Saturday, August 2, 2008

So, Who's Going to Win This Race? The Forecasts are Starting to Come In

With second quarter economic data now available, many of the typical political science presidential election forecasts are beginning to emerge. And this week seemed to be the time for their unveiling. Thomas Edsall, who is writing for The Huffington Post now, had a great run down of some of them earlier in the week and since then Seth Masket and Robert Erikson and Christopher Wlezien have released their numbers. Those add to the back and forth between Abramowitz, et al. and Campbell that I discussed in the electoral college post last Sunday.

Let's look at the numbers here and some of the impressions we can draw from them. [As a side note, I should add that I tried to track down some of the sources from Edsall's piece (Some forecasts have links while others do not.) and stumbled upon a special issue of the International Journal of Forecasting* focused on presidential election forecasts. Now, while the issue does not contain the actual forecasts, it does provide more than adequate insight into the models and debates within the forecasting area of the literature.]

Abramowitz, Mann and Sabato: Comfortable win for Obama (a Democratic environment, Democrats' party ID advantage, and recent state and national polls)

Campbell: Close election (Bush approval does not translate to McCain necessarily, McCain was the best-positioned of the Republican candidates to for the general election, open seat elections are close)

Erikson and Wlezien: Obama = 53% of the popular vote (based on state trial-heat polls and leading economic indicators)

Geer: Close election (electorate's comfort zone with candidate's foreign policy stances post-9-11, McCain is a good candidate for the GOP, the last two elections have been close)

Lewis-Beck and Tien: Obama = 50.6% of the popular vote (based on jobs growth and growth in GNP and including a correction for the race factor in the contest)

Masket
: McCain = 47.7% of the popular vote (based on 2nd quarter real per capita disposable income)

Norpoth
: Obama = 50.1% of the popular vote (based on support of major party candidates in the primaries among other factors)

I withheld Sandy Maisel's prediction because it doesn't neatly fall into either category, landslide or toss up. He sees an Obama win unless the Illinois senator does something to lose it. With that said, we have seven forecasts; 4 close calls (two unambiguously finding Obama winning) and 3 -- I hesitate to call them landslides -- more comfortable victories for Obama. Five of these forecasts see Obama wins, but find different margins based on the underlying factors included in their models.

I should also mention that there were two forecasting models discussed at this past spring's Western Political Science Association conference: De Sart and Holbrook and Gurian and Cann. De Sart's forecasting web page hasn't yet updated for the 2008 election, but the forecast is based on both state and national polls (While we're citing, I should go ahead and include Thomas Holbrook's blog (again) as well.). The paper that Paul Gurian -- an occasional FHQ contributor -- presented at WPSA wasn't intended as a forecast, but he and Damon do have a forecasting component to it. They are still waiting on another couple of factors to add in to complete the forecast portion, though.

H/T to The Monkey Cage for the head's up on the Edsall article and Erikson and Wlezien's latest forecast.

----
* The pdf files of those articles are gated (for purchase), but the abstracts should be available to those who, at the very least, want to check those out. Simply click on the title of the article to get the abstract.


Recent Posts:
VP Announcement Timing

5% of Democrats Say They'll Vote for McCain

The Electoral College Map (7/30/08)

7 comments:

Anonymous said...

fivethirtyeight.com currently predicts Obama getting 49.7% of the vote, and McCain getting 47% of the vote.

(I believe it gets this data through averaging the results of 10,000 test simulations)

It gives Obama an average of 303.7 electoral votes, with McCain having an average of 234.3 electoral votes.

It predicts that Obama has a 67.7% chance of winning.

Anonymous said...

Thanks for including 538's results, Glenn. I thought about including it and then forgot. Not much of an excuse. Nate has actually been advised by Erikson (a family friend apparently) on some of his methodology.

FiveThirtyEight is cited often enough around here, but for those interested, there is a link to the site in the right hand side bar.

Robert said...

Again, as Paul has said, most of these models assume that the campaign doesn't matter. I still have problem with that concept. I heard Charles Cook on NPR this morning indicating that he thinks it will be an extremely close election that is impossible to call at this point.

Robert said...

Did you see

http://politicalticker.blogs.cnn.com
/2008/08/02/obama-mccain-camp-cynical
-not-racist/

There was a comment from Obama saying "They're very good at negative campaigning. They're not so good at governing."

I think we have heard that before, and I think it will be heard again. I think if the Republicans keep up the complaining about Obama playing the race and age cards, he will play the age card early and often, particularly in accusing McCain of living in the past.

Unknown said...

Do you remember the old computer game President Elect? It was a remarkable simulation for getting across the interaction between the economy, incumbency, political views, the candidate's abilities, and shrewd campaigning. Random gaffes and debate results were included too. If the macro factors were too far against you, though, there was nothing you could do. But if the race was finely balanced (say, 1960, or presumably 2000), then campaign decisions and the candidate's skills could make the difference.

In the sense of actually teaching something, I think it's amongst the most effective educational computer games ever made.

Jack said...

Tried to play the game but when I attempt to begin a simulation, three numbers show up on the screen, and apparently there's supposed to be some question that comes up that you need the manual to answer.

Anonymous said...

Here's that link from Rob.