I will freely admit to being a FiveThirtyEight junkie over the past few months. The excellent analysis and projection that they do over there is more than a little reminiscent of my favorite baseball site, Baseball Prospectus, in that we have a very self-conscious quantitative methodology in which numbers are used to produce good insights, and not simply crunched for the sake of crunching numbers. Marvelous stuff, and basically updated in real time — hats off to Al Gore for inventing the InterNet that makes this all possible.
That said, I worry just a bit about this obsessive poll-watching, the same way I worry just a bit about the obsessive sabremetric-ification of the game of baseball. And I worry for the same reason in each case: too much of a focus on individual-level statistics can obscure the broader sweep of the endeavor. Neither a baseball game nor an election are simply the linear sum of individual behaviors; the process is considerably more complicated, since the actual play of the game is qualitatively different from the statistical snapshots that we derive from the game’s playing. This is not to say that such statistics are not useful — far from it, I remain a junkie for both opinion poll and sabremetric data — but simply a caution that we ought not to confuse the map with the territory, the derivative with the product, and the lived experience of the activity with the quantitative analytic that we sometimes use to parse and explain it.
Baseball first. Individual players engage in a series of acts — at-bats, pitches, fielding chances, etc. — over the course of a long season with many games. In the course of doing so, they produce a record, a summation of how they have performed each time they engaged in a given activity over the course of the season (and, ultimately, over a career). On-base percentages, strikeout-to-walk ratios, range factors: these and other quantitative measures give us some sense of how well someone has performed, and baseball being the kind of effectively closed social system that it is, these numbers give us a reasonably reliable basis on which to forecast how a player is likely to perform in the future (even though the actual extrapolation of past performance into the future is a technically complicated activity). Baseball, by design, produces statistical data, and that data is a pretty good way to explain what is going on in the game over the course of a long season — although only over the long-term, because day-to-day averages fluctuate quite a bit.
But baseball statistics are not the game of baseball. I, like many baseball fans, play fantasy baseball, a derived game in which you score points based on how “your” chosen players perform on the field. (I’m far less obsessive a fantasy baseball player than many, mainly because I don’t have time to spend on gathering and evaluating the data that I’d need to have in order to really play well.) In fantasy baseball, the name of the game is baseball statistics, since that’s all you have available to you. It’s kind of like the derivatives market for baseball, but without the opportunity for the actual baseball teams themselves to use it to hedge their investments. And it’s a lot of fun, but baseball statistics are not the activity itself. An individual game is won or lost on a contingent combination of factors, not on someone’s lifetime batting average against lefties in late-inning “pressure” situations. Rather, the play of the game is something qualitatively different from the statistical records that a player accumulates, even if those statistical records can serve as the basis for explaining or even predicting how events will likely turn out. There’s the play of the game, and then there’s the analysis of the play of the game, and the wonderful and majestic thing about the sport of baseball is that the game generates its own data that can be aggregated and then reapplied to the game itself: it’s a recursive system.
What I worry about when it comes to sabremetric analysis is that the beauty of the game in its playing can be lost amidst the quest for better statistical measures. A good baseball analyst avoids this problem by treating the statistics as an analytical tool, and using them to inform case-specific accounts of individual games in all of their glorious and sometimes wacky contingency. But the close connection between the organized sport of baseball and the statistical data it generates — statistical records are derivatives, but they are derivatives of the game itself — means that in the end I worry less about baseball being overwhelmed by its own statistics than I worry about the electoral process being overwhelmed by opinion-polling statistics. People have to keep playing baseball in order to generate statistical data about baseball. Some of the measures constructed out of that data — especially my pet peeves, “batting average with runners in scoring position” for the batter and “wins/losses” for the pitcher — are silly and unrevealing, but they are still reasonably close to the activity itself.
This is not true of public opinion polls. An opinion poll measures the response of likely voters to a question about how they are likely to vote; it asks those individuals to express a preference, to declare an opinion. There is no commitment implied in such an act, and it has few direct consequences, especially since people are free to change their minds as often as they’d like. Opinion-polling aggregates those individual responses, and the percentages we get are a more or less straightforward sum of individual expressed preferences. But an election — that’s qualitatively different from a poll, in that the act of voting is not merely a declaration of an individual preference. To vote is to participate in a common process of determining an outcome that will be binding on everyone; votes have consequences in a way that expressions of preferences do not, and those consequences are general (community-wide) rather than specific (limited to the individual). Voting, in this sense, is not strictly an individual action; it is a collective act of determination by the body politic about where it wants to go. I’m perfectly comfortable with anthropomorphic language to describe this, since a vote — like any other process designed to produce a public affirmation of a course of action — constitutes something like a collective decision in favor of being a certain kind of actor proceeding down a certain kind of road. Public affirmation is binding, and constitutive — it has consequences, both for a community as a whole and for the individuals inhabiting it.
Participating in an opinion poll, on the other hand, does not have these kind of consequences. One can say whatever one wants to say, and nothing is decided about the identity or activities of the community — although in our system of campaigning, partisan advocates might use that data to craft their message differently or to figure out where to best allocate their scarce (or, sometimes, abundant) resources. And even if a governing official “watches the polls” and then uses the poll results to inform her or his decisions, that does not collapse the qualitative distinction between opinion-polls and votes — between the polls and the polling-place — because that official’s paying-attention to the polls is situationally contingent rather than constitutively necessary. I can ignore the opinion polls, but if I ignore the recorded and tabulated vote, that’s another matter entirely; ignoring the opinion polls might affect my popularity rating and my job approval rating, but ignoring the election deprives me of legitimacy. These are not the same thing, conceptually speaking; what individual people (say that they) think is not the same as what people as a group or as a community performatively do when they cast their ballot. Hence, opinion polls are not even a derivative market for votes in the same way that fantasy baseball is a derivatives market for actual baseball.
That said, there is clearly a connection between opinion polls and electoral results. Under many conditions, it is reasonable to make an assumption that people are not deliberately falsifying their expressions of preferences, and it is reasonable to assume that what people say at time t is a good predictor (within a certain degree of statistical accuracy) of what they might say at t+1, and even how they might vote at time t+2. But this does not address the basic conceptual issue, which is that voting is not merely an expression of an opinion. Because of this qualitative difference, because of the gap between the two activities, and because of the difference from a media-marketing perspective between one election and hundreds of opinion polls, I worry more that the distinctiveness of voting will be obscured (or overwhelmed) by the reams of opinion poll analysis, such that we may — and I have heard people refer to elections in this way already — come to think of elections as mere expressions of individual opinions. And in that case we would come to minimize the civic character of the electoral process in favor of a wholly individualistic one. That strikes me as deeply problematic.
I’m all for analyzing polls and trying to predict the results of the election, but I sincerely hope that doing so doesn’t obscure what is truly profound about an election: here are the people collectively determining their identity and their future course of action, as a community and not merely as a collection of self-interested individuals. Once we get past the barrier around a polling-place where electioneering is no longer permitted, we leave mere preferences behind and enter a different kind of space: a space of collective self-determination. That’s what voting is all about, and why it’s so damn important that you get out there and vote NOW if you haven’t already.