For those who are worried about Pat Toomey’s sudden slip in the polls in Pennsylvania — and I was among them — you’ll want to read this analysis from “Number Cruncher,” a finance/political junkie who spends even more time breaking down polls than I do.
Your readers should calm down. This race isn’t that close, or even closing. The only thing closing is the pollsters’ prediction of who to consider a likely voter. Looking at the Quinnipiac results, I find them very encouraging for Toomey, specifically among Independents. Taking the Quinnipiac crosstabs and applying 2004, 2006, 2008, and my predicted 2010 Likely voter model, the results are as follows:
2004: The last good GOP year, Pennsylvania exit poll crosstabs came in with party ID at 41 percent Democrat, 39 percent Republican, 20 percent Independent. If you apply this breakdown to Quinnipiac’s sense of how each group is voting, it results in a 2.1 percent Toomey win. This year is the most likely turnout comparable to use based on history only, and one I believe understates GOP support in general.
I think Quinnipiac is using something akin to this year as their turnout mode, but 2004 has me scratching my head, namely the 41 percent Democrat party ID support.
It was a very good year for the GOP, yet if you look at party splits we see only a 2 percent upward tick for Democrats in the 2006 exit polls, when the Democrats had a landslide election — in this election Santorum lost by 18 points!.
This is the danger of comparing Presidential and Mid-term elections. A plausible explanation of this can be found in this study from Univ. of PA The exit poll Bush support versus actual support hurt Bush by 6.5 percent. In other words, this year’s exit poll party ID is likely off base.
In fairness to the exits on CNN, they appear to have been adjusted somewhat (they all seem to tie out to the actual results), but I recall as the article study does there might be problems.
Simply stated: How does Specter win by 9 percent in 2004 and Santorum lose by 18 percent in 2006, where the electorate only turned out 2 percent more Democrats? The answer something is wrong — a 2 point party split does not make up 27 percent points, not with a candidate like Santorum who did not have any significant ethics problems. So while I won’t fault Quinnipiac for targeting 2004 as a sanity check, it’s the best one historically, it is nevertheless a problematic data point.
So let us take a few breaths and look at the worst case scenarios for Toomey — applying 2006 and 2008.
2006: Assuming a 2006 turnout percentage — the last big Democrat midterm year, with 43 percent Democrat, 38 percent Republican, 19 percent independent — the race is a dead heat, with Toomey trailing by only 0.6 percent under Quinnipiac’s cross-tabs.
This is fatal news for Sestak! When applying the Quinnipiac cross-tabs for the biggest Democrat midterm election since 1974, yields at best a .6% win for Sestak . . . Can we stop worrying?
2008: So then we look at the biggest Democratic Presidential year since 1964. Cross tabs came in with party ID at 44 percent Democrat, 37 percent Republican, 17 percent Independent.
To believe the race could duplicate this is to say you need to be smoking peyote. Assuming this turnout, including historic Philadelphia turnout . . . at best puts this still a dead heat, with Toomey trailing by just 1.4 percent.
So where is this race really at? Follow me.
Scenario One: Presume Pennsylvania’s turnout will be 36.5 percent Democrat, 37 percent Republican, 24.5 percent Independent. Applying these targets to Quinnipiac’s split yields a pretty easy to call 52.6 percent to 46.9 percent, or 5.7 point victory for Toomey. The way I came up with my splits is to take the 2008 Rate 44-37-17, and adjust based on Rasmussen’s summary of Party Affiliation percentage change in Nov 2008 versus Sept 2010.
Scenario Two: The problem with Scenario One is that while I capture the change of party affiliation, I assume a presidential year base model rather than a mid-term election. In a midterm turnout will be lower — 50 to 70% of a presidential year, which means only the more enthusiastic voters will turn out. Some expect this to be anywhere from 3 to 7 points — overall advantage to the GOP. My model takes a more conservative view with the Dems falling to 34 percent, the GOP up to 38 percent, and Independents to 28 percent. This gives Toomey a very comfortable 9 point win (and is consistent with many polls, with the same cross-tabs as Qunnipiac and PPP).
Scenario Three: But there is another problem with this assumption, which is the Tea Party as an affiliation. As noted, the GOP support has gone down since 2008. How is this possible? The emergence of the Tea Party. Tea Party Independents are going to vote and to be blunt, they will not be voting Democrat. Many of these voters also will not be telling any pollster they “are” Republicans. Going with a cruder model, but one that tries to capture the Tea Party, using flimsy data and guess-timates, I believe the Electorate in Pennsylvania is 34 percent Republican, 37 percent Democrat, 19 percent Tea Party-Independent, and 10 percent Moderate-Independent. Applying the Quinnipiac cross tabs where Republican and Tea Party have the same levels of support for Toomey, and where Moderate-Independents favor Sestak 53 percent to 47 percent yields . . . a 12 point Toomey win.
Thus I believe the margin race will end up between 9.1% and 12%, which is to say I will go out on a limb by taking the midpoint, and predicting Toomey beats Sestak by 10.5 percent.
Bottom line: nothing has changed with this race other than the pollsters’ predicted turnout models.
Separately, a few weeks ago Number Cruncher offered to bet a Ruth’s Chris steak dinner that Tom Tancredo would be the next governor of Colorado. I took the bet, and now I’m starting to worry I’m going to be paying for a filet mignon.