Election Expectations, PollyVote, and the Charismometer

In a previous post, I wrote that undecided voters may be influenced by a less-obvious bit of persuasion - the momentum of a narrative, the apparent shift in a candidate's prospects from one day to the next, the poll-based stories that no news team can resist. Given a choice, all other things apparently being equal, who wouldn't want to back a winner?

I had this in mind when I came across a new study by Andreas Graefe, a German forecasting guru who found that one tool for predicting election outcomes has been largely overlooked in recent decades, despite it's astounding accuracy: the voter expectation survey. Rather than asking people who they intend to vote for, these surveys ask, "Who do you think is going to win?" Graefe looked at 190 of them conducted between 1932 and 2008 and found them to be 91 percent accurate. Looking at the six elections between 1988 to 2008, voter expectation surveys did better at predicting the winner than polls, prediction markets, econometric models, and the judgment of pooled experts).

Despite this record, Graefe doesn't think we should neglect those other forecasting methods. Better to combine them all into one -- thereby canceling out the biases and shortcomings of each. He and some colleagues have done this with an election forecasting model they call PollyVote. In a 2011 study of the model, Graefe and colleagues looked at daily forecasts for 100 days prior to each of the five elections between 1992 and 2008 made by polls, econometric models, the Iowa Electronic Markets (prediction market established at the University of Iowa), and PollyVote. On average, PollyVote reduced the error in forecasts at least 30 percent, except when compared to the 7-day average of the IEM, where it did about 8 percent worse. However, PollyVote did better than IEM in daily predictions of who would win: PollyVote predicted the winner on 96 of the 100 days, on average, compared to 76 by IEM. In case you're wondering, PollyVote has projected Barack Obama winning at least 51 percent of the popular vote every day this year and never gone above 52.5 percent. Try to make a narrative out of that. Boring!

If you've slogged this far, here's your video reward, featuring John Antonakis, a psychologist and professor of organizational behavior at the University of Lausanne in Switzerland. He is predicting the presidential election with something he calls the charismometer. Simplified, it is the index of "Charismatic Leadership Tactics" (CLTs) used by the presidential candidates in their nomination acceptance speeches. There are 12 CLTs, such as using metaphors, drawing contrasts, asking rhetorical questions, and utilizing gestures. And, according to Antonakis, this is precisely the sort of election -- where the economic picture is muddled and voters aren't sure whether to punish or reward the incumbent party -- when charisma really matters. According to the charismometer, Obama will prevail. I'll let Antonakis explain:

Bragging Season

There's been some nice early coverage of Mind Over Mind, including a stellar review by renowned psychologist Irving Kirsch in New Scientist magazine, and a short feature on the book in O, The Oprah Magazine by Emma Haak. Also, many thanks to Jeff Glor and CBS News Online for the Author Talk Q&A!

Voting for a Narrative

It's not news that campaigns try to win debates by lowering the expectations for their candidate's performance to the point where sentience and continence while on stage counts as victory. Here's The Daily Show's take on the strategy.

The Daily Show with Jon Stewart	Mon - Thurs 11p / 10c
Democalypse 2012 - Negate Expectations - The Presidential Debates
www.thedailyshow.com
Daily Show Full Episodes Political Humor & Satire Blog The Daily Show on Facebook

It's such a well-worn ploy, and so easily mocked, that one wonders who remains to be taken in by  pre-debate expectations limbo. Certainly, it's not the more than 90 percent of the electorate who has already made up their minds. They are unlikely to be swayed by the debates. And I don't really think it's the much sought after undecided voters, not directly anyway. They just aren't paying enough attention. They may watch the debates themselves, but most of aren't tuning in for all the pre-debate punditry on the Sunday talk shows.

No, the expectations gambit is aimed squarely at one audience: the media. It's a solipsistic little game with big potential impact on the election. Because the news, particularly political campaign coverage, lives and dies on narratives, especially "horse race" narratives of who's gaining and who's losing. Among the forces that remain for influencing undecided votes in one direction or another, the momentum of a good narrative is among the most powerful, and the most superficial. As New York Times columnist Ross Douthat points out in "The Media Bias That Matters":

"As a presidential candidate part of your job is to be aware of how easily the horse race narrative can overwhelm whatever story you want the country to be hearing, and to do everything in your power to actively shape a narrative that will inevitably be shaped by the press’s zeal for “who’s up/who’s down” reportage as well."

On Needles and Nothing

As reported yesterday by Amanda Gardner at Health.Com (and re-posted on CNN.Com), "Acupuncture Works, One Way or Another." The article reports on a recently published meta-analysis of 29 studies comparing pain relief from real acupuncture to placebo acupuncture (usually needles that are placed outside of known pressure points) to no treatment.

The common measure of efficacy for such studies is a 50 percent reduction in pain, as measured on a 100 point scale (dropping from 80 to 40 for example). Accordingly, real acupuncture achieved this rate of pain reduction 50 percent of the time, placebo acupuncture met this mark 43 percent of the time, and people who weren't needled at all felt this level of pain relief 30 percent of the time.

This is better than acupuncture's usual showing. In many studies, it does no better than placebo. On the other hand, about two thirds of the difference in pain relief between no treatment and acupuncture is accounted for by placebo.

This likely isn't the final word on the efficacy of acupuncture. But I'd like to bring up one other interesting finding about the treatment. A few years ago, the German health authorities (the volks who decide what treatments will be covered under their national healthcare plan) compared acupuncture to sham acupuncture and to several other known (and already covered) treatments for chronic lower back pain, including medication and physical therapy. Acupuncture did no better than sham acupuncture in these trials. But both real and sham acupuncture beat the therapies that were already approved!

Subliminal Placebos!

We don't need a doctor's suggestion to trigger placebo effects. Maybe we don't even need to think, not consciously anyway, according to a just-released study in the Proceedings of the National Academy of Sciences.

Alongside studies of verbal-suggestion ("this will dull your pain" etc), much research has looked into conditioned placebos -- in which people learn to associate some inactive stimulus (a medicine smell, for instance, or a sweet taste) paired with active drugs (much in the same way bells were paired with food in Pavlov's dog experiments). Soon enough, the medicine smell or the sweet taste by itself is able to have some of the same effects as the active drug.

In pain research, placebos and nocebos (the expectation of more pain), scientists usually condition people with a visual cue -- a specific color or shape paired with more or less pain. In this new pain study, however, both placebo and nocebo effects were triggered with images that flashed by the subjects so fast they could not detect them consciously. The researchers, led by Karin Jensen, a psychiatrist at Massachusetts General Hospital, were affiliated with the new Program in Placebo Studies at Boston's Beth Israel Hospital.

Here's how it worked: At first, subjects' arms were singed repeatedly with high or low heat pain, while viewing one of two faces (shown below).  The subjects were asked to rate each burning pain from 0 to 100. The mean high heat score was 63, while the mean low heat score was 24.

HIGH HEAT

LOW HEAT


[From Jensen et al., (2012) PNAS, "Nonconscious Activation of Placebo and Nocebo Pain Responses"]



In the second round, every heat stimulus was the same, medium intensity, randomly paired with one or another face. As expected, medium heat was felt more keenly (mean 53) when paired with a high heat face compared to a low heat face (mean 19). Finally, subjects sat through one more round of arm singeing paired with faces. This time, however, the faces blinked onto the screen for just 12 milliseconds, well below the threshold for conscious awareness. Still, they got the same results -- a mean pain rating of 44 for subliminal high heat faces and a mean of 25 for subliminal low heat faces.

These findings support the idea that patients' pick up on the subtlest of cues from the context of medical care, and that, in many cases these cues have a medical effect. For instance, consider the possible expectations communicated, without a word being said, by a doctor giving a treatment she is fully confident will work versus one about which she has significant doubts. Whether we're aware of it or not, our brains are constantly learning, and creating links in our minds that can have real effects on our bodies. 

Settle An Argument

Is a positive attitude the key overcoming obstacles and achieving a better life, or is it the sort of magical thinking that leads to burned feet at fire walks and mortgage meltdowns? It can't be both. Can it?

In case you're wavering, here's a recent piece in the Wall Street Journal, "Self Help for Skeptics: Train Your Brain to be Positive, and Feel Happier Every Day: It Only Sounds Corny." And here's one to counter it from a few weeks ago in the New York Times, "The Positive Power of Negative Thinking." 

Admittedly, these two articles aren't really arguing with each other. But, I feel like they are in spirit -- much like articles that talk about digital culture as a boon or a bane to our social lives, education, political discourse, etc. There's legitimate reasons for both, depending on where you look, what your values are, and how you define your terms.

Returning to the question of attitude, among the more interesting studies I've read on optimism was a German study* of positive expectations v. positive fantasies on achieving personal goals (these goals varied from university students aiming for good grades to people desirous of a quick recovery from hip-replacement surgery). The researchers measured the extent to which their subjects had positive expectations or positive fantasies about their goals (way too much to explain in a blog post, but the paper goes into it in depth) and then checked in on their subjects' progress some months later. The results? More positive expectations correlated with more progress and goal attainment. But more positive fantasies correlated with less.

*Link is to an abstract. You can download the full text for free from Google Scholar, but it takes a while.

Feel The Future

That's the title of a forthcoming article in the Journal of Consumer Research that says people who trust their feelings make better predictions than people with a low trust of their feelings about everything from the weather to the winner of American Idol to the stock market's closing numbers a week out to the winner of the BCS (college football) championship bowl game. The high-trust folks also out-predicted control group predictors. The researchers found a similar "emotional oracle effect" across eight experiments, involving about 1,250 participants between 2008 and 2011. Before we get to their theories of why this might be, here's a bit more on the experiments.

In some cases, the researchers simply asked participants to rate how much they trusted their gut feelings when making predictions and then compared the accuracy of the predictions along these lines. Other times, they manipulated that trust with a fairly counter-intuitive little trick. People in the "high trust" condition would be asked to recall two incidences when they made a judgment or decision based on their gut feelings, which turned out to be the right thing to do. People randomized to the "low trust" group were asked to recall ten such situations in which they trusted their gut and it turned out to be right -- a much harder task, which supposedly leads people to infer that feelings aren't super reliable guides.

 
[In the 20 hours between the final performances and the winner's being announced for the 2009 season of American Idol, the researchers asked more than 100 people to predict who would prevail. The degree to which people said they trusted their feelings when making predictions was significantly related to their ability to predict that Kris Allen, featured above, would pull a surprise upset of Adam Lambert.]

In every case, subjects were screened to assure they had a basic level of familiarity with the subject matter they were making predictions about (this background knowledge was very important, as we'll see) and weren't obviously biased (e.g. ardent supporters of either team contending for the BCS title were excluded from the football prediction experiment). Importantly, trusting one's feelings was not a proxy for confidence in one's predictions. The researchers had subjects rate the confidence in their predictions, and it bore no relation to how much they trusted their feelings as guides.

The researchers speculate that our emotions encode and summarize vast amounts of information we accumulate "consciously and unconsciously, about the world around us," and that people who trust their emotions are thereby given a "privileged window" into that huge storehouse of information on which to base their predictions. According to this theory, gut feelings aren't airy notions or premonitions -- they are information packed. The researchers bolster this account by including a few conditions in some of their experiments when subjects are likely to have little or no background knowledge of what they're predicting (e.g. American college students asked to predict the weather in Beijing several days from now). In these cases, the "emotional oracle" effect disappears.