Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

The undecided brain

October 27th, 2008, 11:00pm by Sam Wang

In Tuesday’s New York Times, Joshua Gold and I have an op-ed on the neuroscience of being an undecided voter. It draws upon themes found in my book Welcome To Your Brain: Why You Lose Your Car Keys But Never Forget How To Drive. Full text, with scientific references, after the jump.

As we enter the final week of a seemingly endless election campaign, opinion polls continue to identify a substantial fraction of voters who consider themselves “undecided.” Although their numbers are dwindling, they could still determine the outcome of the race in some states. Comedians and other commentators have portrayed these people as fools, unable to choose even when confronted with the starkest of contrasts.

Recent research in neuroscience and psychology, however, suggests that most undecided voters may be smarter than you think. They’re not indifferent or unable to make clear comparisons between the candidates. They may be more willing than others to take their time — or else just unaware that they have essentially already made a choice.

Neuroscientists have begun to tease out the brain systems that make some simple decisions. Even when it takes a second or less, decision-making is thought to involve two parts, gathering evidence and committing to a choice. In tasks as simple as deciding whether a shifting pattern of dots is moving to the left or to the right, brain activity in the parietal cortex rises as evidence is gathered, eventually reaching a tipping point of choice – though it is not yet known what brain regions drive the final choice.

Inherent to this kind of process is a trade-off between speed and accuracy. Commit early and you can get on with your life. Take more time and you might make a wiser or more accurate decision. Since a commitment to John McCain or Barack Obama is not required until Nov. 4, for the greatest accuracy, one should gather evidence until that date. So then why aren’t there even more undecided voters? For simpler decisions, after there is enough evidence to reach a decision threshold, the brain tends to ignore further input even when it might improve accuracy. The brain goes ahead and decides, freeing up mental resources to deal with other problems.

This logic suggests that undecided voters might simply require a higher degree of confidence before they commit. Pollsters know this, and so push “uncommitted” voters to state a preference. Although this approach may seem heavy-handed, it gives a fairly accurate reading of a candidate’s support. In psychological studies, people who describe themselves as undecided often reveal a pronounced preference when they are forced to choose. When someone reports being only “moderately sure” of a decision like whether to accept a new job, his eventual choice is all but certain.

Still, the person may not be aware of that internal commitment. In one study, people were asked to play a gambling game in which they could choose cards from several decks, some of which were secretly stacked against them. After losing repeatedly, most subjects began to nervously avoid the less favorable decks but were unable to say why until after much further play. People with damage to the ventromedial prefrontal cortex lack this intuition, and so they take inordinate time to make decisions in general.

Of course, undecided voters aren’t suffering from brain damage, it’s just that their brains may require an especially long amount of time to develop confidence in or awareness of a choice. In these cases, hidden commitments can be queried in creative ways. In a recent study, 33 residents of an Italian town initially told interviewers that they were undecided about their attitude toward a controversial expansion of a nearby American military base. But researchers found that those people’s opinions could be predicted by measuring how quickly they made automatic associations between photographs of the military base with positive or negative words.

If decisions are lurking somewhere in the brains of undecided voters, could brain imaging methods reveal their inclinations? Not yet. Recent research has shown that when undecided voters looked at images of candidates, their brains’ emotional centers were often activated. But this reveals little information about the content of their thoughts. This type of research serves mainly to demonstrate how difficult it is for scientists to physically trace complex concepts like preference.

It is more effective to pose indirect questions. Pollsters can learn which way “undecided” voters lean by using questions they already ask that are likely to correlate with support for either candidate: Who do you think understands your problems better? Are you more concerned about the economy or terrorism? Which candidate has the better temperament? The answers of decided voters could be used to predict the final choice of undecideds.

No matter how deeply they delve into people’s thought processes, however, polls will never be perfect predictors of election results. Like the brain of an undecided voter, the electorate as a whole may lean toward one candidate or another, but until the ballots are cast on Nov. 4, it remains undecided.

Sam Wang, an associate professor of neuroscience at Princeton, is a co-author of “Welcome to Your Brain: Why You Lose Your Car Keys but Never Forget How to Drive and Other Puzzles of Everyday Life.” Joshua Gold is an assistant professor of neuroscience at the University of Pennsylvania.

Note: As of today, the states where undecideds outnumber the median polling margin are: Arizona, Florida, Indiana, Missouri, Montana, Nevada, North Carolina, and North Dakota. Removing these states leaves safe states totaling 306 EV for Obama, 147 EV for McCain. So Obama would win today even if undecideds all went for McCain.

Tags: 2008 Election

21 Comments so far ↓

  • jmg

    Sam –

    Slightly off-topic, but… undecided voters seem like many jurors I’ve encountered as a prosecutor. Are you familiar with any studies that shed light on how to identify uncommitted ‘late deciders’ or how to indirectly query hidden commitments?

  • egc52556

    Fascinating insight into decision making. When it comes to the practicality of actually casting a vote, however, I think there is an additional variable, one that motivates the decider to take the action necessary to go to the poll. This, I suspect, is a higher barrier than simply answering a poll question and thus probably requires a higher level of decisiveness. If I’m right, then what does it mean to polling and voting systems? Should pollsters give more weight to voters who have voted before (and apparently have whatever combination of factors it takes to actually cast a vote)? Should voting be so easy that even the most casual “decider” will likely vote?

  • Bill R

    Speaking of the undecideds…Sam, I assume you’ve read Bill Greener’s “undecided = won’t admit to pollsters that they’re voting for the white guy” article in Salon. What’s your take on that? Do you have anything to add to (or, for that matter, subtract from) Nate Silver’s response?

  • David

    Great post. And it is REALLY interesting to compare the list of states where undecided voters exceed the median and them post it on an interactive map leaving these states as toss-ups and assigning the rest per the median OR using’s projections.

    (The answer is left as an exercise….)

  • Observer

    Science aside, very nice to know that the undecideds are bunched in states that McCain should be winning. Suggests that McCain could lose some of these states to the Obama momentum?

  • Integrator

    Sam, great article!
    Do you really believe that undecideds are trying to maximize their accuracy by collecting more information? I wish it were so. A well known failure of such integrator models of decision making is their inability to account for long reaction time errors.

  • Sam Wang

    Integrator – no, my opinion is what’s expressed in the second half of the piece. I think undecideds are just as committed as the rest of us. They just don’t know it yet.

  • Matt Incantalupo

    What if we pushed undecideds a little harder? What would be the consequences of asking them “Do you think you will be ‘decided’ by Nov. 4th?” A follow-up could be “If you are still undecided on Nov. 4th, will you still go to the polls and vote?”

  • Frank

    Sam: The advantage in using an even number of polls (when available in a short enough period) is that the median margins are in half rather than whole points, producing a more continuous distribution of win probabilities. With an odd number of polls, the win probability jumps from 50 to 69 to 84% as the margin jumps from 0 to 1 to 2%. (There are no current examples of a margin of 0.5%, but NC is now at 1.5% with a win probability of 77%.) In a closer race, this might matter.

  • Sam Wang

    Matt Incantalupo – I believe Survey USA and Rasmussen do this. It’s a good idea.

    Frank – you are focused on the numerator. The denominator, the estimated standard error, is much more continuously distributed. It is at least one percentage point, so that fractional percentage points would largely add only apparent precision.

  • Frank

    Observer makes a good point. By Sam’s numbers, McCain is expected to lose 11 of 31 Bush states, though this may go down to 5 or up to 13 depending on what the undecideds do in the 8 states that Sam lists where undecideds exceed the margin. (Of these 8 states, my own feeling is that McCain will gain the lead in ND, IN, and NC, and keep the lead in MT and AZ. NV, FL, and MO, where he’s trailing, are another story.) However, this implies a range of 306 to 391 electoral votes for Obama.

  • ndam

    David Sedaris has a very funny description of being undecided in this weeks New Yorker:

  • Andrew Foland

    Chris Hayes also wrote of his experience of canvassing undecideds in 2004. My favorite line:

    More often than not, when I asked undecided voters what issues they would pay attention to as they made up their minds I was met with a blank stare, as if I’d just asked them to name their favorite prime number.

  • Helen Good

    Norman F. Dixon’s 1976 book ‘On the Psychology of Military Incompetence’ gives early decision making as one cause of military incompetence. He gives examples of intelligence reports being ignored because a decision had already been taken. He describes a psychological inability to hear new information that corrects the defective information on which the decision was based.

    If you see this elsewhere, it was posted in error, it is meant to be here.

  • Paul

    Sam – I wonder if the attention on undecided voters isn’t a bit misplaced, as it is the difficulty in persuading decided voters to switch that would seem to predict the accuracy of polls. It seems that a lot of the analysis of polls (both in the media and in some of the projection websites) treat percentage differences as equivalent (i.e., leading 46%-42% is the same as leading 53%-49%). However, I think that the value of polling at or above 50% (accounting for margin of error) has a particular value in that it requires voters to change their previously stated decisions. I am a psychologist (PhD is in clinical, but I have a lot of background in social psych as well), and the majority of research in my field suggests that people look to make confirmatory decisions rather than differential ones – which is to say that once people have made a decision, they are more likely to seek out and attend to information that supports their decision rather than undercuts it, and are also more likely to interpret ambiguous information as supportive. Indeed, I believe that studies have even shown that forcing someone to make a decision (i.e., without giving them enough information to decide) influences their future behavior and processing of confirmatory vs. incongruent information.
    Given that Obama is polling at or around 50% in numerous swing state polls, this would seem to suggest (given the accuracy of the polls) that it is the currently decided rather than the undecided voters that will determine the outcome of this election, as McCain can not win without convincing some decided voters to change their minds.

  • Sam Wang

    Paul, these are good points. I wrote about biased assimilation of information in a previous NYT editorial, here. These types of phenomena are also covered in my book.

  • Frank

    Sam: I think it’s unrealistic that (a) there are only 11 states that are not sure bets (p > .995) for Obama or McCain and (b) all of those states favor the leader by more than two-to-one. (Yesterday it was 10 states and, except for a tie, four-to-one.) It seems to me that the margins are all being divided by around 2% to get the z-scores that yield the probabilities in your input file. That is, the z-scores are basically scaled versions of the margins, and because the margins tend to increase by 1% the z-scores increase by ½ which raises the probabilities rapidly. If you’re tied one day and lead by 1% the next, your win probability jumps from 50% to 69%. Please straighten me out if I’m wrong, as I assume I am.

  • Michael S

    Rather than considering that stunned deer in the headlights look as a sign of stupidity consider instead the reaction as being confirmation that the decision is innate.

  • Mathphysto

    Sam, thanks for the great site! I’ve been lurking for a while here and have made it part of my election-coverage diet.

    What about errors in the decision-making process? Specifically, what is happening neurologically when a voter decides to support a candidate that they mistakenly believe agrees with their own views on several issues?

    I’m thinking in particular of cases where cognitive overload leads a voter to forget a candidate’s platform/past record on some issues when integrating the evidence to make a decision. Many people make their decisions without writing down the evidence and organizing it in a systematic way, so there seems to be plenty of room for such errors. Who can easily juggle evidence on a few dozen major multi-faceted issues to make a decision that optimally represents their own intrinsic opinions? Seems like it would easily exceed the short-term memory abilities people have.

    Do the neural patterns corresponding to certain pieces of evidence just ‘drop out’ from the decision-making process? Or maybe there is some sort of piece-wise evidence integration process that starts with a decision based on one ‘chunk’ of evidence (perhaps the most important), then adds another chunk and ‘recalculates’ the decision, and repeats this process until the evidence is accounted for? Essentially like a finite series starting from a zeroth approximation, then adding a first, second, etc, until you get the final result.

    Relatedly, do you have any idea about the degree to which emotion can alter the decision-making process? My intuition says it would be able to cause evidence to be (a) re-weighted (a special-case of this being ignorance, i.e., zero weighting), and/or (b) altered in perception/interpretation. But I’m not sure how well the neural end of this is known.

  • Sam Wang

    Frank – Your point is basically correct. Just as donations are most effective in knife-edge races, a 1% jump has a relatively big effect on the win probability. Realistically, a 20% jump is a reasonable measure of change in certainty. Also, at this point in the season I use more than 3 recent polls in most close states.

    Mathphysto – You are asking fascinating questions. Contrary to popular belief, emotions play an essential positive role in decision-making. As I have written in Welcome To Your Brain, emotions tell us when an event is important, positive or negative. When people have brain damage that impairs their ability to link emotion with higher processes, they become horrendously indecisive. It’s fairly well understood – read my book! Of course such a mechanism can be overactivated.

    In regard to biases, there’s no such thing as unbiased integration of information. All information comes in against a backdrop of prior evidence. These “priors,” as they are called, guide us to accept some evidence more than others. This is essential for survival in the natural world.

    In political decision-making, there’s no feedback that corresponds to survival. Voting for Democrats or Republicans doesn’t get you eaten, mostly. Even when you vote for a party that drives the economy into the ground (or allows the U.S.S. Cole to be attacked, or whatever your favorite bad event may be), it wasn’t your personal doing. So biases stick around.

  • Mathphysto

    Thanks, I do plan to get the book, perhaps as an Xmas gift :)

    My main question wasn’t about how evidence is biased via perception, etc. Rather, I was wondering how overloading of working memory is understood at the neural level, particularly in the context of decision-making. In other words, I’m wondering how information-loss functions at the neural level.

    More generally, why is there a limit to short-term memory, even outside of decision-making tasks (e.g., Miller’s 7 +/- 2 result)? What happens to the information neurally, how does it just ‘disappear’?

    Is it a lack of necessary pre-existing synapses, so that some firing patterns just hit dead-ends? Or some sort of signal attenuation when there are too many firing patterns competing for shared resources?

    Maybe I should just read the book, huh?

Leave a Comment