Aug 26 2010
Update: ABC News even notes the fact this year’s polls are not proving to be accurate. Which means the only thing one can trust is the trend – which has been heading towards rout of Dems come November – end update
What if this year’s political tsunami was also like an iceberg, where just the tip of the wave was visible right now? In fact, all tsunami waves look like any other wave in the deep ocean. It is not until they reach the shore that the shallowing water forces the wave up to its true height. But specialized sensors can detect these waves even out to sea because they do have features (like wavelength) which set them apart.
As we approach the November elections, we may discover that the primaries we have seen to date are just the tip of the tsunami, which is now hidden from pollsters because of two very basic and acceptable reasons:
- The angriest element of the electorate have tuned out until November. And part of that tuning out is boycotting pollsters. And what if these were mostly independent voters fed up with both parties as well as the entire political process who will be voting come hell or high water? How would that effect polling models?
- The resistance of voter turnout models to massively changing views. Turnout models assume a state’s electorate will behave reasonably close to the way it has behaved in previous, similar election cycles. That is how they turn a 600 person sample into a representation of millions of diverse and independent people. The problem is, if you are off by even a fraction in the turn out model, you will be off by huge amounts when voting day comes around. And if the voters change their mood dramatically, the models will be tuned to the past, not the present.
So here we have two factors that could be hiding a wave of discontent in the poll numbers. In addition, we have two primary elections this week which behaved as if these conditions actually do exist. Though some put it down to the quirks of small voter turnout in primaries which exaggerate forces that will be muted and swamped in the fall when the ‘normal’ voters arrive. That also may not hold water.
In Alaska, Lisa Murkowski’s cruise to renomination for her senate seat looked all but certain. No poll even hinted at an upset. Yet here we are awaiting the counting of absentee ballots to see if she survived. Even if she does go on to survive, what is interesting is the anti-incumbent, anti-big government wave finally appeared out of nowhere with incredible force.
Same thing in FL in the GOP primary. And in fact, FL also was a good method to see the trend where the Democrats are still voting for big government, political elites (or incumbents) while masses of people are voting GOP and voting out the professional pol (just look at the governor primary results).
Sean Trende over at RCP did the analysis I was going to attempt (and he did much better than I would have) that discusses this concept of a hidden wave. He too, like me, looks at the governor races in 2009 and the MA senate special election in January 2010 as indicators of a massive wave heading towards to the Democrat Party – the party of intrusive big government. His approach is very simple and defendable. He tries to ascertain the general mood of the country based on the VA, NJ and MA elections, to see if a common shift in the turn out models has already occured. He finds one:
As IÂ explained in December 2009:
In Virginia, the Republicans’ share of the electorate increased by 12% from 2008 to 2009; in New Jersey it was 10%. In Virginia, Democrats were at about 84% of their 2009 level; in New Jersey it was 93%.
In both states, the Democrats’ share of the Republican vote dropped by about 50% (50% in Virginia, 56% in New Jersey), and their share of the Independent vote dropped about 66%. The Democrats’ share of the Democratic vote was pretty stable; up 1% in Virginia and up 2% in New Jersey.
As we entered December of 2009, Republicans began whispering about a potential upset in Massachusetts. In the absence of much polling, IÂ asked what would happen if the Massachusetts electorate were to shift in the same way that the Virginia and New Jersey electorates did. The results showed a 51.1%-48.9% Coakley win – a much closer race than almost any analyst was suggesting would be possible at the time.
The fact that Scott Brown performed only a few points better than this model suggested implies that the Massachusetts electorate did in fact move in much the same way as the New Jersey and Virginia electorates did (unfortunately we don’t have exit polls to verify this conclusion directly).
What he then did is apply this same general mood adjustment to all the pending senate races, based on previous election results!
The composition of the electorate was altered in each state so that the electorates would be 11 percent more Republican, 5 percent more Independent and 11 percent less Democratic than they were in the 2008 presidential race.
This is important to note in his methodology. He simply models a universal shift in attitude from 2008 to 2010 and determines what would be the outcome of the Senate races. Interestingly, most of the races are now trending towards his predicted outcomes. A few are not, because of the strength or weakness of individual candidates (which actually proves his theory is right). In AR he nailed Blanche Lincoln’s sad situation, and in IL his model shows the two equally flawed candidates would be in a tight race. Trende’s model shows the GOP picking up 12 seats (without accounting for quirks of individual races).
That means the Democrats are NOT starting with +10 (as everyone assumes), but instead are starting at -12. that means they will need to win back 3 seats to keep the majority if the wave is out there and ‘baked in the cake’ so to speak.
Back in late 2009 and early 2010, anyone who claimed WI, CA, WA, DE, PA and others states would be within reach of the GOP would have been deemed nuts. But not now, even left leaning pollsters are starting to sense the swell rising. Take the respected Nate Silver:
The Democratic majority is in increasing jeopardy in the Senate, according to the latest FiveThirtyEight forecasting model. The Democrats now have an approximately 20 percent chance of losing 10 or more seats in the Senate, according to the model, which would cost them control of the chamber unless Gov. Charlie Crist of Florida, who is running for the Senate as an independent, both wins his race and decides to caucus with them.
His prediction now is the Dems lose 6-7 seats. He still has CA, WA, IL and WI going Democrat (as do most pollsters). I like his approach because he is looking historically at how strong poll leads at this stage before an election hold until the election:
The forecasts are based on a program designed to evaluate current polling and demographic data, and to compare these present-day conditions to outcomes in United States Senate races over the past six election cycles.
This approach can be more accurate because it simply determines how solid the November picture right now, based on previous cycles. However, it is also weak because it is based on previous cycles, and assumes poll accuracies this year are as good as previous cycles. That is a questionable assumption (as Sean Trende indicates). What if the turn out models in place this year are still too heavily weighted towards past cycles (instead of the tuning process they will go through in the coming weeks, weighting new poll data to adjust the turn out models to current conditions)?
What if that 5% of the electorate that is so angry and so energized it lifted up the Tea Party movement is ignoring pollster calls? What if there is a largely center-right, libertarian, Tea Party movement not completely visible to the pollsters out there?
If both Trende and Silver are basically correct, then Silver’s model is going to shift right and consume states like IL, CA and WA. It may even take out IL and CT. We don’t know. What we do know is no one saw Alaska coming, and only one pollster detected the Scott win over McCullum in FL – for very interesting reasons:
We used a loose screen in determining who to call that may have picked up more non-typical primary voters who went for Scott. Instead of calling a list of people who had a history of voting in past primary elections, as we usually do, we just called folks who had a history of voting in general elections and then screened on voting intent for the primary from there. If the folks who voted yesterday had been exactly the same as the folks who voted in the 2006 primary I imagine McCollum would have won. That’s because he was the Republican establishment choice and the kinds of folks who vote in every primary likely went to him. But there were hundreds of thousands more people voting yesterday than in 2006 and my sense is the newbies went strongly for Scott.
We picked up a Republican electorate that was exceedingly conservative. In 2008 exit polls showed 61% of Presidential primary voters were conservatives. Our poll over the weekend suggested 72% of primary voters this year identified as conservatives. Given that Scott was winning conservatives and McCollum was winning everyone else, identifying that conservative shift in the Republican electorate probably helped contribute to our poll’s accuracy.
The one pollster that basically nailed Scott’s wave in FL is the one who shifted their turnout model from the 2006 and 2008 parameters to one more conservative and with higher energy on the anti-incumbent side. Coincidence?
I think not.
Especially since now upwards of 100 house seats are in play and the tsunami may have already consumed the Democrat House (not to mention the governor mansions). The tide, she is still arising!