Chris Good at the Atlantic Monthly expands on Conor Clarke’s case (discussed by John below) against opinion polls.
At the Atlantic Special Ideas Report, Conor Clarke makes a case against polling, for, among other reasons, polls’ ability to influence mass opinion by reflecting it, accurately or inaccurately, and to effect a herd mentality: … Polls illustrate, in other words, the power of perception in politics. Fatigue sets in when one’s favorite candidate is down. If he’s down ten points, why even take the trouble of voting? In that sense, polls corrupt the experiment of an election by suggesting results beforehand; for all the statistical science that goes into them, they’re fundamentally anti-scientific. And perhaps more significantly, Conor suggests, people can change their opinions to side with the front-runner.
I’m not at all sure that I buy ‘they’re fundamentally anti-scientific’ bit – elections, despite Good’s metaphor, are not experiments, and are not supposed to be. But Good’s underlying animus is, I suspect, a reasonably common one. Good believes that voters should vote on the basis of their actual preferences for candidates, rather than their perception of whether other voters support the candidate in question. Opinion polls – because they provide evidence of others’ voting intentions – make voters more likely to be swayed into voting for candidates who would not be their first choice in an ideal world.In the language of political science, voters should vote sincerely rather than strategically.
This is not a ridiculous position to hold – as the social choice literature makes clear, strategic voting can lead to all sorts of indeterminacies, possibilities for manipulation via agenda control etc. And that’s not even to get into the social psychological literature on tendencies to group conformism etc. But there is a normative case to be made for strategic voting too. Sunshine Hillygus has a nice article on strategic voting in the 2000 Presidential election. She finds strong evidence that supporters of minority parties (in this case, Ralph Nader) are more likely to vote strategically, changing their vote to support the candidate closest to their own views (for the most part, Gore) who has a decent shot at winning. High information voters (as best as the data indicates) were more likely to switch support from Nader to Gore, as were voters in swing states. But clearly, a number of people did not vote strategically (according to the usual political science definition of the term) in this election, since Nader had no realistic chance of winning, instead expressing their sincere preferences for a no-hoper candidate.
those supporting Nader for expressive reasons were undeterred by the wasted vote appeal, and were the most likely to remain loyal. These voters might not be voting so as to alter the outcome but instead to send a message or signal. For these Nader supporters, they might consider their vote to be much like cheering for their favourite team in the stadium. If a voter views the vote as an end in itself, then there is little reason to respond strategically to the electoral environment.
Nader’s vote, as is well known, collapsed in the 2004 presidential election. There are many possible explanations for this collapse. But casual empiricism would suggest that one plausible reason why left-leaning 2000 Nader voters were unlikely to vote for him again in 2004 was that they felt they were badly burned by the 2000 experience – what had seemed like a good opportunity to signal their sincere preferences turned out to have momentous consequences for the governance of the country. This is, of course, highly unlikely to have been the only reason for the precipitous drop in support for Nader, but it does suggest that sincere voting may have its downside in a winner-take-all electoral system. And opinion polls – by providing information on who is likely to win or to lose (especially in multiple candidate primaries and the like) can allow voters to cast their votes in more efficacious ways.
{ 10 comments }
Actually the social choice literature makes clear that sophisticated (“strategic”) voting restricts the possibilities for agenda manipulation compared to sincere voting. Compare McKelvey and Schofield (sincere) to Miller and Banks (sophisticated).
Yep – you’re right – my fault for writing in a hurry late at night …
Sure, there’s nothing wrong with voting strategically, though I would say that there is something wrong with voting systems that promote strategic over sincere voting, eg first past the post.
On a seperate point, what about the possibility that in 2000 some people voted strategically for Nader – i.e. they didn’t want Nader to rule the world but did feel that the Democrats had drifted further to the right than they would like? One’s sincere preference could hae been Gore over Nader, but one might vote strategically for Nader (especially perhaps in safely Democratic states) in the hopes of pushing the Dems leftwards (though presumably not as far left as Nader).
James, while FPTP promotes some strategic voting, I’m not as sure (though there likely is deeper literature on this) that PR doesn’t also promote strategic voting. If your ideal party is not one of the larger 2 parties, then you may, in fact, have an incentive to vote for one of the larger ones instead of your true preference. A larger coalition partner has more power to gum up the works than a smaller one, and the big party that does for the government might choose to do so with a coalition of smaller parties rather than with just a few, because then no one party could hold the government hostage with a threat to leave the coalition.
Granted, this is likely less than the strategic incentives in FPTP systems, but I’m not sure that the general election is the big concern in either system. Primaries seem to pose the bigger problem vis-a-vis polling (to me), and where I would disagree with the Clarke/Good argument is that people will still factor electability into their votes, they’d just have less information to do it with.
I’d like to make a point about the argument that election’s are an experiment.
Well, Although not ortodoxy, that would be the opinion of Hayek, for example, who believed that competition (political or of economy) is a mean to discover things we don’t know in advance.
If an election is a mean to discover people’s will, then everything that interfere with this is disturbing the purpose of an election.
So, in your example of strategic voting, the election was discovering people (strategic) will (to send a signal to dems).
However, if people really do not go to vote because of polls, then the election is not discovering people will!
I think this is a differente point!
> though I would say that there is something wrong with voting systems that promote strategic over sincere voting
Um…so my friends Gibbard and Satterthwaite ensure that by this standard there’s something wrong with pretty much all of them.
not the lottery vote!
A good place to start delving into that “deeper literature” that Matt alludes to would be the new Grofman, Blais, and Bowler volume. I recall (from the conferences, which I attended, though my contribution went into a different volume) that several of the papers show that the extent to which there is actual strategic voting under plurality-rule systems is much exaggerated.
Maybe there are reasons why it is higher in the US. But I suspect there are just fewer voters who actually favor Nader, Greens, Libertarians, etc., than favor the NDP, LibDems, etc. In fact, that seems like a no-brainer!
As for Nader’s 2004 vote, its collapse compared to 2000 presumably had something to do with his not being on the ballot in several states (including some of his stronger 2000 states), as well as with the deeply divided Greens having nominated a different candidate. How much it was factors such as these and how much it was updated strategic information from 2000 (and updated information on just who this Bush fellow really was) is hard to disentangle.
(And James pretty much sums up how I felt back in 2000!)
[My first attempt to post this two days ago failed; apologies if it ultimately shows twice.]
A good place to start delving into that “deeper literature” that Matt alludes to would be the new Grofman, Blais, and Bowler volume. I recall (from the conferences, which I attended, though my contribution went into a different volume) that several of the papers show that the extent to which there is actual strategic voting under plurality-rule systems is much exaggerated.
Maybe there are reasons why it is higher in the US. But I suspect there are just fewer voters who actually favor Nader, Greens, Libertarians, etc., than favor the NDP, LibDems, etc. In fact, that seems like a no-brainer!
As for Nader’s 2004 vote, its collapse compared to 2000 presumably had something to do with his not being on the ballot in several states (including some of his stronger 2000 states), as well as with the deeply divided Greens having nominated a different candidate. How much it was factors such as these and how much it was updated strategic information from 2000 (and updated information on just who this Bush fellow really was) is hard to disentangle.
(And James pretty much sums up how I felt back in 2000!)
Oh, here is the link to the Grofman, Blais, and Bowler volume (in case that first attempt, with two embedded links, never appears).
Comments on this entry are closed.