party deserters will only fall from here, and it’s already at a good level relative to history.

Evaluating Obama Campaign’s Predictions

April 23, 2008

Back on February 6th, the Obama campaign leaked an internal spreadsheet (.xls) projecting the rest of the primary season. Since then, the predictions have proved prescient, calling races with remarkable accuracy. The campaign has correctly picked the winner of every single contest, with the one exception of Maine (projection – Clinton; actual – Obama). Beginning March 4th, the average error in calculating Obama’s share of the vote is a minuscule 1.71%. To put it into perspective: the projections, made more than one or two months before the actual voting, are more accurate than any pollster’s record.

Using the spreadsheet to examine upcoming contests, both Indiana and North Carolina should be Obama victories by respectable margins. Plugging the percentages into Jay Cost’s handy spreadsheet, Obama could net 150,000 votes out of the two states – partially offsetting Clinton’s 200,000+ gain in Pennsylvania. Though more importantly, winning both could spell an end to the nomination fight.

Note: Here’s a closer look at how the predictions stack up.

National Polls on the Democratic Side

November 6, 2007

After the Oct. 30 debate in Philadelphia, national polls across the board show a dip in Hillary Clinton’s lead.  Before the debate, CNN had: Clinton – 51%, Obama – 21%, Edwards, 15%.  After: Clinton – 44%, Obama – 25%, Edwards – 14%.  Pre-debate, Rasmussen had: Clinton – 49%, Obama – 22%, Edwards – 12%.  Post-debate: Clinton – 41%, Obama – 22%, Edwards – 13%.

Note that in the CNN poll, Obama jumped up, while Edwards actually slipped a point.  Obama’s debate strategy of (mainly) staying above the fray and letting Edwards do most of the heavy hitting seems to have worked.

UPDATE: Another poll to back up this storyline.  The latest WNBC/Marist poll, taken before and after the debate, shows Clinton much weaker in the post-debate sample, dropping from 52% to 43%.   The biggest gainer was “undecided”, moving up from 16% to 22%.

Surprising Poll Numbers

October 24, 2007

The latest Los Angeles Times/Bloomberg poll reveals an interesting tidbit about the Republican race.  One question: “Could you vote for a candidate for president who supports abortion and gay rights if you agree with him on other issues, or could you only vote for a candidate for president who opposes these issues?”  Out of the 39% of Republican primary voters who could only vote for candidates opposing these issues, 19% are supporting Giuliani in the primary.  Another surprising number: a plurality (20%) of Republicans believe Giuliani would be the best among the field at handling social issues such as abortion and gay rights.

There are some fascinating general election numbers revealing Obama’s immense cross-over appeal.  In a hypothetical Clinton vs. Giuliani matchup, Clinton gets 10% of registered Republicans.  On the other hand, Obama is supported by 23% of registered Republicans in a Obama/Giuliani contest.  Clinton vs. Thompson – the junior NY senator gets 12% of Republicans.  Obama vs. Thompson – Obama gets 22% of Republicans.  Obama also beats Clinton in Republican support against McCain and Romney by five and three points, respectively.