Misreporting Voter Turnout

President Obama is currently leading by as much as 4-5 points in the polls if attention were restricted to registered voters. Yet, polls based on the subsample of “likely” voters show a different story, with Mitt Romney equal or slightly in the lead. How sure are we that these “likely voter” corrections adequately assess who will turn out on November 6?

One problem is that people often misrepresent whether they voted. The people who feel the need to give the socially desirable answer in surveys may differ systematically from those who don’t. In a new article in Political Analysis (ungated version here) Stephen Ansolabehere (Harvard) and Eitan Hersh (Yale) validate what survey respondents in all 5o states said they did against public records from the 2008 elections. Their key conclusion is that:

[..] standard predictors of participation, like demographics and measures of partisanship and political engagement, explain a third to a half as much about voting participation as one would find from analyzing behavior reported by survey respondents.

This finding occurs because:
Well-educated, high-income partisans who are engaged in public affairs, attend church regularly, and have lived in the community for a while are the kinds of people who misreport their vote experience.

So: surveys exaggerate the differences between the characteristics of non-voters and voters. Yes, non-voters are less educated than voters but education is not nearly as important a correlate of actual turnout decisions as it is of respondents’ answers to the question whether they voted (and presumably whether they intend to vote).

This may cause trouble for “likely voter” corrections. If I understand it correctly, “likely voter” corrections are usually based on what Ansolabehere and Hersh call the “standard predictors of participation” and/or self-reported voting intentions or past behavior. If Ansohalabehere and Hersh’s findings generalize, then the difference between likely voters and registered voters may be smaller that currently reported, which could be good news for Obama. This conclusion relies on some additional assumptions and admittedly a cursory understanding of the actual likely voter corrections that pollsters use. Am I wrong?

17 Responses to Misreporting Voter Turnout

  1. Nadia Hassan October 26, 2012 at 3:12 pm #

    LV screens usually work pretty well! This post at pollster.com shows they usually do help identify those most likely to vote, and they were very accurate in 2008 and 2004. The gap was less pronounced then, however, with RV polls averaging .8 for Bush and LV polls averaging 2.2 points.

  2. Bo October 26, 2012 at 3:16 pm #

    No, you are not wrong. Most national polls also assume that minority voters will not turn out like they did in 2008, even though they are the fastest growing subsection of the electorate. Further more, early voting returns have reflected 2008 levels of minority turnout. If that continues, we’re looking at another Obama landslide marketed by the media as a nail-bitter.

  3. Bob"Romney writes Cocession already" Jones October 26, 2012 at 5:12 pm #

    I agree, Registered voter to Likely voters is assuming wayyyy too many registered voters won’t vote. This under counts all the new people that campaigns register . The only reason these people register at all is simply to vote.

    It seems that every demographic that is out there that favors the Democrats is under sampled every election. Thats why while so many pollsters were close in the final outcome they were horribly off in the demo numbers in 2008. Polls are oversampling the South, Under sampling Hispanics, Asians, African Americans, women, cell phone users, tablet only users and younger voters. Those demo’s all favor the democrats.

    It’s funny the one area tat pollster have right consistantly is white men.

    Obama will probably get an estimated (by me) on my computer model 358 electoral votes. Everyone will be shocked when Obama wins Arizona, but acording to my calculations every poll has under sampled Latino’s there. By my calculations you can add 4 to 6% points to every Arizona poll you see in the Dems favor.

    On Nov 7th the media will say how in the end every undecided voter must have swung to Obama, but the truth is the polls use out dated demographics.

    • Lee in Iowa October 26, 2012 at 10:50 pm #

      I agree! Also undersampled are cellphone users, who differ both in demographics (cells being the only phone of the under-30, educated voters) and psychographics (cell phones being a key part of life for the over-30, but forward-thinking, early-adopting, progressive-thinking, educated, upwardly mobile voters). I figure this gives Obama another nearly invisible 5 to 7 point edge.

      • Chris October 27, 2012 at 11:17 am #

        A FIVE to SEVEN point edge? You mean the difference between a dead heat and a landslide? No sir. That’s unfathomable, and completely inconsistent with all the data in past elections.

        Cell phone only users are already far less likely to vote, and pollsters weight to demographics for the younger people they do get to compensate for things like this. It’s not perfect, but it usually works quite well. Moreover, the pollster that has started including the most cell phones–Gallup–has the best Romney results.

  4. Ben October 26, 2012 at 5:38 pm #

    If one looks at the Census Bureau’s data on voting and registration (http://www.census.gov/hhes/www/socdemo/voting/), it is clear that roughly 88+/-% of registered voters vote in presidential general elections. However, if we look at the polls which report results for both RVs and LVs, all such polls have percentages of LVs/RVs of far below 88%, and generally have below 80% of LVs to RVs.

    This suggests likely voter screens are far too stringent, and therefore are likely introducing a bias to the results.

    • Lee in Iowa October 26, 2012 at 10:51 pm #

      And Ben, I am starting to think that bias is deliberate, since it magically screens out more Democratic votes than Republican ones. (And how better to fiddle the e-voting numbers than in a supposedly close election?!)

    • Chris October 27, 2012 at 5:58 pm #

      Which chart, specifically, are you looking at? At first glance it looks like 70% of registered voters voted in 2004, and fewer than that in 2000. I think you might be looking at something else.

    • Chris October 27, 2012 at 5:59 pm #

      Whoops, forgot the image:


  5. Chris October 27, 2012 at 11:18 am #

    If they introduce regular bias into the results, why aren’t the results consistently several points more favorable to Democrats than the polls say? At some point these conspiracy theories have to be exposed to data. And when they are, we see that they simply do not fit.

    • Craig October 28, 2012 at 2:37 pm #

      Well, in 2010 Democrats significantly outperformed the polls, despite losses everywhere.

      You’re wrongly assuming the electorate is static, while in reality the groups that are being purportedly undersampled – Hispanics and cellphone only voters especially – are the most rapidly growing subpopulations. Pollster assumptions circa 2004-2008, when polling was very accurate, are becoming less appropriate every year.

      • Chris October 28, 2012 at 3:12 pm #

        I’m not assuming the electorate is static at all; it’s clearly not. The question is whether or not the polls routinely and significantly underestimate Democrats. And it’s pretty clear, comparing the polling to the actual results, that they haven’t.

  6. Will October 27, 2012 at 3:51 pm #

    Chris, I believe Lee answered your question when he referred to “e-voting numbers”…see the following for evidence that is suggestive of a “red bias” in e-voting machines:



    I realize that its important for your side to refer to such concerns as “conspiracy theories” in order to delegitimize them, but I do not think such fears should be so quickly dismissed (also conspiracies do happen, don’t they?)

  7. Chris October 27, 2012 at 5:52 pm #

    I don’t think it’s just “my side” that thinks of this as a conspiracy. The idea is positively littered with holes. For one, a 7-point edge last election was roughly 10 *million* votes. This isn’t electioneering around the margins in a close election, this is an accusation of fraud on a mind-blowingly level. I’m not calling it a conspiracy theory to discredit it: it would, in fact, be a conspiracy. And a big one.

    Second, voting machines are relatively new inventions. How then to explain polls that predate them? The voting machines have no observable connection to the polls. Is this all a big coincidence, or is the suggestion not only incredible levels of systematic fraud, but voting machine makers and pollsters (almost all pollsters, it would appear!) would have to be actively and constantly collaborating, too.

    Third, we even have explicitly Democratic and Republican pollsters, and while they probably have their biases, they almost never show anything on the order of 5-7 points consistently. How could this be?

    Fourth, as I pointed out, Gallup surveys more cell phone users than other pollsters, yet this cycle seems to be finding better results for Romney, not worse.

    And these are just off the top of my head.

    • Karenj October 28, 2012 at 12:49 am #

      Chris do you think that Rasmussen is reporting unbiasedly? He (and Gallup) are way out there in no way consistent with most other pollsters. Do they have a magic ball or are they just plainly wrong? Rasmussen is simply biased because he is a republican. I also find it odd that we dems and liberals are called out for being unfair when we conceded the 1st debate to Romney. We just had to admit it, that IS fair. But when I hear from GOP that Romney won all 3 just because they say so, I wonder. I live in Ohio and when I go for long rides I am finding so many clues that Romney supporters do not have one( A clue I mean) Me and my husband laughed so many times to see Romney signs next to John Boehner signs. Um er ah Boehner is not running for anything people. I see signs with Romney and Sherrod Brown in the same yards as well. Do they know what party they are voting for and if the 2 men are completely opposed to each other in views? No. Then I see many many signs saying Obama your fired, but no Romney signs, so who would replace Obama if he is fired guys? You would think the signs would go hand and hand but….. I also have seen thousands of Josh Mandell signs EVERYWHERE. Today we saw Romney and Mandell signs along the highways. Even one at a very dangerous exit. That means that their volunteers go along our highways stopping and putting up signs while endangering drivers! I can not imagine anyone stopping at that exit and posting signs. it’s just too dangerous. But they should understand that signs do not make anyone change votes. Mandell has proven one thing, he can order the most signs. Since he has proven little else I voted for Brown 3 weeks ago… Oh yeah and had the honor of voting yet again for Mr Obama.

      • Chris October 28, 2012 at 9:59 am #

        Boy, this comment is all over the place. Placing signs in dangerous places? Eh?

        Most of this has absolutely nothing to do with the topic, which is whether or not it’s actually plausible that there’s a conspiracy among all pollsters and election monitors to reduce Democratic votes and responses in polls at commensurate levels to steal elections to the tune of 5-7 points. That’s what was actually suggested above (by two different people), and that’s what I was rebutting. Most of what you’re saying doesn’t have any relation to that.

  8. Ray October 29, 2012 at 10:55 am #

    Chris, you are correct. The conspiracy theory about “rigging” voting machines is just that, a “conspiracy theory.” The voting systems are tested rigorously by certified labs and by state authorities and subject in many states to post-election audits to ensure accurate tabulation and reporting of each vote. Election administration certainly can be improved and some excellent recommendations are made in a recent report from the Cal Tech/MIT Voting Technology Project, where two of the principal authors are Michael Alvarez at Cal Tech and Stephen Ansolabehere at Harvard. The report is available on the web: “Voting: What has Changed, What hasn’t, & What Needs Improvement.”