« Bush's Hispanic Support Headed Downwards | Main | Unmarried Women Now Key Element of Democratic Base »

The Time-for-Change Model and the 2004 Presidential Election: A Post-Mortem and a Look Ahead

By Alan Abramowitz

Based on George Bush’s net approval rating of -1 percent in the final Gallup Poll in June, a 3.7 percent growth rate for real GDP during the first two quarters of 2004, and the fact that the Republican Party had controlled the White House for only one term, the time-for-change model predicted that Mr. Bush would win reelection with 53.7 percent of the major party vote. This forecast was made in late July, before the Republican National Convention and more than three months before Election Day.

In fact, George Bush was reelected with 51.4 percent of the major party vote. The forecast error of 2.3 percentage points was slightly larger than the previous average out-of-sample error of 1.7 percentage points for the time-for-change model. Nevertheless, the model maintained its record of correctly predicting the winner of the popular vote in every presidential election since 1988 and the 2.3 percent forecast error only slightly exceeded the average error of 1.9 percent for 14 national polls that were released in the last four days before the election.

George Bush’s net approval rating (approval – disapproval) of -1 percent in late June was well below the average net approval rating of +11.7 percent for all incumbent presidents who have run for reelection since World War II. In addition, the 3.7 percent growth rate of the U.S. economy during the first half of 2004 was below the average growth rate of 4.5 percent for all presidential election years since World War II. So why did George Bush defeat John Kerry and win a second term in the White House?

George Bush did not win because he ran a brilliant campaign, because John Kerry ran a poor campaign, or because millions of evangelicals turned out to express their opposition to abortion, gay marriage, and Janet Jackson’s wardrobe malfunction. He won because he was an incumbent whose party had controlled the White House for only one term. In the past century there have been 11 elections involving such first-term incumbent presidents. Ten of these 11 incumbents were reelected. George W. Bush’s 51.4 percent share of the major party vote was actually the second worst showing for a first-term incumbent in the past century: only Jimmy Carter in 1980 received a smaller share of the major party vote than George Bush.

The good news for Democrats is that 2008, unlike 2004, will be a time-for-change election—one in which the president’s party has controlled the White House for two or more terms. There have been 16 such elections in the past century, with the incumbent party winning 7 times and losing 9 times. Since World War II, the track record of the incumbent party is even worse: 2 wins and 6 losses.

Reestimating the time-for-change model based on the results of all presidential elections since World War II, we obtain the following estimates:

V = 50.3 + .81*GDP + .113*NETAPP – 4.7*TFC,

where V is the predicted share of the major party vote for the incumbent party, GDP is the growth rate of real gross domestic product during the first two quarters of the year, NETAPP is the incumbent president’s net approval rating in the final Gallup Poll in June, and TFC is the time-for-change dummy variable. TFC takes on the value of 0 if the president’s party has controlled the White House for one term and 1 if the president’s party has controlled the White House for two or more terms.

The estimated coefficient of -4.7 for the time-for-change variable means that once a party has controlled the White House for 8 years or longer, it is penalized by almost five percentage points. This obviously makes it much more difficult for the incumbent party to win another term. For example, if real GDP grows at an annual rate of 3.7 percent during the first two quarters of 2008, as it did during the first two quarters of 2004, and if President Bush’s net approval rating in late June of 2008 stands at -1, as it did in late June of 2004, the Republican presidential candidate would be predicted to receive only 48.5 percent of the major party vote in the 2008 presidential election.

Forthcoming in PS: Political Science and Politics

Note: You can read the original article to which this piece is a followup, as well as the entire pre-election symposium on election forecasting, on the PS website.


Doe this model work for 2000? Were Clinton's approval ratings that low? Was the economy doing well? What did the model predict?

I just want to take this opportunity to complement the operators of Donkey Rising for foing what they can to encourage the political opposition during what has to be a tough time. I'm an independent, but I've always wondered if the tide has been turning on the Right Wing in this country since the collapse of Communism in the late 1980's (9-11 and Bush-Rove or no 9-11 and Bush-Rove). Having said that, from looking at some of the comments your articles have inspired are more discouraging than anything any Republican might write. You people are almost as bad as Birchers. As for Donkey Rising, keep up the good work and hang in there. You may actually get your people to cheer up.

This is all very interesting and probably encouraging for most Democrats but for the more pragmatic, like me, it doesn't ring true.

Bush may not allow elections by 2008, the Constitution may be amended to allow "Arnie" to run, it may be amended to give Bush a third term, the GOP will have changed so much of the political and social landscape by 2008 that even if the Democrats win they will have a hard time making any really progressive changes and may not be able to stop the regression.

In addition, the Republicans (as a troll on the Dean Web site kept reminding us) do own the voting machines. I don't believe we can ever have an honest election in this country again. I felt this after 2000 and 2004 just strenghtened my belief that the Republican interests own and control the voting process. They are never going to allow Democrats or any other party to win again. There isn't much we can do about it. We sent out thousands of observers" this time and they still disenfranchised thousands of voters and they'll continue to do so.

We just do not, as things stand today, have any capability of getting a fair election and the media, certainly, are not going to admit this and anger their GOP masters.

What do the authors of this suggest? 2008 may be the year for change, but if the voting is controlled by the Rebublicans the only change we may see is from Dubya to Arnie or Jeb. We won't see any Democrat win.

Besides, by 2008 the GOP will have redistricted every state in the nation in such a way that no Democrat can win in any district. They will have recalled as many other Democrats as they possibly can and gotten Republicans elected in their stead so how is the Democratic Party going to win even if the people want them?

The answer is they are not. As long as Republicans own the voting machines and the voting process they will control the elections and the country.

There is a big error or at least a very misleading statement in this post: it is either not accurate or extremely misleading to say that George W. Bush’s 51.4 percent share of the major party vote was actually the second worst showing for a first-term incumbent in the past century. Bill Clinton received a smaller percentage. William Howard Taft received a smaller percentage. Herbert Hoover received a smaller percentage. Now, Taft and Hoover had succeed Republican presidents, so their situation was not exactly analogous. But Clinton had succeeded a Republican president, so his situation was exactly analogous. Now, it may be that Alan is factoring out Ross Perot's share of the vote, but did he factor out John Anderson's? Or even those of the minor party candidates this year? It doesn't look like it, so the statement that Bush's percentage was the second worst is either simply wrong or extremely misleading.

Another error: 10 out of 11 incumbents reelected in the last century? Just in the past twenty-five years, we had defeats of both GHW Bush and Jimmy Carter.

This is not an error. He says 10 out of 11 "first-time incumbents". A first-time incumbent is " an incumbent whose party had controlled the White House for only one term". Carter was a first-time incumbent. GHW Bush was not, because he was preceded by a Republican president.

As a physician and, therefore, a small businessman, it seems to me that doctors and small businessmen are a natural constituency for Democrats - we are getting squeezed by inadequately regulated corporations (health insurance companies that refuse to pay when we bill them, etc)

Gov Schweitzer did a wonderful job in Montana to court small business...where are the Democrats - why don't they make their presence known so that by 2006, they have laid the necessary groundwork to win?

Also, Mr Texeira, are there recent opinion polls of physicians re dems vs repubs? I understand in Rochester, Minn, the polls showed a large shift of support towards Dems by MDs.

This kind of "gas molecules in a box" analysis is simply misleading. It treats the electorate as some kind of an undiffentiated collection of clones that like simple physical systems, "if you do this and this, then it will do that every time". It's simply not that simple.

Without data that accounts for variance in the electoral demograhics ACROSS all the elections cited, as in: regional voting results, age group voting results, gender voting results, race voting results and socio-economic class voting results, it's an irrational conclusion. All these catagories' results in each election would have to show a proportional consistency when compared to the total results in order to even begin to support a causal link between the Time-for-Change Model and the elections' outcome.

Therefore, this is not science. It's not even social science. It smacks of the kind of crank voodoo stuff that economists have been touting as science for decades and getting away with aka: perfect rationality etc. Every college physical science major learns in their first year that just because event A occurs and then B occurs, even if relatively freguently within a SMALL sample, that does not in itself indicate a causal link. Period. And Abramowitz's claim with finality that the outcome is due to his INTERPRETATION and not any others, such the as evangelical vote, is merely his opinion and no more credible than anyone else's ABSENT OF MORE DATA.