Who's Going to Vote This Year?

Who's Going to Vote This Year?

By Sean Trende - October 22, 2010

The most difficult job a pollster has is trying to figure out who the actual voters are going to be in a given election year. This is easier said than done, because we know that (a) almost all survey participants say they will vote in the midterm election and (b) historically, only about 40 percent will.

Pollsters do their best to solve this problem by screening out those who are unlikely to vote using a question or series of questions probing interest in the election and/or prior voting behavior. These techniques vary widely from pollster to pollster. Some pollsters use especially "loose" voter screens: asking only, for example, if someone is certain to vote, without probing any deeper.

For example, simply asking respondents if they are certain to vote (used by Suffolk) will sometimes let more than 90 percent of respondents through a screen. In such a situation, nearly half of the respondents who are counted will not actually vote.**(update below) Even tighter voter screens tend to let through a significant number of respondents who will not actually vote.

In most years, this isn't such a big deal. If these "excess voters" are more-or-less spread evenly between the supporters of both candidates, then they will cancel each other out.

But what happens if one party's supporters are significantly more enthusiastic about voting in an election? In 2008, for example, national pollsters did a good job of modeling for an electorate that included historic African-American turnout and many "first time voters," while polling at the state level - particularly in battleground states - struggled with the dynamic of the election and in some cases ended up understating Obama's eventual support on election day.

This year, it appears to be the reverse. A preponderance of the data seems to indicate a considerably more energized vote from Republicans this year. But given the discrepancies we seen in some of the polling data, it's clear pollsters are struggling to come to a consensus on who is going to turn out and vote on November 2.

Let's start by looking at some concrete evidence of what the electorate looked like in last year's elections in Virginia and New Jersey. The dynamics of the electorate haven't changed much since then; if anything, they've probably gotten somewhat worse for Democrats.

If we compare the New Jersey and Virginia 2009 exit poll results to the progressively more Democratic electorates of 2004 (near partisan parity nationally), 2006 (slightly Democratic nationally) and 2008 (very Democratic nationally), we get the following results:

As you can see, the electorates in New Jersey and Virginia in 2009 looked a lot more like 2004 than 2008.

In particular, note the ratio of Republicans to Democrats. In both states the ratio was much higher than 2008, somewhat higher than 2006, and about as high as 2004 (a touch higher in Virginia). Put simply, if the trends that drove 2009 have continued, we should expect to see the partisan composition of the electorate that is, on average, somewhere near 2004 levels.

But it doesn't appear that the screens being used by pollster this year are creating an electorate that resembles 2004 at all. Instead, the partisan environment these pollsters' screens are producing is somewhere between 2006 and 2008 - the type of electorate you'd expect to see in a good Democratic year.

A quick examination of the crosstabs indicates that these pollsters are almost all seeing the same basic environment in race after race: Republicans and Democrats are drawing near unanimous support from their partisans, while Democrats are drawing support from 35-40 percent of Independents.

In other words, these pollsters don't disagree on how Republicans, Democrats and Independents are going to cast their votes. Instead, they disagree on how many Republicans, Democrats and Independents are going to cast their vote.

Two recent polls in Washington State provide a perfect example. Consider the following table:

As you can see, the two pollsters agree nearly 100 percent on how Republicans, Democrats, and Independents are going to vote in this race. This means that the difference between CNN/Time's finding that Murray is ahead 8 points and above 50 percent, and PPP's finding that Murray is ahead 2 points and below 50 percent, is almost entirely attributable to differing numbers of Republicans, Democrats, and Independents making it through the respective pollsters' screens.

The following table shows the ten most competitive Senate races (data for every competitive Senate race is appended at the end). The "Polls with Crosstabs" column is an average of all recent polls in a given race that publish crosstabs of voting intentions by party. It is current as of about 10pm on Monday night.

I then took the crosstab data from each of these pollsters and weighted them by exit poll data for each state for 2008, 2006, and 2004. Note that there weren't exit polls in some states for 2006, so I couldn't also perform the appropriate calculation:

So, for example, let's look at Washington again. The pollsters in my sample for Washington State show a 4-point lead for Patty Murray. Had the electorate that their likely voter screen allowed through resembled the 2008 electorate, Murray would lead by 5.95 points. Had it looked like 2004, Murray's lead would shrink to 1.84 points. Looking across the row, we can infer that the pollsters are seeing an electorate in Washington State this year somewhere between 2004 and 2008, but closer to 2008.

The far right column shows the difference between the "Polls with Crosstabs" column and the "2004 Weight" column. As you can see, in every state except West Virginia, the pollsters are showing an electorate that is, on average, 2-3 points more favorable to the Democratic candidate than we'd expect to see in a 2004-type environment. (Incidentally, this is roughly consistent with the difference between the RCP Averages in Virginia and New Jersey in 2009 and the final results.)

It is also the case, somewhat contrary to my expectation, that this tendency is only slightly more pronounced in "Obama surge states" such as Colorado - where registration drives really might have changed the composition of the electorate since 2004 - than in states that were largely ignored by the campaigns, like Washington and Kentucky. Perhaps this should not have been so surprising, given in 2009 Virginia reverted to its 2004 electorate, even though it was the quintessential Obama surge state just a year earlier.

All of this is consistent with what we've seen in pollsters' measurements of the enthusiasm gap. Pollsters who ask respondents to rate their interest in voting on a scale of 1-10 are consistently finding Republicans far outpacing Democrats among those who describe their interest as a 10, with Democrats performing better among low-interest voters. Some pollsters, even those who are showing bad results for Democrats, are letting these low-interest voters through to their samples. This is true even though research indicates that only 1/3 of those who describe their interest level as seven or below on a scale of 10 will actually vote, compared to 84 percent of 10s and 71 percent of 9s.

In sum, no one can know for certain what the electorate is going to look like this year until after Election Day. But it appears that the variations in some pollsters' likely voter screens may be envisioning an electorate that is less slightly Democratic than the 2008 electorate but more Democratic than 2004. Such a model could well turn out to be true.

But an alternate reading is that the exit poll data from 2009, as well as other data pointing to a significantly more energized Republican party, suggests the electorate this year may be far closer to 2004. If this turns out to be a more accurate assessment of the electorate, we could well see some Republican candidates outperforming the polls by 3-4 point margins similar to the results for Bob McDonnell in Virginia and Chris Christie in New Jersey last November.

UPDATE: David Paleologos of Suffolk writes and states that the numbers reported on their questionairre only includes voters who made it through their LV screen, and that the percentage of "adults" included in their poll is much less than 90 percent.  He informs me that the 500 respondents listed in the survey were selected only after thousands of respondents were screened out, which would equate to at most a 25 percent turnout.  (Further update) -- David Paleologos further clarifies that when he was referring to thousands of adults, he was referring to the larger body of non-respondents and unlikely voters.  Altogether, their poll numbers suggest an electorate with 50-60 percent turnout.


Sean Trende is senior elections analyst for RealClearPolitics. He is a co-author of the 2014 Almanac of American Politics and author of The Lost Majority. He can be reached at Follow him on Twitter @SeanTrende.

Sean Trende

Author Archive

Follow Real Clear Politics

Latest On Twitter