Mandatory Voting Isn’t a Solution to Polarization

William Galston:

The third argument for mandatory voting goes to the heart of our current ills. Our low turnout rate pushes American politics toward increased polarization. The reason is that hard-core partisans are more likely to dominate lower-turnout elections, while those who are less fervent about specific issues and less attached to political organizations tend not to participate at levels proportional to their share of the electorate…A distinctive feature of our constitutional system — elections that are quadrennial for president but biennial for the House of Representatives — magnifies these effects. It’s bad enough that only three-fifths of the electorate turns out to determine the next president, but much worse that only two-fifths of our citizens vote in House elections two years later…But if you think that today’s intensely polarized politics impedes governance and exacerbates mistrust — and that is what most Americans firmly (and in my view rightly) believe — then you should be willing to consider reforms that would strengthen the forces of conciliation.

Jon Bernstein is dubious.  Andy perhaps less so?  Me: I am definitely dubious.  To be clear, I’d be more than happy to give mandatory voting a whirl, but is it a solution to polarization?  Probably not.

A couple pieces of circumstantial evidence, for starters.  First, the period of the highest voter turnout (of those eligible)—the mid- to late- 1800s—coincided with a lot of party polarization.  And the decline in turnout in the first half of the twentieth century occurred during a decline in polarization.  Compare turnout in presidential elections and polarization.  And the increasing turnout in presidential elections during the past decade has coincided with increasing polarization.

Second, although states vary in whether they allow independents to vote in party primaries, more “open” primaries do not tend to produce more moderate members of Congress or less polarized state legislatures.  See this old post.

Okay, but this is just comparing levels of turnout within the limited bounds.  What if we turned turnout up to 11, as it were, by making it mandatory?  Here, studies of the opinions of voters and nonvoters are important.  After all, if requiring habitual nonvoters to vote is to mitigate polarization, nonvoters better have political attitudes that are (1) different than voters and especially (2) more moderate than nonvoters.  Unfortunately, neither is consistently true.

Take the research (pdf) of Jan Leighley and Jonathan Nagler that Andy cites in his post.  Scroll through that pdf file to Table 2 and look at the differences between reported voters and nonvoters in the 2004 American National Election Study.  Consider these survey items and the attendant differences:

  • Ideological self-identification: 42% of nonvoters consider themselves “moderate,” vs. 30% of voters.  Now, a 12-point difference hardly seems to presage the end of polarization, but this is at least supportive of Galston’s conjecture.

  • Should government guarantee jobs: Nonvoters are 12 points more likely than voters to say that the government should do this.  They are more liberal, not more moderate.

  • Should government provide health care: Ditto.  Nonvoters 7 points more likely than voters to give the liberal answer.

  • When should abortion be legal: Nonvoters more likely to say “never”!  They are not consistently more or less likely to give “moderate” opinions.

A parallel data analysis in a different 2004 survey turns up even more modest difference on most items.  And again, non-voters are not consistently more moderate.  In fact, the central finding, according to Leighley and Nagler, is this: “…voters are substantially more conservative than non-voters on class-based issues.”

Part of the problem is that, although Galston is correct that non-voters are not as strongly partisan as voters, they are also less well-educated, less wealthy, and more likely to be an ethnic minority.  Which is to say, they are more likely to want the government to do stuff.  And given that the parties are increasingly polarized on precisely this question—what should the government do?—bringing non-voters to the polls is not an obvious recipe for less polarization.

One final thing: the central effect of political campaigns—one identified in over 60 years of research—is to solidify and reinforce the existing social identities as well as partisan, ideological, or policy views of voters.  That is, campaigns tend to bring potentially wayward voters back “in the fold.”  This only tends to polarize voters.  Indeed, it can happen even in the space of a single 30-second ad.  If, under a mandatory voting system, candidates no longer have to worry about mobilizing voters to turn out and can concentrate on persuading voters to support them, I suspect that we would see this same effect, but magnified over the entire electorate.  So it’s entirely possible that mandatory voting may even increase polarization.


9 Responses to Mandatory Voting Isn’t a Solution to Polarization

  1. Kevin November 8, 2011 at 9:22 pm #

    Galston writes, “The reason is that hard-core partisans are more likely to dominate lower-turnout elections, while those who are less fervent about specific issues and less attached to political organizations tend not to participate at levels proportional to their share of the electorate.”

    Certainly, at the individual level partisanship is associated with turnout, but is there empirical evidence that high turnout US elections tend to see less polarized electorates than low turnout elections? If so, how strong is this association? It would seem to be a first order test of Galston’s hypothesis.

  2. Adam November 8, 2011 at 10:03 pm #

    Ansolabehere and Hersh look at the views of voters and non-voters and find that they’re even closer than previously thought…

    Stephen Ansolabehere and Eitan Hersh. 2011. “Who Really Votes?” In Facing the Challenge of Democracy, eds. Paul M. Sniderman and Benjamin Highton. Princeton: Princeton University Press.

    Past scholarship has shown that voters and non-voters have similar political preferences to one another, implying that if everyone voted, political outcomes would hardly change at all. We re-evaluate this claim while correcting for the many non-voters who falsely report on surveys that they voted, and find that voters and non-voters are even more similar to one another than previously thought.

    • Andrew Gelman November 9, 2011 at 7:45 am #


      The link you gave doesn’t link to an actual copy of the paper. But given the research I’ve seen on the topic, I’m doubtful of the claim, “Past scholarship has shown that voters and non-voters have similar political preferences to one another, implying that if everyone voted, political outcomes would hardly change at all.” Perhaps we could lock these people in a room with Leigley and Naglerl until they could come to agreement!

  3. DavidN November 8, 2011 at 10:11 pm #

    Suppose you model polarisation by a bimodal distribution of voters on an ideology scale, then if non-voters are less conservative then wouldn’t mandatory voting lessen the influence of more conservative elements in both parties i.e. the median voter ex-post will be more liberal, this will have the effect of moving the median in the distribution towards the left, but of itself doesn’t say anything about polarisation (the distance between the two modes). Ex-post polarisation will depend on the distribution of non-voters which according to table 1 seems to suggest non-voters are more concentrated in a central mode as opposed to bimodes.

  4. Peter T November 8, 2011 at 11:09 pm #

    One thing this misses is that the more reliably “liberal” electorate produced by compulsory voting would raise the stakes for the radical right. In effect, they would have to choose between radicalism and electability. The experience of other countries suggest most would go for electability – ie, polarization would lessen as the right moved to the centre. The experience of the late 19th century suggests otherwise, but was not the politics of the time dominated by slavery and race issues – bitterly divisive in themselves in a way that more purely economic conflicts rarely are?

  5. Claire H November 9, 2011 at 10:21 am #

    Peter T. makes an excellent point. The model may also be missing some nuance in the elite response to ideology in the electorate, even if that larger (because mandatory) electorate is just as polarized. If the parties didn’t have to worry about turning out (by riling up) their respective bases, they might moderate their positions – choosing electability over radicalism.

  6. Nolan November 9, 2011 at 3:52 pm #

    I have added my $.02 at

  7. Kevin November 9, 2011 at 6:02 pm #

    I should clarify my earlier comments. By high / low turnout election, I meant midterm and odd year elections versus presidential years, rather than looking polarization and turnout in presidential election years only which, as John pointed out in his post, are both steadily increasing

  8. ZC November 9, 2011 at 9:08 pm #

    Third Peter T’s comment. It really does depend on how “liberal” the nonvoters are – if they were -really- liberal (which I doubt), the net effect would probably be to decrease polarization. Polarization to my mind increases when both parties smell blood and all the branches are up for grabs, as they are now. (That would fit, with some shoehorns here and there, with much of the mid to late 1800s). We’re more polarized on some social-cultural issues because many traditional religious positions are losing ground, in a secularizing America, and the sides are more evenly matched; we weren’t polarized on gay marriage in the 1950s because no pol in his right mind would have risked standing up for gays and lesbians. And we’ll be less polarized on gay marriage 20 years from now, the same way that the “women’s role in society” question in the ANES has become largely an asterisk and sometimes isn’t even factored into measures of polarization.

    I really doubt that nonvoters are -that- liberal, but I would wager that the increase in nonvoters would tamp down the radical right somewhat and more our politics a bit to the left. Which to my mind would =be= more “moderate.” If we’re allowing Michele Bachmann to redefine what it means to be conservative, than “moderate” has shifted right already, and the inclusion of a broader electorate, while not perhaps reducing polarization per se, might shift the spectrum back left a bit.