Don’t Compel People to Vote; Use a Voting Lottery Instead

I’m not in favor of a voting lottery. However, I think the best arguments for compulsory voting never succeed in justifying compulsory voting. Instead, at best, they justify replacing our current system with a voting lottery.

Consider the “Representativeness Argument” for compulsory voting: Under a voluntary voting regime, the median, mean, and modal voter is white, maler, richer, more advantaged, straighter, etc., than the median, mean, and modal eligible voter. That is, when voting is voluntary, more advantaged people are more likely to vote than less advantaged people. It seems reasonable to worry that the government will thus fail to properly represent the real interests of the less advantaged.

Now, on the contrary, I think forcing the less advantaged to vote will tend to harm them rather than help them. A rational poor person who knew the literature on how ignorant poor people are wouldn’t want to have poor people vote en masse. But suppose I’m wrong about all that and the above argument is right. Compel people to vote? Nah. Use a voting lottery instead.

Here’s an excerpt from chapter 2 of Compulsory Voting: For and Against:

In a voting lottery, all citizens have the same equal fundamental political status. While in universal suffrage, every citizen has one equal vote, in a voting lottery, every citizen has equal eligibility to vote. Elections proceed normally, with candidates working to gain support from voting-lottery eligible citizens. Shortly before the election, the system selects a pre-determined number of citizens at random. These citizens—and these citizens only—become “electors”, imbued with the power to vote. To ensure turnout, the government pays each elector a substantial sum to vote.[i] They are not forced to vote. We might perhaps ask them to sign a contract committing them to voting (in exchange for the payment), and then allow them to be punished for breach of contract if they renege. This involves compulsion, but only compulsion to which citizens genuinely consent.

For instance, in a US presidential election, we could select 20,000 citizens randomly from all eligible voters. We pay them $1000 to vote. They and they alone decide the election. In a local election, we might select a much smaller number of local citizens, and pay them significantly less.

This is just one way to instantiate a voting lottery. The exact details will not be important here. The important idea is that we select a small, but not too small, subset of the population, given them the right to vote, and then pay them to vote. Note that we could also have voting lotteries for any set of elections where we are worried that disadvantaged citizens are not voting proportionately.

Since the 20,000 citizens are selected at random, they will be representative of the country at large. In terms of their demographics, socioeconomic statuses, political opinions, and so on, the 20,000 will be a mirror image of all eligible voters in the US.

Of course, there remains the problem of “random sampling error,” i.e., that the 20,000 citizens will not be perfectly representative of all eligible voters in the country. Yet, the greater the number of electors, the smaller the margin of error. Let’s do some statistics. The United States has about 207 million eligible voters. 20,000 electors produce a statistical confidence level of 99 percent, at a confidence interval of 0.9, with a margin of error of only 0.69. So, the lottery would be extremely accurate.

You may be familiar with Gallup and Rasmussen polls, which attempt to predict the outcomes of the election ahead of time. These polls use a sample set of only 1500 or so citizens to try to gauge the voting behavior of the citizenry at large. The polls are quite accurate. But the voting lottery is far more accurate than these polls, with a far lower margin of error at a higher confidence level.

Voting lotteries are not just as equally representative as compulsory voting. At current levels of technology, they may even be more accurate—more representative—than compulsory voting. It often surprises the public to hear this, but vote counting in real elections is not perfectly accurate. Every vote counting technology or method has a significant margin of error. Political scientist David Kimball estimates that the residual vote rate (one kind of error rate) in presidential elections tends to be between 1 and 2 percent.[ii] One average, hand counting of ballots results in a 2 percent error rate.[iii] In general, if the Democratic and Republican candidates in a US presidential election are within one percentage point on the popular vote, you should treat the result as a statistical tie. In general, the more votes that must be counted, the greater the margin of error. Thus, if we want the voting population to be representative of the voting eligible population, a voting lottery may be superior to compulsory voting.

Also, in the United States, the Electoral College makes it so some people’s votes count for less than others. A Virginian vote is more likely to decide a presidential election than an Oregonian vote. A voting lottery eliminates this bias.

Note that even under compulsory voting systems, a significant portion of the population still abstains from voting. In Australia, typically 5-7 percent of registered voters do not vote. The 93 percent of voters who do vote are not perfectly representative of all eligible voters. Poor minorities are still less likely to vote than rich whites. Also, Australia’s much-lauded 93 percent turnout is misleading. In the 2011 parliamentary elections, 13.1 million citizens voted out of a voting-age population of 16.2 million. In fact, only about 81 percent of eligible voters in Australia vote. 93 percent of registered voters vote, but over 10 percent of the population does not register to vote.[iv]

Voter lotteries have other advantages over compulsory voting. Voter lotteries are less expensive, even if we pay electors to vote. For a federal election, the voting lottery would cost $20 million in electors’ fees. Add administrative expenses, such as expenses for conducting the lottery, counting votes, and so on. Suppose the voting lottery is extremely expensive to conduct—suppose it would cost $50 million total. That may seem like a lot. It’s not. In contrast, the administrative costs of the 2000 US federal election were about $1 billion, according to Caltech and MIT’s Voter Technology Project.[v]

Voter lotteries not only cost state and local governments less money, but they cost citizens less time. Suppose we force all 207 million eligible American citizens to vote. Suppose compulsory voting advocates get their dream outcome—suppose everyone votes. Now suppose voting takes each citizen on average one hour, including time spent driving or walking back and forth to the polling station. Compulsory voting would thus cost 207 million hours of our time. This is equivalent to 300 full average American lifetimes spent voting. It is equivalent to 5 million American work weeks spent voting. This time has an opportunity cost. What other valuable things could Americans do in this time? In contrast, the voting lottery takes far less time.


[i] We might require employers to give electors a day off to vote, as well. However, this trades compulsion against citizens as voters for compulsion against employers.

[ii] The residual vote rate is the percent of votes that were cast but not counted for one reason or another. See David C. Kimball, “Summary Table on Voting Technology and Residual Vote Rates,” (St Louis: University of Missouri, 14 December 2005), URL: <> See also Martha Kropf and David C. Kimball, Helping America Vote: The Limits of Election Reform (London: Routledge, 2011): 37-44; Stephen Ansolabehere and Charles Stewart III, “Residual Votes Attributable to Technology,” Journal of Politics 67 (2005): 365-89.

[iii] Stephen N. Goggin, Michael D. Byrne, and Juan E. Gilbert, “Post-Election Auditing: Effects of Procedure and Ballot Type on Manual Counting Accuracy, Efficiency, and Auditor Satisfaction and Confidence,” Election Law Journal 11 (2012): 36-51.

[iv] See the tables available at the International Institute for Democracy and Electoral Assistance, URL: <>

[v] Caltech/MIT Voting Project, “Voting: What Is, and What Could Be,” (Caltech/MIT, July 2001): 13, URL: < voting_what_is_what_could_be.pdf>

Later, I consider an argument from Bill Galston, in which he argues we should use compulsory voting to stop non-voters from free-riding on voters. He says that non-voters receive the benefits of democracy without paying for its upkeep, and the right way to pay for it is through voting. But, I respond, this doesn’t work:

Galston uses the analogy of jury duty. Juries play a vital role in the criminal justice system. No one—except foaming-at-the-mouth libertarians and anarchists—thinks it is unjust for government to compel citizens to serve on juries.

Galston also uses the analogy of the draft. If we are under a dire existential threat, we might have no choice but to compel young citizens to fight for our defense.

However, these analogies might undermine rather than support the case for compulsory voting. Consider jury duty again. Imagine someone said, “Jury duty plays a vital role in maintaining criminal justice. Therefore, we should compel all citizens to serve on the jury every time there is a criminal trial.” That argument would seem absurd. We don’t need or want everyone to be a juror on every trial. Rather, at most, we want a small number of people to serve as jurors for each trial, say, 12 people per trial.

We don’t compel every citizen to serve on every jury. We don’t even compel every citizen to serve on at least one jury. Instead, we use a lottery system. To make sure we get the jury members we need, we randomly select a small subset of qualified citizens to serve as jurors. Most Americans never serve on a jury their entire lives.

Similarly, when using a draft, we would not force every American to fight unless we needed every American to fight. Rather, we would use a lottery. If we need 50,000 troops, we randomly select 50,000 citizens from a pool. Maybe we instead select 70,000 to ensure we meet our needs. But we don’t force everyone to take up arms.

And so, again, Galston’s free rider argument doesn’t support compulsory voting, but, at most a voting lottery.


Published on:
Author: Jason Brennan
  • Jerome Bigge

    The Greeks of Classical Athens used a lottery to select which citizens would serve as representatives of the citizens of Athens. Apparently it was rather like jury duty in that if you were selected to be a representative, you were expected to serve unless you had a very good reason why you couldn’t do it. Of course those of Athens were were “citizens” only applied to free born males. Women, slaves, those who lived in the city but where not citizens of the city-state were excluded. In early America, in order to be legally allowed to vote, you had to be white, male, and an owner of income producing property. Objectively not a whole lot different in some aspects from the way that the Greeks did it.

    The current make up of Congress consists of people with an average net worth in excess of one million dollars. As a percentage of population, the professions are vastly over represented in Congress relative to their numbers among the general population.

    Which brings up the question of do the “selected electors” end up voting from the same class of people that now make up the House of Representatives and the Senate? If we use the concept of a true cross section of the American people, the lower income people will be voting as a larger percentage of the electors than is the case today. We’d likely be electing more Democrats and fewer Republicans given that the well off who more likely vote today tend to vote Republican. Whereas with a true cross section of population voting we’d get more Democrats and fewer Republicans. The US would move from the conservative side of the political spectrum to the liberal (modern) side to a certain degree. On a whole, people “might” be better off, but as I’m rather strongly opposed to the excessive growth of government, I would prefer to reduce the size of government, not increase it to more European levels.

    • Les Kyle Nearhood

      I brought up the Athenian model in another forum, It is alien to our way of thinking but often is the times I have thought that I would rather that random voters be in charge than the career politicians we have ruining things.

      • ThaomasH

        Why not got straight to a Congressional Lottery?

      • AP2

        “For the last century, almost all top political appointments [on
        the planet Earth] had been made by random computer selection from the
        pool of individuals who had the necessary qualifications. It had taken
        the human race several thousand years to realize that there were some
        jobs that should never be given to the people who volunteered for them,
        especially if they showed too much enthusiasm. As one shrewed political
        commentator had remarked: “We want a President who has to be carried
        screaming and kicking into the White House — but will then do the best
        job he possibly can, so that he’ll get time off for good behavior.”

        Arthur C. Clarke,

        Imperial Earth

    • Damien S.

      AIUI you had to volunteer for the Athenian jury pool; also they got paid a decent rate. Conversely it wasn’t perfect representative: you had to be 30 or older, and could be screened for good character. But otherwise, yes, the Boule or Council of 500 was selected by lot for a year, and set the agenda for the Assembly (basically, proposing laws that still had to be approved by referendum), and juries of 201 people or more tried cases and sometimes laws.

      “On a whole, people “might” be better off, but as I’m rather strongly
      opposed to the excessive growth of government, I would prefer to reduce
      the size of government, not increase it to more European levels.”

      So you prefer your idea of liberty to democracy.

  • Jameson Graber

    I agree with you that the “voting is a duty” argument suggests a voting lottery a lot more than it does mandatory voting. Which just reinforces my belief that voting is not a duty.

    Personally, I always get irritated when people try to encourage more voting. There are already a lot of people who vote but shouldn’t. And they’re probably making life worse for all of us by persisting in their voting.

    Seriously, why does anyone think they have a duty to vote, but not a duty to be smart enough to understand politics or political theory? As if the mere expression of your preference somehow helps government get better. Um. What?

    • ThaomasH

      I agree that citizens have a duty to try to be informed, not vote out of mood affiliation, read Robin Hanson daily, etc.

      • Sean II

        Shouldn’t the phrasing on that be “citizens have a duty to vote correctly, for which being informed is usually a pre-condition, though hardly a sufficient one”?

        This whole “well-informed voter” thing has the ethics of voting exactly backwards. For example..

        If Comrade Gletkin and G.A. Cohen both go to the polls and cast a vote for Stalin, a reckoning of consequences would say they both did equal harm. But even folk morality (moral intuition, if you must) won’t give us a result where Cohen is the better voter.

        Indeed, I can perhaps forgive Gletkin on the grounds that he is an ignorant peasant who couldn’t have known better. But Cohen I do not forgive, because he has all the tools needed to figure out why voting for Stalin is a terrible idea. The fact that he defends his decision with elaborate sophistry is an aggravating factor, not a mitigating one.

        • ThaomasH

          I’m not sure “who to blame most” is a good way to discuss policy, but for better or worse I DO blame a deluded Democrat for supporting an increase in the minimum wage less than I do a Republican for not pointing out that he could achieve as much redistribution at lower cost with an increase in the EITC.
          BTW, without knowing who else was on the ballot, I don’t know if we can fault either Cr. Geltkin or Professor. GA Cohen for voting for Cr. Stalin, but I did not read anything in The Mises Review: Why Not Socialism? by G.A. Cohen
          The Mises Review: Why Not Socialism? by G.A. Cohen
          View on Preview by Yahoo
          Why Not Socialism?

          Why Not Socialism?
          If anyone could have given socialism a plausible defense in the current milieu, it would have been the late Marxist philosophy professor G. A. Cohen. But rather tha… View on Preview by Yahoo

          that led me to think that Professor Cohen had any sympathy for Cr. Stalin.  None of the examples he uses imply a Stalin-like situation in which one person had life and death power over the other.  His obituary in the Gardian says “Continually modifying his theories in the light of history and his own experience, he was an early critic of abuses in the Soviet Union, and ultimately described himself as an ex-Marxist.”
          Therefore I don’t know what lessons we can draw from the hypothetical voting behavior of GA Cohen.

          • Sean II

            Well then, swap in any two examples you like where two people vote the same horrible way, but one does so out of ignorance and the other does so from atop a mountain of epistemically elegant bullshit.

            Hell, forget that and just consider this: imagine two middle school students commit a murder. One is developmentally disabled, and his reason for taking part in the killing is “it was something to do”. The other kid is a genius who defends his crime with an elaborate theory a’la Hitchcock’s Rope.

            Everyday moral intuition says the second kid is worse, and yet…if we follow the logic of the “informed voter” ethic, we’re supposed to prefer the genius killer on the grounds that he had better reasons. What…the?

            Now let’s slip back out of analogy mode, and you tell me who is worse: a) Some ignorant mom who voted for murderer George W. Bush on the grounds that “he has such nice hair”…or b) some erudite neo-conservative history professor who voted for murderer George W. Bush thanks to a bunch of intricate baloney about Plato, Strauss, etc.

            Again, everyday moral intuition says both are bad, but the second is clearly worse.

          • ThaomasH

            Actually, THAT part I don’t really object to.   Being better informed CAN sometimes lead to worse voting behavior than being less informed.  I agree with you that a naive mistake based on a simple misunderstanding of one’s self interest or misplaced sympathy, is less culpable that the same position taken out of ideological positioning.
            But what implications should this have for how easy or difficult we make it for people to vote?

          • Sean II

            Well, I suppose the big implication is that voting lotteries or poll qualifications wouldn’t make much of a difference.

            The only real solution is the one we’ve always known: to make voting less important by reducing the amount of power available to be captured.

            A moral vote is one which does less harm, little harm, or no harm.

            The best path to that is a vote which does little.

          • ThaomasH

            I agree.  Government should be as big as necessary but no bigger. And one way to make no bigger than necessary is to propose less intrusive/costly/politicized ways to deal with problems.

  • ThaomasH

    I’m not in favor of compulsory voting, either, but I think that it’s advantaged people who are voting “ignorantly” meaning not in the way I personally think is best, favoring things like a less progressive tax system, for example. But I would not on those grounds try to discourage the richer people from voting.

    I think a much better approach is to remove obstacles that make it more costly for poorer people to vote than rich people. Elections could be held on Sundays for employers subsidized for giving time off for voting. If certain kinds of identification are the lowest cost method of preventing voter fraud (I seriously doubt this in several cases), then states should make it easy for people to obtain such identification.

  • StephenMeansMe

    This is really interesting. When I finally set up my night-watchman-plus-LVT-funded-BIG leftish libertarian futarchy, I’ll make sure to have a voting lottery.

    There might be some Constitutional issues with lottery voting in certain cases (the legislature specifically), whereas the Presidential elections could *easily* be turned into lottery voting given the Constitutional electoral framework. It’d be as simple as changing how electors get chosen.

    And since voting is a fairly trivial exercise, a lottery might make it “special” enough to get more people invested in the outcome by other means! (It’d be fun to see how a voluntary-voting legislature compares against a lottery-voting executive, if there’s any appreciable change in voting patterns there.)

  • Damien S.

    Of course, you could skip the step of sampling people to elect a legislature, and just have the sample *be* the legislature. There are tradeoffs either way, like losing long term experience; I like the idea of a bicameral legislature, one house elected, one selected by lot.

    • Damien S.

      One catch with any method of creating a small body of voters is whether you’ve made it easier to bribe one’s way to the desired outcome. Even with secret balloting obscuring responsibility, there could be “you’ll all get $X if the vote goes my way” strategies. With full elections you have to bribe everyone, which would at least have egalitarian benefits if it happened.