Number-Cruncher on Polls’ History of Underestimating the GOP
By Jim Geraghty
September 28, 2012 3:05 P.M.
One of my regulars, the accounting-minded poll watcher nicknamed “Number-Cruncher” writes in, describing what he thinks honest pollsters should be saying right now:
“For the past two election cycles the partisan divide in this country has been volatile. In 2008, we could have modeled the turnout in race similar to 2004 and Obama still would have beaten McCain by 1 or 2 points. We knew that there was no way the divide was going to end up even, so even moving to a 1996 model of +3 Democrat advantage would give Obama a 4 point win. Democrats ended up with a +7 partisan ID advantage, given an almost perfect storm for the Democrats.
The 2008 cycle was an interesting race for pollsters, in that while the partisan divide clearly favored the Democrats, we didn’t have to worry too much about overestimating or underestimating too much because we knew the main result: an Obama win.”
Looking back through recent presidential cycles, we see Republicans over-performing their standing in the final polls – sometimes by a little, sometimes by a lot.
In 1992, Gallup’s final poll had Clinton winning by 12 percentage points, he won by 5.6 percentage points. In late October 1992, Pew had Clinton up 10.
In 1996, some reputable pollsters had Clinton winning by 18 percentage points late, and Pew had Clinton up by 19 in November; on Election Day, he won by 8.5 percentage points… In 2004, pollsters were spread out, but most underestimated Bush’s margin. (2000 may have been a unique set of circumstances with the last-minute DUI revelation dropping Bush’s performance lower than his standing in the final polls; alternatively, some may argue that the Osama bin Laden tape the Friday before the election in 2004 altered the dynamic in those final days.) In 2008, Marist had Obama up 9, as did CBS/New York Times and Washington Post/ABC News, while Reuters and Gallup both had Obama up 11.
Now, if this was just random chance of mistakes, you would see pollsters being wrong in both directions and by about the same margin in each direction at the same rate – sometimes overestimating how well the Democrats do some years, sometimes overestimating how well the Republicans do. But the problem seems pretty systemic – sometimes underestimating the GOP by a little, sometimes by a lot.
This is an international polling problem. Look at the polling for the most recent presidential race in France, if you read the tracking polls you would have thought Sarkozy would lose by 20….then the last round of polls showed it in single digits, and Sarkozy ultimately lost by about 3 points.”
Going back to the topic of volatility, in 2008, Gallup provided a model called the “expanded” likely voter model; they knew turnout was likely to be different from past cycles, but they knew that the different turnout was almost certainly going to help Obama. So they used this poll (they ran four different polls Adults, Registered Voters, Likely Voters and New Voters). In the end it wasn’t necessary because the regular bias of registered voters was enough to offset the “new voters”…. but that’s for another day.
Here is what people should know is bothering pollsters, and if you’re a Republican you can feel comfortable that what you are reading is based on guess work assumptions:
In 2010, we saw the country move back to 2004 levels, but we also saw a bubbling of the Tea Party, who are among the most enthusiastic of voters. Also 2010 was a midterm, where the overall turnout of registered voters is considerably lower, and the GOP base turns out better in non-presidential years than the Democrats’ base. So we process this data.
We saw in 1994 the GOP do very well, but in 1996 Clinton won easily. But sometimes a party’s momentum from the midterms carries on to the following year; we saw the Democrats add to their 2006 gains in 2008. So will 2012 be a receding of the tide of the midterms (like 1996) or an acceleration (like 2008)?
Of course in 1996, the economy was soaring and right now, we’re crawling… so you make the judgment on where this should be.
Even using logical deductions, it is difficult to get a read on what the 2012 partisan divide will be because we’ve seen it change so quickly. From 1994 through 2004, the partisan divide was fairly stable, moving no more than 2 points from cycle to cycle.
Personally I think its safe to say that 2008 is not going to happen in 2012, any pollster hanging their hat on 2008 sampling cannot be reasonably relied on…
Number-Cruncher and I part company a bit on this point:
Given the intensity of the Tea Party, it would not be all that surprising if the Tea Party/GOP combination out polls Democrats by a margin greater than 2004, which would turn every pollster except Rasmussen upside down, with Rasmussen being turned on his side.. Simply put, we just do not know.
The problem is that “not GOP, but Tea Party” isn’t listed as an option in most polls, so we don’t know how many Tea Partiers are choosing to identify themselves as independents. It is quite possible that in the polls where we see Romney winning independents, his lead in this demographic is driven by Tea Partiers who refuse to self-identify as GOP. In short, you know how we’ll know the combined demographics of the GOP andTea Party makes up a larger share of the electorate than self-identified Democrats? When Romney wins the election.
One other point to keep in mind, is that Rasmussen has been consistently polling party preference ID, among adults (not likely voters). His latest result was +4.3 Republican and while that is a bit of an outlier, he has consistently been polling Republicans ahead of Democrats by about a 1 to three point margin. Also consider this: In 2008 when the electorate was breaking towards Obama and the Democrats, Rasmussen predicted a +7.8 percent Party Advantage, the exits revealed essentially the same result. . In 2004, Rasmussen revealed a Partisan ID trend favoring Democrats by 1.8% percent. If Rasmussen goes or comes close to three for three on the partisan ID prediction (he was within two points both times), then Romney likely has a 2 to 3 point lead in his polling (Note if you subscribe to the pay side of Rasmussen’s data you know his is polling more Democrats than the Partisan ID study). Simply put, if Rasmussen is correct, then Romney will has an electorate which is MORE favorable than 2004. If this is the case with swing states, the Electoral College will break significantly towards Romney.
I still think a D+3 or D+4 electorate is the most likely scenario, but Rasmussen’s measurements do provide one piece of evidence for a scenario that’s considerably better for the GOP.