The time has come to start matching the reality to the dream. What are your predictions for the US Elections and how do these compare to the professional pollsters and pundits – and most importantly, how do these compare to the actual results when they are eventually declared?
Realclearpolitics is calling it 52/44% popular vote resulting in an electoral college split of 338/200 to Obama. Pollster.com has an almost identical vote split prediction of 52/44% but predicts a 364/174 electoral college vote (EV) split. www.fivethirtyeight.com predicts a more conservative 52/46% popular vote split and an Obama Electoral College advantage of 346/192.
So why do I think all these highly professional pollsters have got it wrong?
Firstly, these predictions are all the result of statistical or arithmetic projections of polls we know to be flawed in many ways. These flaws include:
- Under-representation of cell-phone only voters – this is known to skew opinion polls by as much as 2-4% against Obama.
- Early voting trends have been heavily skewed in favour of Democrats – a vote in the bag is worth a lot more than a “likely voter” who may be put off by long voting lines or other priorities on voting day.
- Obama’s superior ground game – and the much greater enthusiasm/determination being exhibited by his supporters – obvious in the long lines that have formed at early voting and now election day voting places. At the end of the day – even with an unprecedentedly high turnout – perhaps some 30% of the electorate will fail to vote – and these are much more likely to be over-representative of the more lukewarm and disinterested voters or those who do not have the support of a GOTV operation. Its much easier to state a preference to an online or telephone pollster than it is to actually go out and vote.
- The “likely voter” models some pollsters use to screen their results are still often based on the historically lower turnouts amongst minority and youth voters – something which does not appear to be reflective of the early voting patterns to date.
I know there are other factors which may work the other way, chiefly the The Bradley effect, and the Shy Tory effect which postulates that people may be uncomfortable revealing to pollsters their support for a very unpopular (in this case Republican) regime. In the privacy of the polling both, some anxious and insecure voters may also be more inclined to go with what they perceive to be the safer, more known quantity, than with the alleged greater uncertainty associated with a relatively inexperienced and possibly more radical younger President – who also happens to be the first serious African-American contender for the Office.
However is it not also possible that we might be seeing a “Shy Obama Supporter effect” amongst voters who do not wish to attract the ire of their traditionally conservative friends, relatives and neighbours? These are “soft” questions, which the polls, by definition, cannot answer, because they all posit some form of voter dissonance between what they say they will do, and what they actually end up doing. These are questions which a commentator from afar such as I simply cannot answer at this stage except to say that I remain a skeptic that they will end up having a significant aggregate effect either way.
Many pollsters and commentators much closer to the action and more experienced than I seem to be hedging their bets on these factors, and purport to see a “tightening” in the polls as polling day approached. This tightening is certainly visible in the latest pollster.com table:
What is striking is that the 9 closest states from Virginia down to Georgia have all trended towards McCain in the last day, and all except the last 3 have also been doing so over the past week. However the national trend has been towards Obama over the same periods and so what we are seeing here is specific to those states, and not to the US as a whole. If there was significantly heightened anxiety about an Obama Presidency, would we not see it reflected in the national polls?
Far more likely we are seeing the effects of the heightened campaign and media focus on those battleground states and the distortions noted in paras. 1-4 above. In other words I will stick my neck out and project a 10%+ margin of victory for Obama (based on a c. 3% net anti-Obama polling bias for reasons 1-4 above) and a 396/142 Electoral College win for Obama based on his winning all the states in the table above right down as far as the last “toss-up” state of Georgia.
Of course for statistical noise reasons alone, it is far more likely that Obama will lose at least one of those “toss-up” states – Montana and North Dakota seem the most likely because there have been far less polls there and so the confidence levels associated with any prediction has to be a lot less. But equally,on that basis, Obama might even pull off a really stunning upset – such as in McCain’s home state of Arizona – if he really does achieve a 10%+ lead in the national popular vote.
So at the risk of looking very foolish I will go with my instincts and predict a 10%+ margin of Victory for Obama with an absolute and extraordinary 396/142 blow out margin in the Electoral college. At that rate the Democrats also stand at least a 50:50 chance of securing a 60:40 filibuster proof margin in the Senate and we could be seeing the dawning of a new age in American and World Politics.
Obama already has the 270 EVs he needs if he wins all the states (down to Pennsylvania) classified as strong Dem in the table above, and 291 if he takes the “Lean Dem” states of Nevada and Virginia as well. Virginia should be one of the earlier states to declare a result (provided the long voting lines don’t result in some delayed vote counts). If Obama wins Virginia he should be safe. Otherwise we could be in for a long night and many days in court before this thing is finally resolved.
