Morning Jay: A Polling Quandary—in the Buckeye State
October 31, 2012 6:00 AM
There is a peculiar divergence between various public opinion polls at the moment. On the one hand, Mitt Romney has built a narrow but durable lead in the national polls, averaging around a 1 percent advantage over the last three weeks. This has cheered the hearts of conservatives everywhere.
Yet, liberals retort, Obama has a lead in enough swing states to add up to 270 electoral votes, and that is really what matters.
What to make of this?
For starters, they cannot both be right. If Mitt Romney wins the popular vote by 1 percent or more on Election Day, the odds that he will lose the Electoral College are quite small. To put this into perspective, there have only been four times when the popular vote winner and the Electoral College winner were different in predominantly two-way matchups: 2000, 1960, 1888, and 1876.
In 2000, 1960, and 1888, the margins separating the two candidates was roughly half a percent or less. Only in 1876 was the margin greater than that – with Democrat Samuel Tilden winning the popular vote by 3 points over Republican Rutherford B. Hayes, but still losing the Electoral College by a single vote. Today, historians generally agree that Republicans stole that election (although in fairness, the Democrats stole it first by suppressing the black vote in the South).
So, this is a problem: The national and state polls are not in sync. But when you take a closer look at the state polls – in particular, Ohio – you see disagreement among state pollsters
If all the state pollsters were polling the same prospective electorate, then the only differences we would see between them would be random. And if we graphed the Obama-Romney margins, we would expect to see them look something like this:
But in fact that is not what we see. Instead, we see this:
This is not definitive evidence, but it is certainly suggestive of a non-normal distribution. It explains why the average of all the Ohio polls shows Obama up by 1.8 percent over Romney, but the median (i.e. the middle margin) has him up by just 1 point. Again, more evidence that the polls are not behaving the way they should be.
It is even more interesting when we consider that the same pollsters tend to be finding the same results. The next picture recreates the previous one, only this time with the addition of which pollsters are finding which results:
(Ras = Rasmussen; WAA = We Ask America; SUSA = Survey USA; Purple = Purple Strategies; CBS = CBS/New York Times/Quinnipiac; NBC = NBC News/Marist.)
Again, this is not something we should expect if all the pollsters were sampling from the same population. Instead, we sometimes should see Gravis have a better number for Romney than Survey USA, or Rasmussen look better for Obama than CBS.
What does this all mean? In the case of Ohio, it suggests there are basically two views of the race.
The first, shared by Rasmussen, Gravis, We Ask America, ARG, the Purple Strategies Poll, and the Ohio Newspapers Poll, shows something between a dead heat and an Obama lead of 1 point. The second view, shared by PPP, CBS, NBC, Time, and CNN, shows the president with a comfortable lead of 4 to 6 points. (Survey USA and Fox fall somewhere in the middle, and not coincidentally find a relatively large number of undecided voters.)
When you average out these two views, you get an Obama lead of 2 points – but that does not appear to be very meaningful, considering that just 2 of 25 polls in the last month have found him up by that margin!
Just like the difference between the state and national polls, they both cannot be right.