The Palin Effect

As attentive readers of the New York Times are aware, science bloggers are all about being noisy and shouting at people we don't like. As most of us have liberal leanings, that means we can be obnoxious towards people on the political right. And this week we've got great fodder, in the shape of a paper about Sarah Palin.

Sarah Palin addressing the Republican National Convention. Go on, wave back

Actually, it's even better than that. It's also about George W. Bush.

President George W. Bush speaks at Coast Guard commencement, 2007

This paper takes a look at the effects of the different players in the 2008 US election: Obama and McCain (the presidential candidates), the vice-presidential candidates, Sarah Palin and, what's his name, um, err, Joe Biden. yes, that's who it was. and finally the then-president, George W. Bush. The authors, political scientists at Standford and Duke Universities, use panel data to see how everyone fared through the campaign.

Panel data is simply data when a group of people - the panel - are asked the same question at different times. In this case, the questions were first asked in November 2007 (i.e. a year before the election, and then on 9 subsequent occasions, the last one being just after the election (in November 2008). On each occasion, the panelists were asked "For each of the following individuals, please select if you have a favorable or unfavorable impression of that person. If you don’t know enough about the person to have an opinion, you can say that too." They had a choice of answers: "Very Favorable", "Somewhat Favorable", "Somewhat Unfavorable", or "Very Unfavorable", along with a neutral category "No Opinion/Don’t Know Enough to Say". Over 1000 respondents answered every time.

This is all standard stuff. One big advantage of panel data is that we can see how individuals change opinions over time: normal polls can't show who is shifting opinion in what direction, only that the overall numbers change. This means we can see the movements of people from one candidate to another, and how (eventually) they voted.

We can see the overall shifts in the figure below:

Changes in Candidate Favorability During 2008 US Presidential Election

(The reason Palin and Biden only start in Sept 2008 is that they were only publically announced as VP candidates then). There is some movement for most candidates, but the things that stand out are (a) G.W. Bush's figures hardly move. I guess this is not surprising: everyone knew what he was like, and most had formed their own opinion of him, (b) for the other candidates the number of don't knows drops over time: people are learning about the candidates, and (c) Palin's large increase in unfavourables. The latter point mimics the narrative for her election: she was initially popular, until she started giving interviews. Then people saw what she was like.

Perhaps the most important question is how these changes affected voting patterns. From the panel data, the authors could look at how preference for a candidate affected the decision about who to vote for. They present their results as the difference in probability for voting for Obama between two identical people, both of whom start off as undecided but one who changes to a "somewhat unfavorable" opinion of Obama, and the other to a "somewhat favorable" opinion at the survey time. this is what they get:

Predicted effects of candidate evaluations on the probability of transitioning from undecided to Obama.

As we might expect, the closer to election day the greater the difference: there isn't the time to have one's views changed. But it's also worth noting that Joe Biden has a much smaller effect on voting patterns (and Bush none at all). Sarah Palin's effect is almost as large as the presidential candidates. the received wisdom is that VP candidates have very little effect. Our maverick politician certainly went rogue with that.

The authors estimate that, overall, the effect of Palin was to lose McCain 1.6% of the vote. This wouldn't have changed the overall result (Obama won 53% of the total votes cast). Looking at the tables, she seems to have roughly the same size of effect as McCain.

The guy I really feel sorry for, though, is Joe Biden. In a paper about the 2008 election, he's ignored. This is the number of times the following appear:

Obama: 51
McCain: 63
Palin: 67
Biden: 21
Bush: 59
Undecided: 32

Not only does Bush (who wasn't even in the election) get almost 3 times as many mentions, Biden is even beaten by the undecideds! I'm not sure Joe Biden minds though. He still has his place in history.

Inside drawer of Senate chamber desk XCI occupied by Sen. Joe Biden during his time in the U.S. Senate.

Citation

Elis, R., Hillygus, D., & Nie, N. (2010). The dynamics of candidate evaluations and vote choice in 2008: looking to the past or future? Electoral Studies DOI: 10.1016/j.electstud.2010.04.005