Are Voters Different From Non-Voters?

by John Sides on December 1, 2007 · 1 comment

in Political science

Andrew Gelman discusses a paper by Jan Leighley and Jonathan Nagler, in which they find that differences between voters and non-voters in terms of policy preferences are larger than those found in the canonical research of Ray Wolfinger and colleagues, including Steven Rosenstone and Ben Highton. Gelman suggests, “It would be good to resolve the disagreement between the different experts in this area.”

With Jack Citrin and Eric Schickler, I have done some work in this area, and so I can offer at least few thoughts that push toward a resolution:

– If you look at the totality of research in this area, it is clear that there are statistically significant differences between the preferences of voters and non-voters. This is what Jack, Eric, and I found in both of our pieces. Moreover, those differences mirror those found by Leighley and Nagler: on average, non-voters have preferences that are more Democratic and more liberal than voters. – Wolfinger and colleagues are skeptical that any differences would amount to greater support for Democratic candidates. There is merit in this view. In our research, we found that very few races are so competitive that higher turnout would actually lead to a different (simulated) outcome. – Of course, in Leighley and Nagler’s paper, they are more focused on policy preferences than election outcomes. Now, the question becomes: are the differences in policy preferences substantively meaningful? Wolfinger might say that a difference of 5 or 8 or 10 percentage points doesn’t mean much. I am personally less sanguine, especially given evidence that elected leaders better reflect the preferences of voters than of non-voters (see especially this paper by John Griffin and Brian Newman, as well as two papers by Larry Bartels and Marty Gilens, respectively, on the responsiveness of leaders to constituents with higher incomes).

Where does this leave us? With a nascent resolution, I think. Non-voters are more Democratic than voters, though the uncompetitive nature of many elections means that even compulsory turnout wouldn’t produce many more Democratic victories. The policy preferences of non-voters are also more liberal than those of voters, and this may have some consequence for policy (though here I think the research is less well-developed, especially compared to that about election outcomes).

Another consideration here is that social science may not be able to fully comprehend the effects of something like compulsory turnout. How would this kind of shift change the political landscape? We don’t really know. For example, we don’t know if this would affect how candidates and parties position themselves and the kinds of messages they send. It’s possible that the aggregate preferences of the entire electorate might be more different than those of voters today, in part because of campaign effects. And it is also possible that elected leaders would pursue different policies while in office, given the expanded electorate to which they would have to answer once the campaign begins.

If we are to speak to these kinds of possibilities, political scientists need more creative simulations or empirical analyses than we have thus far developed.

{ 1 comment }

Scott McClurg December 2, 2007 at 8:59 am

Excellent post, John. This is really one of the most important and unanswered questions in behavioral research. Two comments.

1. The real difficulty here — which you allude to — is the ceretis paribus condition. All the research in this area simply has to make the assumption that the politics surrounding voting wouldn’t change, which is obviously wrong. Since contextual effects (which this is) tend to have heterogeneous effects, it makes the problem even that more complex. I’d like to see more work that simulates the environmental effects too.

2. Having done a little work on the electoral side of things, its worth pointing out that a lot of work in that areas compares turnout across elections rather than within them. I personally think that’s the wrong comparison. When we talk about the effects of higher turnout, we almost always mean within a single electoral environment rather than between them. That’s especially true if you buy into the contextual argument in my first point. This is one reason to prefer the simulation work in this area over some of the empirical studies.

Comments on this entry are closed.

Previous post:

Next post: