So why were the New Hampshire polls wrong?
With 96% of precincts reporting, we have the following results:
Republicans
McCain 37.2%
Romney 31.6%
Huckabee 11.2%
Giuliani 8.6%
Paul 7.6%
Democrats
Clinton 39.2%
Obama 36.5%
Edwards 16.9%
Now what did the polls say? Since Iowa, there were six main polls. They were:
The sample sizes varied but all were at least 500 which is a margin of error of 4.5%.
Using the 500 sample size, the probability for each poll that McCain was actually ahead of Romney is in order 12%, 98%, 99.5%, 61%, 93% and 86%. So five out of six were predicting McCain and the sixth was close enough that it was saying there is a 10% chance McCain was ahead.
Now on the Democratic side, the six results were saying the chance of Clinton being ahead was 10%, 0.8%, 0.02%, 3%, 0.7%, 1%. So one out of six had some doubt about it, and five out of six were emphatic that Obama was ahead.
Overall the average of the six polls vs the actual is:
So the polls were not too bad on the Republican side. McCain and Romney both did better than the polls said but the margin between them was within 2%.
On the Democratic side, they got fairly close with Obama. The average was 39% and he got 37%. Where they were far out was Clinton’s support at 39%. The polls averaged 30% with a range from 28% to 34%.
So why did they get it so wrong? The fact every company got it wrong suggests we can ignore methodological differences between the companies.
One possibility is that those who voted were different to those who answered the poll. Now as you see in this Suffolk poll, they ask how likely people are to vote, and 96% were very likely so that mitigates that problem. But if for example Clinton had a much better get out the vote machine, then that could be part of it.
Another possibility is the independents voted Republican rather than Democrat. You see Clinton led in some polls amongst Democrats only, but Obama had a massive lead amongst independents. However 43% of those who voted in the Democrat primary were independents so again that is probably not it.
A lot of people are looking at this. Mystery Pollsters quotes the ABC Polling Director:
There will be a serious, critical look at the final pre-election polls in the Democratic presidential primary in New Hampshire; that is essential. It is simply unprecedented for so many polls to have been so wrong. We need to know why.
Next we look at the CNN exit poll. It shows Obama beating Clinton by 11% amongst men – 40% to 29%. However amongst women voters Clinton is 12% ahead – 46% to 34%.Mystery Pollster looks at the difference between exit polls and the pre-election polls and finds that support for Clinton was 17% higher amongst women in the exit poll than before. So it may be as simple as women flirting with Obama and saying they will support him, but when they get to the booth, deciding to go with Clinton.
I could joke about women changing their mind, but that does seem to be the case. Support for Clinton amongst men was only 5% higher in the exit poll than the pre-election polls. That would not have been enough to win it for her. It was the massive swing back of women voters it seems.