This note was originally published November 05, 2012 at 11:57 in Macro
As usual, we had a vigorous discussion in a recent research meeting on the state of the election. I provided the update to the team that all of the indicators appear to be moving more towards President Obama over the last ten days. Specifically, in the national poll average Obama is now up +0.4, Intrade has reached a three week high with a probability of 68% that Obama gets re-elected, and Obama appears to be expanding his lead in Ohio from +0.8 three weeks ago to +2.9 today. According to the expert pollsters, Obama is on the path to re-election. Is it possible that the experts are wrong?
To start let’s look at Gallup, the longest running Presidential pollster, going back to 1936. In the nineteen elections going back to 1936 that Gallup has polled, the organization has correctly picked the winner 16 of 19 times, so it has been correct 84% of the time. The three elections that they missed were as follows:
- 2004 – In their last poll leading into this election, Gallup had Bush and Kerry tied at 49.0%. Bush won the popular vote with an edge of 50.7% to 48.3%;
- 1976 – In their final poll prior to the election, Gallup had Ford at 49% and Carter at 48%. Carter won 50.1% to 48.1%; and
- 1948 – In the classic miscalled election, Gallup had Dewey at 49.5% and Truman at 44.5%. Truman won 49.5% to 45.1%.
If we can use Gallup as a proxy for the professional polling industry, then you would have been correct betting on the experts 84% of the time for the popular vote. The professional pollsters, in aggregate, are currently predicting an Obama win, primarily based on the President’s stronger polling in the battleground states and his edge in the national polls, albeit slight. With a 68% probability on Intrade that President Obama is re-elected, this polling data is all but priced into the market.
The probability on Intrade of course assumes that the polls are accurate. In this election, more so than many prior elections, the validity of polls is being increasingly questioned, despite their long term track record of being accurate. The key doubt related to polls this year is whether the samples accurately reflect the voters who will actually turn out.
In its most recent poll, the National Journal critiqued its own methodology in “The United Technologies / National Journal Congressional Connection Poll”. They wrote:
“The poll’s finding of a 50-45 Obama advantage in the presidential race highlights the central uncertainty surrounding the blizzard of late campaign polling: What will the partisan and racial composition of the actual electorate look like?
In its likely-voter model, the Congressional Connection Poll projected that the 2012 electorate will be virtually unchanged from 2008, with Democrats holding an 8 percentage-point advantage among voters (compared with 7 points last time) and whites representing 73 percent of voters (compared to 74 percent last time).”
In the table below, we’ve looked at the voter ID breakdowns for the most recent polls that provide voter ID from national polls, Ohio polls, Nevada polls, and Virginia polls. In this analysis, we’ve used those three states as they are considered by many to potentially be the deciding states in the Electoral College vote. In fact, Nate Silver from the New York Times put the probability of these three states being the deciding votes at 71% collectively.
The high level takeaway is that across the board the polls have meaningfully higher voter IDs projected for Democrats. Based on the national polls it is a +7 advantage, in Ohio it is +6.3, in Nevada it is +8.3, and in Virginia it is +6.7. If there is in fact more Democrats than Republicans and nationally in these key states, and they turn out in comparable numbers to the Republicans, then President Obama likely has the election won, even with an overwhelming swing of independents to Romney. Conversely, if these voter IDs are incorrect, then Governor Romney remains very much in contention. The prior two elections provide some insight into this voter ID question.
In 2008, there is no question that more Democrats turned out and more Americans identified as being Democrats. In 2004, with Bush’s victory, this was most definitely not the case. The key question is whether this election looks more like 2004 or 2008. In the table below, we look at average voter IDs in the polls above and actual voter IDs from exits polls in 2004 and 2008.
As the table above shows, the voter ID by party in the 2012 polls looks much more similar to 2008 than 2004. This is surprising given that on a national level the race looks much more like 2004 than 2008. Intuitively, it doesn’t make sense for the race to be effectively tied on a national basis but for state level polls to reflect voter turnout comparable to 2008, a year when the Democratic candidate won the popular vote. To be fair, we are comparing apples to oranges on some level as the table above looks at exit polls versus pre-election polls. Regardless, all measures suggest that Democrats are less enthusiastic in 2012 than in 2008.
In the last table below, we return to the nation’s longest running polling firm, Gallup. In this table Gallup compares the demographics of likely voters in 2004, 2008 and 2012. They find the following:
- In 2004, 37% were Democrats, 24% were Independents and 39% were Republicans;
- In 2008, 39% were Democrats, 31% were Independents and 29% were Republicans; and
- In 2012 they estimate, 35% are Democrats, 29% are Independents and 36% are Republicans.
So in contrast to much of the polling industry, Gallup is expecting more Republicans than Democrats to vote in 2012. It is no surprise then that Gallup has Romney up 5% in its latest poll, which was given on October 29th before Hurricane Sandy hit. If Gallup is accurate, Romney will be the next President. But if Gallup is accurate, then the crowd of experts is wrong and the state polls, in particular, will be proven profoundly inaccurate.
We say profoundly inaccurate because historically state level polls have been spot on in predicting the winner of the Electoral College. Nate Silver from the New York Times runs a probability model on the election based on state level polls and he currently has an 86% probability that Obama wins re-election. This is due to the fact that the key battleground state polls currently lean towards Obama on average and going back to 1988 the state poll averages have been accurate in calling the winner of the state 96% of the time.
So in summary, the experts, via their opinion polls, currently show a high likelihood that Obama gets re-elected. For them to be wrong, it is likely because of broad sampling errors based on party affiliation, which as we outline above there is some evidence to support. This also jives with many of the economic predictive models that point towards a Romney victory, such as the recent call we did with Dr. Ken Bickers from the University of Colorado.
We can assure you of one thing: this is going to be one of the closest races in Presidential history and pollsters won’t decide the outcome. As former Congressman from Minnesota Walter Judd famously said:
“People often say that, in a democracy, decisions are made by a majority of the people. Of course, that is not true. Decisions are made by a majority of those who make themselves heard and who vote - a very different thing.”
Get out there and vote!
Daryl G. Jones
Director of Research