These days, electability is the talk of the town - if the town is Des Moines or Cedar Rapids. This is from yesterday's Washington Post:
Strategists for Obama said over the weekend that they see an opening for their candidate on the question of electability, and campaign manager David Plouffe also predicted a "relentlessly negative" barrage from the Clinton campaign in the days ahead.
Central to the new Clinton push will be the argument that only she can beat the eventual Republican nominee, a claim Obama is also seeking to make to voters here.
Advisers said her message will be: "You can't have change if you don't win." Her rivals, meanwhile, are moving aggressively to capitalize on Clinton's weaknesses in Iowa -- and, they hope, block her path to the nomination. [Snip]
"We're picking up a lot more on the ground on electability," Plouffe said. "What voters are looking at is: Who's got the best chance to win the election . . . and second, who can govern."
Also, Jason Zengerle offered an extended meditation in New York Magazine yesterday about electability and the 2008 primary contests. His conclusion is that the whole discussion is nothing more than a maddening puzzle with no solution. He writes:
Unlike political judgments that are based on concrete assessments of, say, a candidate's record or even something as grubby as his fund-raising prowess, those that are based on a candidate's supposed electability can change on a moment's notice. And then change again. Electability is completely ephemeral. Even those hoary and maddeningly indefinable political qualities like "character" and "authenticity" have more meat on their bones.
I think that Zengerle is on to something here - but like so many articles I read in Esquire, New York Magazine, the New Yorker or other such "upscale" mags where writers get 4,000 or more words to work with - he just ends up repeating the same basic point with different anecdotes. Zengerle never gets beyond the fact that electability is "squishy." Why is it squishy? Or, more specifically, why is it a squishy concept that we can't help but talk about? That's the question I would like to tackle today.
From a rational choice perspective - electability should be a critical factor in a voter's preference for the party's nomination. This is because winning the nomination brings about no changes in public policy. Only winning the general election will bring about such changes. So, a purely rational voter who is only concerned about public policy would have to calculate the expected policy benefits from a given candidate's election in something like the following manner:
Ideological Proximity of Candidate and Voter X Candidate's Chance of Winning Nomination X Candidate's Chance of Winning General Election
The latter two terms would combine to be a candidate's electability factor. So, electability is an important feature in the purely rational voter's preference for the nomination (although the purely rational actor might not even vote). We would call a calculation like this part of the act of strategic voting because it implicitly takes into account the choices that you expect other voters to make.
However, we confront here the problems that plagues all of us who make use of rational choice theory: people are not as good at acting "rationally" as mathematicians are at writing about it. Things get much more difficult when we are dealing with our subjective experience, and not the objectivity of a formal proof. In the real world, we might expect some interdependence between electability and ideological proximity. For instance, we might expect proximity to influence a voter's assessment of electability because, as everybody knows, what we want often gets in the way of our understanding what others want. We might also expect the causal arrow to run in the other direction: what others want can influence what we want. That would be a bandwagon effect. Princeton's Larry Bartels finds something like that in his Presidential Primaries and the Dynamics of Public Choice.
My intuition about electability is that it: (a) is an inevitable subject in a presidential primary campaign because the nomination is not an end, only a means; (b) contains some objective elements that would enable us to estimate who is more electable than who; (c) contains some subjective elements that would impede such estimates. This would explain why the conversation about it has been endless and fruitless. We can't help but discuss it because it is, after all, what a party's nomination is all about. We're also compelled to talk about it because there is something "real" to it - and so we are tantalized at the prospect of having an objective statement of who stands a better chance of winning. At the same time, we're confounded because, while there is something objective here, the subjectivity of it prevents any real consensus. Thus - we're left to debate endlessly (or, at least until the election!).
So, I have three basic assertions about electability: (a) inevitable, (b) partially objective, (c) partially subjective. Let me try to convince you that I am on to something here - for there is good evidence for each point. I think that (a) is essentially established. Common sense, rational choice theory, and polling data that shows voters are thinking about electability all conspire to put to rest any doubts of its importance. But what about (b) and (c)? Can we "unpack" the different elements that make up our conceptions of electability to see what in them is objective and what is subjective?
I think we can - but we'll have to do something that Zengerle does not: treat the work of political science with some care. He writes:
Political scientists, for their part, have taken a stab at gauging electability. At conferences and in journal articles, these academics have tried to suss out just what makes a candidate electable--or, at the very least, what makes voters perceive a candidate as electable. Physically attractive male candidates, one group of political scientists concluded, enjoy an electability advantage, but the same group could find no straightforward link between attractiveness and electability for female candidates. Similarly, some political scientists found that male politicians with hair tend to get elected more than balding ones. (There doesn't appear to be any research on the electability of balding female candidates.) And in 1994, two professors from William & Mary and the University of Colorado went so far as to come up with an actual electability formula, which looks like this:
Candidate electability = a + b1 (party) + b2 (evaluation of C) + b3 (C's proximity to R) + b4 (C's proximity to the average voter) + b5 (C's proximity to party) + b6 (C's nomination chances) + b7 (C's TV performance) + e
In my margin notes, I wrote next to this paragraph simply "Boo!" Maybe I am over-sensitive to the way that the work of political science is ignored by the punditocracy - but this seems to me to be downright rude. If you think enough of somebody's work to cite the work, you should also cite the somebody. Let me correct the oversight, and give some credit where it is due. The researchers who developed this idea are Walter J. Stone (formerly of the University of Colorado, now of UC-Irvine) and Ronald B. Rapoport of the College of William and Mary.
I also have to object to the obvious carelessness with which Zengerle read this article. It seems to me that he trolled it for a quickie rhetorical point - for he brings up this "formula" simply to compare its seeming over-complexity to the "straightforward approach" that pundits take. In so doing, he completely misunderstands what Stone and Rapoport are on about.
The point of Stone and Rapoport's equation is not an "electability formula" in the sense that Zengerle means it. This "formula" is an ordinary least squares (OLS) regression equation. OLS regression can be used for predictive purposes - a "formula." But - Stone and Rapoport would surely agree that their formula does a lousy job of predicting how electable a given voter views a given candidate. Their model has an accuracy rate of 25% to 57%. [When political scientists do use OLS regression to predict outcomes - say for congressional elections - they look for an accuracy rate of 80% to 90%].
OLS regression has another purpose, which is what they are actually utilizing in this article and which should not be characterized as the development of a "formula." What Stone and Rapoport do here is test whether some variables are significant causal factors in how voters perceive electability. The idea behind this equation is that you compare (for instance) a voter's evaluation of a candidate's television performance to a voter's evaluation of a candidate's electability. Are they related - and, if so, how strongly are they related? But you do this comparison in relation to comparisons between electability and the rest of the variables on the right hand side of the equation. That way, you can conclude whether any one of these factors influence electability controlling for the other factors. OLS regression, then, enables one to draw causal inferences in complicated situations where many variables interact with one another. This is what Stone and Rapoport are on about here.
In fact, they have a very specific question in mind. Do voters form assessments of electability based upon their perception of candidate moderation? As a candidate appears to be better able to appeal to the middle of the general electorate, will a primary voter view him as more electable? This question has great analytical importance, for obvious reasons. But it is also of civic importance as well. It would not be good for the country if small, unrepresentative primary electorates forced upon the broader public ideologically extreme nominees. So, Stone and Rapoport are interested in whether caucus goers in 1980 and 1984 perceived a link between ideological moderation and electability.
They do indeed find that proximity to the average voter is a factor in electability - but it is not as big a factor as other characteristics. They find that performance on television is a much stronger factor. That is, if we think a candidate performs well on television, we will be more likely to see him as being electable. Other factors influence perceptions of candidate electability more than ideological moderation: evaluations of candidates (the more we like a candidate, the more electable he will seem to us), nomination chances (the more we think a candidate will win the nomination, the more electable he will seem to us), and party (we are more likely to see candidates of our party as electable).
If we look at the list of variables that influence electability - we should appreciate the mixed bag that "electability" presents. On the one hand, there is a lot of subjectivity here. For instance, partisanship is a factor in evaluations of electability. Our perceptions of electability have a partisan skew to them. So also is our impression of the candidate: if we like him, we believe other people will, too. That is about as subjective as you can get! With electability being informed by such subjective evaluations - it is unsurprising that there is such disagreement among everybody about who is electable and who is not.
But, on the other hand, there is also some objectivity here, which I think helps explain why pundits (and campaign strategists) so frequently try to quantify it. For instance, television is a factor in perceptions of electability. As television is a shared experience - it probably gives the concept of electability some objective leverage. If I think a candidate does well on television, other people probably will, too. Also, as a candidate must be nominated to be elected - the fact that the chances of being nominated is a factor in electability makes it somewhat objective for a similar reason: all of us can agree on who could and who could not be nominated.
Relatedly, Stone and Rapoport found some rough parity in ideological placements of candidates among Republican and Democratic caucus goers. For instance, in 1984 Democratic caucus goers placed Gary Hart closer to the ideological center than Republicans. However, both Republicans and Democrats identified Hart as being more liberal than the average voter and more conservative than Walter Mondale. The fact that ideological positioning is a factor in electability therefore infuses the latter with an objective element because we seem to view ideology in roughly similar terms.
So, maybe we have hit upon an explanation for what Zengerle has observed: there is a debate about electability that seems to have no end. The importance of electability and the objective elements of electability compel people to debate it. There seems to be something more than just our subjective evaluations that inform the concept. However, the concept itself is still so subjective that it is quite unlikely that we could ever agree.
All of this also means that perceptions of electability are probably manipulable. This would explain why candidates - and not just analysts - are participating in the conversation. These strategy memos are not designed to elucidate the true state of the race for us. They are designed to do what all campaign actions are designed to do: persuade voters. Candidates have an incentive in creating the impression that they are electable - and the subjectivity of electability means that they might be able to create that impression.
-Jay Cost