Incumbents often do better than final polls

[DL]

I know there has been discussion in the past of this "hypothesis" that people who are undecided to the end break for the challenger or that the incumbent needs to be over 50% etc... I work in public opinion research and I have been looking at a lot of elections across the western world - and the one thing that stands out to me is how over and over and over again...the incumbent candidate or party seems to do a bit better than the final polls indicate.

I'm Canadian so bear with me while i go through some Canadian examples:

1.In the '04 and '06 federal elections the incumbent Liberals surprised even themselves by getting about 3% more on election day than the average of the final polls (its true that they lost to the Conservatives in 2006 but the polls projected a double digit Conservative lead and they only won by 6%). Then in the '08 and '11 federal elections the shoe was on the other foot and the Conservatives were the incumbent and it was there turn to get about 3% more on e-day than the final polls predicted.

2. This past year we have had a series of provincial election where the incumbent was projected to get crushed at the polls and in the case of Alberta the ruling PCs were projected to lose by about 7% - instead they won by 10%. In Quebec, some polls had the incumbent Liberal party falling into third place and facing annhilation, instead they lost the popular vote by less than 1% and and got about 4% more than any poll projected.

3. In the UK election of 2010, Labour was supposed to get demolished and get wiped off the map and likely come in third behind the LibDems - instead they got 30% (3% more than any poll projected) and held the Tories to a minority. Let's not forget UK 1992 when all the polls said Labour would win and the Tories under Major won another majority.

4. In France all the polls said Sarkozy would lose and he did BUT the polls also said he would lose by 6 or 7 points and instead he lost by 3.5 points.

5. In Germany in 2005, Schroeder and the SPD was supposed to lose by 7% and the CDU was supposed to win  easily - instead the popular was almost dead even and Merkel got in by the skin of her teeth.

6. In the US the best examples are 2000 where Gore was the quasi-incumbent as VP. I still remember how all the polls pointed to Bush winning by 4-6 points and any poll showing it dead even was dismissed as an outlier (as we all know Gore ended up actually winning the popular vote!). In 2004 many polls had it dead even between Bush and Kerry, Bush ended up winning by 3%.

I am not saying that Obama as the incumbent will necessarily get an "incumbent ballot box bonus" on Nov. 6 but I am just saying that more often then not - to the extent that polls are "non-predictive" they usually underestimate support for incumbents.

There are many hypotheses as to WHY polls often underestimate incumbent support. I have two: 1) I think that a lot of "late deciders" who are low information ":marginal voters" will often "stick with the devil they know" 2) In this age of very low response rates on surveys, I think that the people who are more likely to take part in a survey are the people who are angrier and more riled up and have stronger emotions - in most cases that describes people who are supporting the whoever is the challenger.

Anyways, juyst my two cents worth

[DL]

Bill Clinton did worse than expected in 1996, right?

I said there are always exceptions...another pattern i have noticed is that independent of who is the incumbent and who is the challenger - anytime the final polls suggest a massive landslide for anyone (as the polls suggested in 1996) - the results are almost always a bit closer (probably because people go back to their natural political home at the last minute - so in 1996 some Republicans who flirted with voting for Clinton in the clutch they stayed loyal to the GOP and so Dole didn't get quite as totally blown out of the water as some expected)

[DL]

Its true there are examples on each side - but in recent years - there has been a very consistent pattern of polls underestimating support for incumbents. To find examples of polls overestimating support for incumbents you typically have to go back 15 years or more. (back in the days when response rates to polls were as high as 30% and not the 2 or 3% it is now)

[DL]

2000 is a strange case because it was revealed right before the election that Bush got a DUI. That may have influenced a lot of last-minute changes who people voted for.

That may have been one factor, but another thing that happened in 2000 was that the Gore GOTV was very strong and very underestimated and polls of "likely voters" tended to exclude a lot of people who ended up voting for Gore