[Cross posted from the European Tribune where it was the third in a series on “Who is really going to Win the US Election”]
To a European observer the most striking thing about the US party nominee elections is how volatile they are. In many European elections a 5% swing is regarded as very significant when it takes place over the course of an entire campaign and opinion polls are often accurate to well within their stated +- 3% points margin of error. In the US elections, however, opinion polls published on the same day by different companies can often differ by much more than this.
In the New Hampshire primary the final polls looked like this:
Poll Obama Clinton Edwards Richardson
RCP Average 38.3 30.0 18.3 5.7
Suffolk/WHDH 39 34 15 4
American Res Group 40 31 20 4
ReutersC-Span/Zogby 42 29 17 5
Rasmussen 37 30 19 8
CNN/WMUR/UNH 39 30 16 7
Marist 36 28 22 7
CBS News 35 28 19 5
To help overcome the variation between polls I have been using the average of all available polls but even that didn’t seem to reduce the margin of error.
Iowa Final Actual New Hampshire
Polls Result Polls Result (with 96% [UPDATE} of precincts reporting)
Clinton 29 29 30 39
Obama 31 38 38 36
Edwards 26 30 18 17
Biden 5 1
Richardson 5 2 6 5
—
Huckabee 30 34 12 11
Romney 27 25 28 32
Thompson 12 13 2 1
McCain 12 13 32 37
Paul 7 10 8 8
Guilliani 6 4 9 9
In Iowa both the Obama, Edwards and Huckabee results were significantly outside the alleged 3% margin of error for a single poll, never mind the average of a whole number of them. In New Hampshire the Clinton and McCain surges were entirely unanticipated.
Pollsters will often try to explain this away by saying that any poll is only a snapshot of opinion on a given day and that a lot can change in 24 hours. If that is the case, why bother polling at all? It is hard to escape the conclusion that the polls are not really about trying to predict outcomes at all. They are there too try and influence the outcome, and to give the media something quasi scientific to talk about.
The other major factor influencing voter is of course the bandwagon or bounce effect. The Iowa result had a major impact on voter perceptions in New Hampshire and resulted in a Clinton lead of 7% in the polls prior to Iowa being turned into an Obama lead of up to 13% in polls taken after the Iowa result.
Even here the polls asking ET readers to predict the next POTUS changed dramatically after the Iowa result.
Before Before
Iowa New Hampshire
Clinton 47% 22%
Obama 11% 48%
Edwards 29% 7%
Giulliani 0% 3%
Huckabee 5% 14%
Romney 5% 0%
McCain 0% 3%
. n= 17 27
Oh what a fickle bunch you are!
I said in my first Diary on European Tribune – Who is really going to win the US Election?
CONCLUSION: Its early days and all to play for, but my money (based on looking at the trends to date) is on a Clinton Huckabee contest with Clinton winning by a narrow majority. It’s a bit like trying to predict the future by looking in the rear view mirror and we all know that big money and powerful vested interests are in the driving seat.
I made that very early and tentative prediction not because I was confident of calling the Iowa and New Hampshire results, but because Clinton had a 20% lead nationally which gave her a lot of downside resilience in the event of some early setbacks. That 20% lead in the National Polls went down to 8% after Iowa, so it was clear that another defeat in New Hampshire could have erased it altogether.
Obama and Edwards needed to to win in either Iowa or New Hampshire to have a chance, which Obama duly did in Iowa in very impressive fashion. Edwards is in now deep trouble and South Carolina is his last chance.
Giulliani’s 3% lead in the national polls has evaporated and he is now behind both Huckabee and McCain by 3 and 2% respectively even before the New Hampshire vote. Huckabee and McCain needed to win either Iowa or New Hampshire, and having done so they will now move significantly ahead. Romney and Giulliani now badly need to win in Michigan and Nevada to stay in the race.
I don’t want to give people here the heebee jeebees as we say in Ireland, but the national head to head polls indicate that a Clinton McCain national contest is the one combination of candidates that the Republicans would actually win!
——
Many years ago I did some market research for a rather famous brand of Irish beer. The company was interested in finding out why there was so much variability in their share of the total beer sales from pub to pub. So we analyse those factors which influed drinkers in their choice of Brand AT THE POINT OF ENTERING A PUB AND PLACING THEIR ORDER.
The results where very very interesting and had a major impact on Company strategy. The most important factors where:
1. Is it a traditional or a modern pub in terms of decor (traditional favoured Guinness, modern favoured lager)
2. What were most people in the pub drinking (no experienced draught beer drinker wants to be the first to drink a particular brand on a particular day, becuse that beer may have been sitting in the pipeline all night and may have oxidised slightly or been stored in warmer conditions.
3. Hot weather favoured lagers
4. The most important factor was if the Barman made some kind of prompt – e.g. its a great day for a Guinness, or Guinness would go great with that meal you just ordered etc.
The research led directly to the company to setting up “The Irish Pub” marketing strategy whereby bar owners where offered a complete service in furnishing their premises with themed traditional Irish furniture, knick knacks and other visual cues.
It also led to Guinness being served at colder temperatures (particularly in summer), a huge emphasis on point of sale materials, advertising on Menus, and the training of bar staff to assist customers in their choice of drink.
All of this goes a long way to explain the huge importance of having your supporters and branding materials strategically located outside polling stations. It also explains the huge importance of media and opinion poll created perceptions of “surges” for particular candidates.
In my (long distance and uniformed view) a critical factor may have been Hilary’s alleged “breakdown” and its treatment by the hostile media. To most people she was just showing she was human and it was the first time many might have warmed to her. But that is just my guess.
—-
I suppose the bigger point I am trying to make here is that if the US electorate IS so volatile, and so many people make up their minds at the last moment, then it is critical, from an analytical and campaigning point of view, to understand what factors influence those final decision making moments the most.
There is a pseudo scientific mythology around the polls and my thesis is that they are as much about influencing as predicting voter behaviour. The margin of error is quoted as being +/- 3.3% no less. To one decimal place!!! With no mention of the fact that such margins of error only apply to yes/no questions with an expected 50/50 split. When was the last time those conditions were valid for a poll? If the poll is that close the margin of error makes it almost useless in any case.
The sample sizes are also generally too small. 1000 is the accepted minimum in Europe for any reasonably large and diverse population. And finally the problems with telephone sampling introducing systematic non random biases are well known.
——
OK Class, one more time…
