The nonpuzzle of the close election polls
August 1st, 2008, by Andrew
The economy is going badly, Bush is unpopular, so why is Obama up only slightly in the polls? This is the question that is (currently) puzzling the political world. A quick answer: the economy isn’t going so badly by historical standards, and Bush’s popularity isn’t so relevant given that he’s not running for reelection. Forecasting models based on past elections predict Obama to get something like 53% of the two-party vote–but these forecasts aren’t perfect; they have a margin of error of a few percentage points. (This is not the margin of error of the polls, arising from sampling variation; rather, it’s the uncertainty reflected by the imperfections of the forecasting model as applied retrospectively to past elections.) In short: macro conditions for the Republicans are not so bad as all that: Obama is a legitimate favorite but there’s no reason to expect that there would be a landslide. Things are going about how one might suspect based on historical patterns of the economy, incumbency, and presidential elections.
To think about it another way, consider this graph adapted from Doug Hibbs of incumbent party’s vote share and economic growth, for all presidential elections since 1952.
As you can see, the incumbent party sometimes loses but they never have gotten really slaughtered. In periods of low economic growth, the incumbent party can lose, but a 53-47 margin would be typical; you wouldn’t expect the challenger to get much more than that. Such things can happen (see, for example, Eisenhower’s performance against Stevenson in 1952) but it wouldn’t be expected.
Similar Posts:
- Doug Hibbs’s latest forecast
- Predicting the election outcome months ahead of time: discussion and link to revised paper with Kari Lock
- Evidence that the state of the economy is crucial to voters
- Erikson and Wlezien forecast Obama will get about 53% of the vote with approximately 2/3 chance of winning
- Close elections really have become more common in recent decades
Entry Filed under: Elections
22 Comments
Add your own1. Yair | August 1st, 2008 at 11:45 am
Great graph! I take it from the last post that this displays average economic growth over the past administration. Have you looked at whether this changes if you use economic growth over the past year (if voters have short-term memory), or even using something like Erikson, Stimson, and MacKuen’s economic mood (if voters aren’t actually monitoring economic growth closely)?
2. Andrew | August 1st, 2008 at 12:47 pm
Yair,
The numbers are from Doug Hibbs and they’re some sort of weighted average that favors the last year of the administration (which is why Carter looks so bad in 1980 and Reagan looks so good in 1984, he’s recovering from the 1982 recession). Larry Bartels talks some about this issue; see here. and here.
3. A. Nonnimous | August 1st, 2008 at 2:02 pm
Economy going badly? You must be employed in the financial industry and look no further than your neighbor for confirmation (bias) of your opinion. Check the data: the economy is fine.
4. Lane | August 1st, 2008 at 3:40 pm
Andrew,
This is helpful. But I think the worry stems in part from the fact that the gap between the candidates typically closes gradually after the conventions. If it’s already small now….
Then again, isn’t it pretty well-established that polling at this stage provides very little information about who’ll end up winning? Here’s a related question that perhaps you or the others at “The Monkey Cage” could answer: What’s the point at which polls become pretty reliable at discerning the eventual winner? I’d suspect a few weeks after the last of the conventions, but that’s just a guess.
Lane
5. Jamie Chandler | August 1st, 2008 at 5:58 pm
I guess the next question is whether Obama’s huge money raising and spending lead is having only a small effect compared to McCain. Obama has deployed resources across many ore states than McCain (50 state strategy vs swng state). Is it possible that it costs Obama more to achieve the same position in the race as McCain?
6. Midday August 1 2008 &laq&hellip | August 1st, 2008 at 6:56 pm
[...] Why isn’t Obama pulling away from McCain? Actually the economic numbers suggest a modest Obama victory, which is just [...]
7. How Bad Are the Fundament&hellip | August 1st, 2008 at 7:45 pm
[...] science junkies (at least check out the working paper (pdf) that turned into the book) has a blog post up noting that historical economic trends would presage Obama victory of something in the [...]
8. Science Applied to Politi&hellip | August 1st, 2008 at 8:10 pm
[...] see. Chart courtesy of Red State Blue State, Rich State Poor State [via The Monkey [...]
9. Rich C | August 1st, 2008 at 8:11 pm
Just to be clear, Hibbs is using real disposable personal income growth, not GDP growth, in his model. Since personal income growth (certainly at the median) has tended to trail GDP growth significantly since 2001, this suggests that 53% should not be viewed as the upper bound on Obama’s performance this year.
10. Bob Hawkins | August 1st, 2008 at 9:05 pm
You’re missing the point. It’s not about the vote margin since no one knows what the vote margin will be.
The Democratic candidate is typically up by 15% at this point. Why is he only up 3% this time?
11. Andrew | August 2nd, 2008 at 12:58 am
Lane:
Sometimes the polls keep moving even after both conventions, such as in 1976; see here for some historical graphs. Oddly enough, the generic congressional-preference polls don’t do bad in off-year elections; see here.
Bob:
The Democratic candidate is not always up 15 points at this point. Bill Clinton, Mike Dukakis, and Jimmy Carter (in 1976) were, but others weren’t. See the above link.
Jamie:
I don’t really have any particular insight into 2008; my main point was that, from historical evidence, you wouldn’t expect a landslide. I’m certainly willing to believe that Obama might be doing a few percentage points less than he “should” based on fundamentals, and I’m assuming that the efforts of both campaigns will be moving voters toward their fundamental preferences between now and November.
12. Finance Guy | August 2nd, 2008 at 7:44 am
Do this: Check the direction of unemployment in the second quarter of every election year since 1948. (The direction is determined by subtracting the March rate from the June rate.)
You will find that in every year other than 1956, when the unemployment rate rose in the second quarter, the incumbent party’s presidential candidate lost. When there was no change, the incumbent party’s candidate lost in a very close election. When the unemployment rate fell in the second quarter, the incumbent party’s candidate won.
The 1956 exception can probably be explained by the fact that the rise in the second quarter that year was slight and was quickly undone later in the year. In other years, rises and falls in the second quarter rate were part of longer trends in the same direction.
Why use the second quarter? On the theory that voters won’t base their behavior on the third quarter because the impact of any change hasn’t had time to sink in.
By the way, the U.S. unemployment rate started the second quarter of 2008 at 5% or 5.1% and ended it at 5.5%. By historical standards, that’s a significant change. Barack Obama is going to be the next president. If the election is close, it will be because he’s black.
13. Kevin T. Keith | August 2nd, 2008 at 4:39 pm
Hibbs is using real disposable personal income growth
Where does that stand currently, using Hibbs’s method?
they’re some sort of weighted average that favors the last year of the administration (which is why Carter looks so bad in 1980 and Reagan looks so good in 1984
I’m sure you’re right about the trends in those years, but, actually, they pull both Carter and Reagan closer to the average on this graph, not away from it.
There are only 4 outliers on this graph: Clinton ‘96, Gore 2000, Stevenson ‘52, and Humphrey ‘68. Everyone else, including Carter and Reagan, is almost touching a perfect linear trend from lower left to upper right. (In fact, if you superimpose the trend line on the graph and note how far off the outliers are, not from the 50% mark but from their expected performance given the income growth rate at that time, you see that Clinton and Gore fell above and below the trendline, respectively, by about 5% each, while Stevenson and Humphrey got absolutely creamed.)
But if Hibbs’s method makes Carter’s economic performance look unfairly bad, then the “real” figure would place Carter higher on the income growth axis, but no further right on the popular vote axis (which is an historical given); that would put him above and to the left of the trendline, indicating much worse performance than expected, while as it stands he is exactly on that line. Conversely, if economic growth under Reagan was actually less than it appears, that would push him down the economic scale, closer to the trendline (i.e., better, in his case), while as it stands he’s a bit to the left of it.
14. Michael | August 2nd, 2008 at 5:55 pm
“The Democratic candidate is typically up by 15% at this point. Why is he only up 3% this time?”
I keep seeeing this thrown around, but it really isn’t true. The week of 29 Jul 04, Kerry was up by 2 pts. Bush beat him by 2%
http://www.rasmussenreports.com/public_content/politics/election_2004/week_by_week_numbers
On July 27, 2000, Bush led Gore by 11 points, Gore beat him by .1% in that popular vote.
http://archives.cnn.com/2000/ALLPOLITICS/stories/07/27/cnn.poll/index.html
The only Dem candidate with a massive lead at this point that went on to lose like you suggest is Dukakis, 1 isolated event 20 years ago, not a trend to fall back on.
In June of 1980, Carter was losing to Reagan 40-38.
In June of 1992, Bush was beating Clinton by 1.
http://dyn.politico.com/printstory.cfm?uuid=ED60C36B-3048-5C12-004D1E289E0C84F4
Welcome to the fact based world…
15. Rasmus | August 2nd, 2008 at 6:30 pm
Says:August 1st, 2008 at 9:05 pm
You’re missing the point. It’s not about the vote margin since no one knows what the vote margin will be.
The Democratic candidate is typically up by 15% at this point. Why is he only up 3% this time?
Nonsense. Kerry was never up by 15 points, 5 points at the HEIGHT of his convention bounce. Gore never had a significant lead over longer time. You just cite Dukakis, but Obama is no Dukakis and we´re not stuck in 1988.
Obama is certainly not a “generic democrat” either, and it´s just stupid to compare different elections. Why? They are different. They follow different schemes.
That´s it. He´s up by 3-4 points, that may translate into a squeaker or a landslide- personally I think it will be something in between. 310-320 EV or so. That´s what he´d get if the election was hold today (unless he has bad luck with the Toss-Up states Florida, Montana, Nevada and Virginia).
And the most likely thing is that nothing will change. Momentum is a lie, especially in a race with a boring candidate like McCain.
16. Structured politics &laqu&hellip | August 2nd, 2008 at 8:08 pm
[...] Via Ezra Klein via Andrew Gelman, check this out: Check the R2 on this [...]
17. Alex | August 3rd, 2008 at 1:22 pm
Also, when an election is based on the electoral college, why do pundits focus on the only stat that seems close. There is more than enough data to do analysis based on the electoral college. Please start doing some. This popular vote data is irrelevant. Why isn’t anyone talking about how Obama is up 60-40 in the electoral college? The states are terribly divided right now. Race is trumping the economy, and there is no precedent to help explain that.
18. Peter Principle | August 3rd, 2008 at 11:37 pm
Most of the outliers in the chart are pretty easy to explain — 1952 and 1968, unpopular wars; 1996, a significant third party candidate who took more from Dole than Clinton.
The hard ones to explain are 2000 (why did Gore do so poorly relative to income growth?) and 2004 (how did Bush do so well despite an unpopular war?)
19. The nonpuzzle of the clos&hellip | August 4th, 2008 at 12:12 pm
[...] the good folks at Red State, Blue State, Rich State, Poor State Blog, which is sure to elicit some discussion over burgers at Otto’s this afternoon: The economy [...]
20. Astor | August 4th, 2008 at 4:41 pm
Predictors of past presidential campaigns only work until they are proven wrong. I remember USA Today printing a list of predictors in 2000 showing economic trends which indicated that Gore would win in a landslide. We all know how that ended.
I would say nothing can predict this election. Yes, the economy is not great, but we are not in a recession. The unemployment rate is at historic low levels despite an rencet uptick. And we never had a black candidate nominated for either party. Noone knows how this will play out.
21. Judy Fritts | August 4th, 2008 at 11:21 pm
Liked your graph. I think Obama needs to choose Biden for his VP. He (with all due respect) has a mouth and is not afraid to use it! He got rid of the former mayor of New York with one sentence! “A noun, a verb, and 9/11.” Priceless! I think a few more one-liners like that and we will see the polls move upward for Obama. What do you think?
22. Boîte noire – Les &hellip | October 28th, 2008 at 4:02 am
[...] Gelman nous dit le nécessaire [...]
Leave a Comment
Some HTML allowed:
<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>
Trackback this post | Subscribe to the comments via RSS Feed