The positive effect of mandatory voting laws on turnout is well documented, and its success in reducing the disparities between voters and non-voters is more or less settled. Next: what is the effect of participation requirements on political information and informed voting? Brennan campaigns against increased turnout, pointing out that non-voters are less informed than voters, and arguing that increasing the number of uninformed votes worsens electoral outcomes. I disagree. Before I respond to this argument, a clarification is necessary.
1. Compulsory Voting as Compulsory Balloting
The names “mandatory voting” and “compulsory voting” (CV) are both misleading. Modern democracies using this voting system do not force all citizens to vote, nor do they penalize citizens who cast invalid ballots.
Rather than combining voting as a single stage, a more accurate model of voter behavior recognizes that balloting and voting are 2 different decisions. Citizens first choose whether to acquire a ballot (including all the costs of registration, transportation, and time spent traveling to the polling place) and then, only if the citizen chose to acquire a ballot, choose whether to mark a valid vote for each contest on that ballot. The widespread existence of secret ballots prevents democratic governments from monitoring if (and how) a person marked their ballot. Therefore, non-participation penalties are attached to the balloting stage, not the voting stage. Citizens are therefore able to submit a blank, spoiled, or invalid ballot without penalty.
Therefore, a better name for the voting system currently practiced in 31 democracies is Compulsory Balloting (CB). CB can be defined as a penalty for not casting a ballot, which allows citizens to ballot and abstain without penalty.
2. CB, Political Information, and Informed Voting
Full voter turnout sounds appealing, but should uninformed citizens cast votes? Most of us would agree this would be counter-productive. However, the argument that CB encourages uninformed voting is based on a misperception of how the institution works. In theory, CB should not increase uninformed voting. The reason is explained just above: CB does not penalize actors for not marking valid votes. It only penalizes actors for not casting ballots. If marking a random vote would hurt the expected outcome of an election, an uninformed voter should rationally abstain from voting, and submit the blank ballot. This suggests that CB will not increase uninformed voting.
Would all insufficiently informed citizens cast blank ballots under CB? Probably not. Ballot order position effects would not exist if some people weren’t willing to vote at random. However, roll off voting trends and other comparative studies demonstrate that uninformed citizens are capable of rationally abstaining.
There’s more. Imagine all the informed people who don’t vote. If incentivized to submit a ballot, most of them would probably invest the extra few seconds to mark votes on that ballot. So the number of informed votes would increase. Furthermore, because CB penalties effectually sink part of the cost of voting, CB makes it more likely that a previously uninformed voter will find it rational to invest in both information and informed voting.
Information levels are not static. They are dynamic, and adapt to the cost and incentive structures shaped by electoral systems. When we allow the decision to become informed to be endogenous to participation costs, this demonstrates that lower costs of participation increase incentives for uninformed actors to invest in both information and informed voting. The net result is that CB increases informed voting among previously informed people, increases informed voting among previously uninformed people, and theoretically does all of this while not increasing uninformed voting at all. See this paper (pdf) for an extended demonstration of this effect.
Again, there’s more. Information environments aren’t static either. Campaigns are dynamic and respond to the incentives shaped by electoral rules. Introducing non-participation penalties will shift the composition of the electorate, and campaign strategies should adjust accordingly. A likely effect is that moving from voluntary voting to CB will shift campaign strategies from a focus on mobilizing the base to a focus on persuading the centrist, undecided, and most uninformed people. These people have weaker prior commitments and are the easiest to persuade. This might, in effect, make information more readily available to populations who need it most, further increasing aggregate information. This paper (pdf) discusses the mechanisms that link CB to increased information and demonstrates that Austrian citizens who lived in CB provinces had more political information than those in voluntary vote provinces.
3. CB and Polarization
The information effect also ties into the polarization debate. A person will vote when their perceived benefit of voting exceeds their perceived cost of voting. In a 2-party system, people on the extremes experience a bigger benefit differential between the two candidates, as compared to people in the middle. This suggests that extremists naturally have more at stake in an election, and so have a higher perceived benefit from voting. This difference increases with polarization, as the candidates are placed further apart.
When voting costs are high, it is difficult to motivate a centrist citizen to endure the cost of voting, since that actor has little to gain from affecting the outcome of the election. As voting costs decrease, the probability that a moderate can be incentivized to vote increases. Because CB effectually decreases the considered cost of voting, it should also increase the participation among moderates.
However, if you buy my argument above that CB would increase aggregate information, this might cause individual-level extremism to increase as well. A la Converse, one might argue that increased information will decrease internal inconsistencies and increase ideological constraint. A la Zaller, one might argue that extreme citizens require information in order to realize their true preferences.
4. CB and Left-Wing Advantage
Are non-voters more moderate? Probably. Are they more liberal? Usually. One possible exception worth noting: Austria used to practice wahlpflicht (CB) nationwide, but later left this policy up to the provinces. If left-wing citizens are more likely to be voluntary non-voters, conventional wisdom would suggest that the liberal provinces would keep CB in place and conservative governments would eliminate CB. But in fact, just the opposite occurred. The four provinces that kept CB the longest (Tyrol, Vorarlberg, Styria, and Carinthia) are among the four most conservative provinces in the country.
The explanation? A colloquial rumor I heard several times was that conservatives put CB in place after women were granted the right to vote. Why? Conservatives were concerned that only “liberated” (aka left-wing) women would vote, and this would shift victories to the left-of-center SPO. Whereas with CB in force, all women would vote equally, and conservative parties like the OVP would maintain their majority within the active electorate.
I asked Erhard Busek (OVP, Vice Chancellor of Austria 1991 – 1995) about this theory, and he told me it has no merit. His explanation: The OVP has always been more likely to not only provide rights but also to expect obligations from citizens, and voting is just another such duty. A third explanation I heard: after reversing the embarrassing laws that stripped voting rights from Jews, the Republic instituted wahlpflicht in order to make a strong statement: not only do all persons now have the right to vote, they also have the obligation to vote.
5. Full Turnout and the Composition of the Electorate
Another comment: In the debate over full turnout, a common strategy is to simulate previous elections as if everyone had voted. These studies are clever and useful in many ways, but I note two common assumptions to watch out for in this area of research.
First of all, these simulations assume that non-voters would have the same preferences if they were voters. However, if we accept that campaigns and information acquisition are dynamic and endogenous to electoral systems, we must accept that measured preferences might be different under a full turnout scenario. Information and preferences under one electoral system are not necessarily a good estimate of preferences under another system.
Second, even if all preferences remained the same, a full turnout simulation should go beyond comparing partisan victory totals. Especially when one considers the number of non-competitive districts within the US, it is not too surprising if the number of districts changing party control is small.
However, maintaining party control while significantly changing an expected margin of victory is still an important difference. This would, in turn, cause existing candidates to adapt by either (with a smaller margin) being more attentive to the voters, or perhaps (with larger margin) by being more attentive to their own policy agenda. Furthermore, primary elections should adapt too; as a party increases or decreases its margin of general support, the pool of potential candidates it can successfully nominate will also adjust to include more extreme members.