My post last week Polling as a Commodity in a Saturated Market generated some interesting comments in the 140 character world of Twitter. I’d like to indulge in a few extra characters for some of the issues that arose. Important questions arose about whether anyone outside the community of political junkies even notices polls, whether polling influences or simply measures attitudes, and polling and citizen engagement.
Yes I responded and offered three pieces of evidence, first an academic paper on polling effects, second the argument that if no one pays attention to polls it would be an expensive and irrational business decision for profit-seeking media to run them, third a study by the Boston University College of Communication’s Project for Excellence in Journalism on polling effects on media in the Scott Brown-Martha Coakley race.
The paper I cited was Philipp Denter and Dana Sisak, “Do Polls Create Momentum in Political Competition?” The core of the paper is focused upon whether or not polls induce different candidate strategies and thus may affect outcomes but the authors accept that polling does influence voting behavior as well. They provide a literature review on both the bandwagon effect theory and the mobilization effect theory. The bandwagon effect literature shows how polls can distort voters’ decisions in favor of the leading candidate by means of information cascades (more on this when we get to the BU study), desire to vote for the winner, or voters’ aversion to uncertainty. Interestingly enough other scholars argue that there may be a negative mobilization effect that harms the candidate leading in the polls. As to the question at the heart of the Denter and Sisak study, they find that “polls generally cause momentum for the front-runner. The reason is that the front-runner has an additional incentive to invest in her campaign. By investing early it becomes more likely that she comes out ahead in the poll in the future. This in turn defuses competition and thus she can save on campaigning expenditures closer to the election. In closer campaigns, when campaigning is very effective, also the trailing candidate may adopt a tough stance and outspend the leader.”
For the purposes of this argument I offer a telling insight from Denter and Sisak:
While an informed electorate is generally considered essential for a well-functioning democracy, one exception concerns polls for candidates’ relative standing. Critics claim that polls undermine both the incentive to vote as well as the vote itself; thus distorting voting decisions. As a consequence the preferences of the populace are warped by the echo chamber of opinion polls. For this reason, many countries have imposed a ban on the publication of pre-election polling results.
As for my second reason, I fully respect the argument that in providing poll coverage media institutions are providing information of interests to their consumers. They are also businesses and polling is expensive. It would seem to me to be irrational to repeatedly engage in a practice that is economically harmful to the business. Perhaps I am wrong.
Finally regarding Mr. Bernstein’s question, on to the effect of polls and the coverage they engender. The Boston University College of Communication Project for Excellence in Journalism published an interesting report on the 2010 senate special election, “Hiding in Plain Sight, From Kennedy to Brown: Press Coverage of the 2010 Massachusetts Senate Special Election,” honing in on the period of Jan. 6-19, after a Rasmussen Poll that showed Brown within nine points of Coakley. After that poll, coverage of the race exploded. “Coakley went from being portrayed as a cautious but competent and clean politician to an incompetent bumbler. . . . Brown went from receiving a polite but dismissive portrayal of a good looking non-contender to a surging populist star, with positive stories in the general election outweighing negative ones by more than 2-1.” This would be an example of the cascade effect Denter and Sisak mention. As John Sides and Lynn Vavreck write in their compelling study of the 2012 presidential election The Gamble, candidates attempt to generate coverage for a good reason: what is prominent in the media influences their electoral prospects with measurable effects.
Mr. Walsh’s pollster friends are right in a narrow sense but underestimate their influence in a broader sense. A poll could be conducted and merely placed in a file and that measurement would have no effect, but that is not what is done with polls. As much as the Denter and Sisak paper says about polling effects on the public, the compelling aspect of the paper was that it sought to measure effects of polls that are not made public. Rather the paper studied the influence of polling on candidate’s spending decisions, which in turn have an impact on the vote. “[A] ban on the publication of opinion polls, as many countries impose in the pre-election period, will likely not eliminate all effects on voters. … there exists an indirect effect of polls on candidates’ campaigning investments, which in turn influence voters decisions at the ballot, and which is still in place because candidates are still allowed to commission polls in that period.” Of course, this is exactly why campaigns pay pollsters.
So in answer to the Serious Q, yes I would suggest that there is evidence that people other than political junkies notice polls.
I attempted to answer the first question about late polls affecting the Coakley race in the comments section to Polling as a Commodity in a Saturated Market so go there if you want the full response. So let me take a stab at “Is info from other voters an acceptable input to decisions?”
I started down the road of questioning the ubiquitous authority of polls from re-reading essays by the late Wilson Carey McWilliams; how I wish Professor McWilliams was here to answer this one. He was concerned with the soul of republicanism and the endangered place of the citizen’s voice. Like just about everyone who studies politics he read and cited survey research but his engagement was more humanistic than mathematical.
I think Prof. McWilliams would want voters to have information from other voters in making their decisions, but I question whether he would think of polling as the kind of conversation that should be championed. He often criticized the role of media in denying the citizen the ability to engage – media presentations are one way conversations. The presentation of a public poll is often in the form of a summary of what X number of citizens had to say in response to questions pre-selected by pollsters, with limited options for even open-ended answers and no opportunity for exchange among citizens. They are attended by an overlay of scientific precision that makes the capacity of a citizen to engage with them even more remote. To repeat the opening sentence of the Denter and Sisak paper, “While an informed electorate is generally considered essential for a well-functioning democracy, one exception concerns polls for candidates’ relative standing.”
By the way the notion of the difficulty of fostering citizen engagement is at the heart of the defense of party caucuses, conventions and the fifteen percent rule offered by Professor Ubertaccio and me. We believe that “the caucus system contributes to a system in which local voices matter.” At least a slice of the citizenry directly engages with candidates and have real face to face discussions with each other. Understanding is fostered and differences confronted. It is in such small local gatherings, Tocqueville recognized, that the individual learns to master the tools of engagement as a citizen.
Finally, in Polling as a Commodity in a Saturated Market I wrote this hyperbolic statement about polling in the primaries: “Warning: Following information has no predictive value.” Like most hyperbole that remark couldn’t be considered strictly accurate, either. Perhaps a caution like this would be more suitable: “Warning: polls are less predictive the further from Election Day, and primaries are trickier to poll than general elections.’” Something like that. For more on polling accuracy, see Nate Silver at FiveThirtyEight Politics, Is the Polling Industry in Stasis or in Crisis? or Hans Noel at MischiefsofFaction, Stop Trying to Predict the Future for more of the academic nerd fight over forecasting models and polling.
I’m glad for the commentary on my prior posts. Each horse race story in the next several weeks will garner the attention of the “poli junkies” and I think the research shows, a much wider swath of the electorate. It is worth reflecting on whether this preoccupation deserves its prominent status in our democratic discourse.