Can we predict the “undecided voters”?

Lars Plougmann / Creative Commons

Now that Election Day is upon us, can you remember way back (a whole few weeks ago!) when there were still those mystical Undecided Voters? Even aside from the sketch comedy skits that mocked their very existence, many people were still wondering how voters could possibly have been “undecided” so close to the election.

However, according to research from Silvia Galdi, Luciano Arcuri, and Bertram Gawronski, those voters might have been “decided” after all – even if they didn’t realize it at the time.

Attitudes are traditionally measured using either explicit or implicit measures. Explicit measures are what people tend to think of when they picture attitude questionnaires. They consist of verbal, conscious attitudes that the participants can explicitly endorse or reject (e.g. “How do you feel about Romney?” or “Do you plan to vote for Obama?”). Implicit measures, on the other hand, are sometimes referred to as “hidden” attitudes (though this is a bit of a misnomer). For example, implicit measures might use cues like body language or reaction times to gauge how people truly feel about something, without them having to directly verbalize those opinions. This can mean that it’s an attitude that the participant is intentionally trying to keep secret (hence the “hidden” qualifier), or simply that it’s an association of which the person is unaware.

In one recent study, Galdi, Arcuri, and Gawronski wanted to see if implicit attitudes could add anything to our understanding of voting behavior that conscious attitudes don’t already tell us. They began by asking Italian citizens about their opinions on a political issue – namely, there was a proposition to enlarge an American military base in Italy, and the participants would eventually have to to vote on whether or not they approved of the military action. Some of the Italians had formed conscious attitudes about the proposition far in advance, while others were self-professed “undecideds.” After 2 months, the researchers returned and asked the participants to indicate how they had actually voted.

To measure the voters’ conscious (or explicit) attitudes at the first time point, the researchers simply asked them to indicate a) if they were in favor of the proposed military expansion, and b) their responses to a 10-item survey on various conscious political beliefs. To measure automatic associations, on the other hand, voters took an Implicit Association Test (or IAT) about the proposed military expansion. The IAT is the most well-known indicator of implicit attitudes; it uses people’s categorization reaction times as a split-second, uncontrollable indicator of automatic positive and negative associations that people have for different attitude objects, regardless of whether or not the respondents would (or could) consciously voice those attitudes. For this task, voters had to categorize pictures of the base as quickly as possible by saying that it either “was a picture” or “was not a picture” of the base, and then categorize positive or negative words as quickly as possible by saying that they were either “positive” or “negative” words. Next is when it gets tricky and the “implicit attitude” starts to peek out. Voters were shown either a picture or a positive or negative stimulus, and then had to categorize whatever they were shown as “either a picture of the base or a positive word,” or “either not a picture of the base or a negative word” as quickly as possible. During the next round, the pairings were switched. According to the logic of the IAT, if the participants had faster reaction times when the pictures were paired with positive words, that means that the voter has a positive implicit attitude towards the proposition (because it was easier for them to make the categorizations if the military base was paired with positive words); if the participant has a faster reaction time when the pictures were paired with negative words, that implies a negative implicit attitude.1 In other words, it uses reaction times to see if people implicitly associate the military base expansion with “good” or with “bad,” regardless of their consciously expressed attitudes.

What the researchers ended up finding has profound implications for how we think about those “undecided” voters. Although undecided participants expressed no strong conscious beliefs about the issue when they were first asked (hence their being “undecided”), they did show differences in their automatic associations – and these implicit attitudes about the proposition significantly predicted how they actually voted 2 months later. Put another way, even though these participants considered themselves to be undecided in October (and consciously expressed truly undecided attitudes), implicit measures of their unconscious associations indicated that they weren’t quite so undecided after all — in fact, these implicit attitudes were meaningful enough that they significantly predicted the actual voting choices that the participants made two months later. In US-Election terms, that’s like saying that even if someone believed that he was truly “undecided” back in September, an implicit measurement of his positive and negative associations with the words “Romney” and “Obama” might have predicted how he will actually end up voting tomorrow.

For self-professed decided voters, on the other hand, these automatic associations had no such predictive power. Instead, the only thing that reliably predicted how these participants voted were the actual, conscious attitudes that they expressed at the first time point. This would be like someone saying that an explicit endorsement of Obama in September is a significant predictor of that person voting for Obama in November. Not nearly as interesting.

As we head for the polls tomorrow, I’m sure we all hope that every voter has made up his or her mind by now. However, even those recently-decided voters might not have been truly undecided two months ago. The choices that voters make tomorrow might actually have been “hidden” in their automatic associations all along.

1. For more on the Implicit Association Task and implicit attitudes, you can visit the Project Implicit website or the Wikipedia page about the IAT.
2. Figures in this post are adapted from Galdi, Arcuri, & Gawronski, 2008.

Galdi, S., Arcuri, L., & Gawronski, B. (2008). Automatic mental associations predict future choices of undecided decision-makers. Science, 321, 1100-1102


2 responses to “Can we predict the “undecided voters”?

  1. Really interesting, as always! Ran into something yesterday discussing the “undecideds”…questioning that those that identified as such were actually undecided, or just chose not to publicly commit. The interesting story was in those that claimed commitment to one candidate but swapped in the face of new information. The argument was that these were the people to target, not the “undecideds”.

  2. Though obviously that refers to the difference between “declared undecided” and “actual undecided”. (I posted that link mostly because I thought you’d find it interesting, rather than directly relating to your post.) I suspect that many actual undecideds have strong feelings, but are conflicted. For example, black pastors that are upset about Obama’s gay marriage stance. Or people who aren’t thrilled with how the last 4 years have treated them financially, but hate Romney’s approach to women’s rights. I wonder how those would fit in the models you’ve discussed here, since “base” is a single issue, and “Obama” or “Romney” might invoke a lot of complex emotions, both positive and negative.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s