Ben Saunders: Combining Lotteries and Voting

Ben Saunders wrote a comment on Claudio López-Guerra’s The Enfranchisement Lottery:

Combining Lotteries and Voting

In recent years, a number of theorists have turned to the Athenian practice of sortition to inspire proposals for democratic reform. Some simply propose that politicians can be appointed by random selection, thereby producing a statistically representative sample of the population (Callenbach and Phillips, 2008). Others, however, seek some way of combining lotteries with the more familiar modern practice of voting. I shall confine my comments to two recent proposals. López-Guerra (2011) suggests abolishing universal suffrage, instead having only a randomly-selected sub-set of the populace vote in elections. Though he is reluctant to endorse this policy all things considered, he argues that it would likely produce better outcomes, since it would be possible for selected voters to participate in a program of education and deliberation prior to the actual election. The proposal is therefore similar to that of a national ‘deliberation day’ (Ackerman and Fishkin, 2005), except that participation in both deliberation and the subsequent vote is restricted to a sample of the population, largely on grounds of cost (pp. 8-9). The second proposal is one that I have developed elsewhere (Saunders 2010), namely lottery voting. This is proposed as an alternative to majority rule. Elections are held as normal but, rather than the majority winning automatically, a single randomly-selected vote determines the outcome. Consequently, each side has a chance of victory proportionate to the level of support it enjoys amongst the electorate.

At first sight, these two policies are quite different. López-Guerra suggests that democracy is possible without universal franchise, provided that all have a chance of being enfranchised. My own contribution is to argue that democracy is possible without majority rule, provided that all have a chance of getting their way. In principle, these proposals could be combined: we could restrict the franchise to a random sample of the population and have their decision determined by a randomly-selected individual. It is, however, fruitful to compare these two schemes because there are greater similarities than may initially be apparent. Both combine lotteries and voting; but López-Guerra uses the lottery first, to decide who can vote, whereas I propose holding a vote and then using a lottery to determine which vote is decisive. Both proposals can be considered forms of lottery voting, but one ex ante (the lottery taking place before the voting) and one ex post (the lottery taking place after voting) (p. 8). The proposal I have defended, which I simply called lottery voting, is the more familiar ex post variant. López-Guerra, however, contends that his ex ante form of lottery voting is quite different (which I grant) and has certain benefits. In what follows, I question whether an ex ante lottery is really preferable to an ex post one.

There are a number of differences between our proposals that I shall set aside for present purposes. Firstly, López-Guerra is concerned with the election of representatives, whereas I suggest that lottery voting can be extended to direct decision-making (Saunders, 2010, p. 151, fn. 11). Let us assume that the matter in question is indeed the appointment of a legislative assembly. Secondly, López-Guerra assumes that participation will be mandatory for those selected (p. 2). I did not assume this, but let us suppose that voting is compulsory. A third difference concerns the number of vote(r)s selected. López-Guerra is vague on what percentage of the potential electorate ought to be selected (pp. 4-5), but it is clear that the electorate will be much larger than a single individual. Conversely, I propose selecting a single vote to decide on any given issue (Saunders, 2010, p. 151). This, I think, may be important to the central question that I wish to focus on here, viz. whether we have reason to prefer an ex ante or an ex post lottery.

López-Guerra favours an ex ante lottery because it allows educational and motivational benefits (p. 8). If we identify the relevant electorate before the vote then we can target education where it will do most good. Moreover, those who are selected will hopefully be motivated by a sense of public duty, strengthened by the fact that each vote (because part of a smaller electorate) will be more weighty than under universal suffrage. These effects, he claims, would not occur if the lottery took place only after voting; people need to be selected ex ante if their behaviour is to be modified (p. 10). But the ex ante identification of voters could have negative effects, offsetting these advantages. If it was known in advance which people would decide the election, then it would be possible to target them not only with education but also with threats or bribes. An ex ante lottery could increase the risk of corruption, thereby worsening electoral outcomes. Moreover, though López-Guerra hopes that increasing the weight of each vote will increase citizens’ sense of civic responsibility, leading them to vote for the common good, it has been argued that it is the very inconsequentiality of each vote, in a large election, that allows each voter to set aside their private interests and focus on the common good (Brennan and Lomasky, 1993). Once voters are given more influence, there will be a greater temptation for them to seek to promote their private ends, rather than the common good.

I think it is wrong to assume that citizens must be selected ex ante for their behaviour to be changed. Voting behaviour is, to a significant extent, endogenous to electoral systems. In defending ex post lottery voting against the accusation that it would allow extreme minorities to win, I argued that we cannot assume that the levels of support apparently enjoyed by extremists under majoritarian systems will carry over to lottery voting (Saunders, 2010, p. 172). If many of these votes are actually protest votes, then it may be that voters feel free to vote for extremists under majoritarian systems, because they are confident that those extremists will never win a majority. If outcomes are to be decided by a single, randomly-selected vote, then this may encourage all to be more responsible, since their vote could be decisive. Of course, it could be argued that the chance of being decisive remains small, so there is little impetus on individual voters to exercise greater responsibility. Nonetheless, it seems that voter behaviour may change, although the lottery happens only after the votes are cast.

Conversely, I hope the fact that no individual expects to be decisive encourages them to focus on the common good, rather than their own private interests. It is, I concede, a potential danger of ex post lottery voting that it may encourage extremism, since individuals may as well demand the whole cake if their chance of getting it is undiminished (Saunders, 2010, pp. 172-3). If voters do not know, beforehand, that they will be the decisive individual, then I believe that it will be easier for them to vote for the common good. An individual identified ex ante as decisive, however, is more likely to succumb to the temptation of promoting their own interests. Thus, the choice between an ex ante and an ex post lottery depends on the number of voters selected. If the decision is to be made by a single individual then I strongly prefer an ex post lottery to appointing a random dictator ex ante. The more people that we enfranchise, the weaker my preference for an ex post lottery becomes. The question is whether an ex ante lottery ever comes to seem preferable.

López-Guerra concedes that his enfranchisement lottery is probably not, all things considered, preferable to universal suffrage – his claim is merely the more modest one that it has some advantages (p. 16). It seems that he thinks the same advantages favour the ex ante enfranchisement lottery over both universal suffrage and ex post lottery voting. I have cast some doubt on the relative advantages of the enfranchisement lottery over ex post lottery voting. I hope to conclude by showing that the reasons of legitimacy and stability that favour universal suffrage over the enfranchisement lottery also give us reason to prefer ex post lottery voting to the ex ante version.
López-Guerra notes that one important function of elections is ensuring the peaceful transition of power, which they usually do by making plain the relative support that rival leaders enjoy (pp. 12-16). In any system that involves a lottery, there is the danger that the outcome of the process will diverge from the popular (majority) will and the public may always suspect manipulation of the random process, thereby casting doubt on its legitimacy (pp. 14-15). Universal suffrage, combined with majority rule, leaves less room for doubt about the outcome of the election and therefore enjoys the advantage of publicly demonstrable legitimacy.

These considerations suggest that a determinate process will be preferable to any process involving a lottery, whether that takes place ex ante or ex post. I agree that these are important concerns. Citizens must be able to trust the workings of the electoral system if it is to enjoy their support. The problem is not, however, confined to (partly) random systems – sham elections are frequently run under dictatorships, in which ballot boxes are stuffed and voters intimidated. If our concern is with descriptive legitimacy, then almost any system can be accepted as legitimate, provided that the people are given some reason to accept its outcome (for instance, the lottery could be interpreted as the will of God). Nonetheless, I believe that these considerations give us reason to favour ex post lottery voting to its ex ante variant, even if they also support a determinate process (universal suffrage and majority rule) over any lottery. Ex post lottery voting is open to all agents, even potential revolutionaries, in a way that a random franchise is not (p. 13). Moreover, I think the fact that it allows all to vote in any given election will increase perceptions of legitimacy and have a salutatory educational effect on the citizenry (Saunders, 2010, p. 154). The enfranchisement lottery will forcibly exclude many from participation. If we value voting, then I think it best to allow all to vote, only holding the lottery afterwards.

López-Guerra argues that the franchise lottery is likely to lead to better outcomes than universal suffrage or ex post lottery voting, but suffer in terms of legitimacy. Ex post lottery voting may be a mean between the two in terms of legitimacy – preferable to an ex ante lottery, though inferior to universal suffrage and majority rule. Whether it is, all things considered, to be preferred depends on how we weigh these competing values and whether we can find alternative ways of satisfying either (pp. 17-19). I have argued that an ex post lottery may not be inferior to an ex ante one with respect to outcomes, which may give us reason to favour it all things considered. The final choice between the two is unlikely to be made in the abstract. In choosing electoral systems, we must attend to the concrete circumstances in which they will be used (Saunders, 2010, p. 177). Nonetheless, those interested in democratic innovations have reason to attend further to these two schemes, including investigating whether they can profitably be combined.


  • Ackerman B and Fishkin J (2005) Deliberation Day. New Haven, CT: Yale University Press.
  • Brennan G and Lomasky B (1993) Democracy and Decision. New York: Cambridge University Press.
  • Callenbach E and Phillips M (2008) The Citizen Legislature. Exeter: Imprint Academic.
  • López-Guerra C. (2011) The enfranchisement lottery. Politics, Philosophy & Economics. [cited from online first m/s, pp. 1-23]
  • Saunders B (2010) Democracy, political equality, and majority rule. Ethics 121: 148-77.

11 Responses

  1. The Athenians and the Italian Renaissancers tried this very thing, dabbling in Sortition, trying it out, using it here but not there. And what happened to such “reform” experiments? They all failed. Isn’t there anybody off-campus out here who will advocate Election-by-Lot as PRINCIPLE rather than only as a feature to be mingled into the Popularity-Contest system of electing Public Servants? Any proposal that is short of BLANKET Election-by-Lot is headed down the same old dead-end road of the Ancients. “BLANKET” means for all all political Offices at all Levels of government and in all Branches of said governments. Somebody?


  2. I suppose in some parts of academia, it’s advantageous to have a novel idea, even if it’s a silly one. I find Saunders arguments entirely unconvincing.

    With this system, you will have loony decisions once in a while. Never mind extremists. Regardless of the beneficial effect on people’s decision making being randomly chosen as dictator supposedly has, mentally ill people will make the decisions in a non-negligible number of cases.

    It’s a bit of a moot point anyway, since all but the most crazy extremists would prefer a policy “smoothed” by voting on decisions, to one where their opposites were randomly superpowered on occasion.


  3. The principal advantage of ex-ante lottery voting is, as Fishkin puts it “A representative microcosm offers a picture of what everyone would think under good conditions. In theory if everyone deliberated, the conclusions would not be much different.” Sortition is not an end in itself, merely an impartial way of reducing the number of voters to the extent that the individual vote regained the causal force that was lost as a result of universal suffrage. In an ex-post lottery the outcome itself is randomly determined, so ceases to be a form of descriptive representation. It would be unlikely to appeal to anyone who values democracy.

    “If it was known in advance which people would decide the election, then it would be possible to target them not only with education but also with threats or bribes.”

    This could be prevented with a combination of secret ballot and purdah.

    Regarding the issue of civic responsibility, I think informed decision making is about as much as we can hope for; even Rousseau accepted (reluctantly) that aggregating interests was a (sub-optimal) way of arriving at the General Will (Grofman and Feld, 1988, pp.571-2).

    The Italian Renaissance experiment lasted 1,000 years, and was only ended by Napoleon’s invasion, so I don’t see how Richard Ward can argue that it failed for intrinsic reasons.


  4. Ben, I have several comments.

    First, I think your readers should be aware of the two papers by Amar on lottery voting – Choosing Representatives by Lottery Voting (1984) and Lottery Voting: A Thought Experiment (1995). You refer briefly to one of them in your Ethics paper, but the context here – where the focus is on selecting delegates – is exactly the same context that Amar considers.

    Second, I think there is a lot of ambiguity in the idea you discuss here. Two points of crucial importance that are unclear to me are:

    (1) Are we considering (a) the appointment of a single decision maker with full decision making powers, or will there be (b) multiple lottery voters, each chosen independently, who make decisions together? The former is quite dangerous, I believe, since it would quite often result in decisions that reflect the ideas and interests of small minorities in the population.

    (2) Can the lottery voter (a) select any delegate he or she chooses, or is there (b) a set of qualified candidates from which the voter must choose? The former possibility makes lottery voting a form of sortition, since the lottery voter is allowed to select himself or herself as the delegate. The second possibility, leaves us very close to the current system since the qualification process would serve as a filtering mechanism that assures that only elite candidates are available for the lottery voter to choose from.

    My understanding of the Amar proposal is that he answers (b) to both questions. I am guessing that you are answering (a) to the first question and (b) to the second. Amar’s proposal, I think, leaves us very close to the current system, although it does do away with the relatively minor problem of gerrymandering. Your proposal, as I understand it, is much more radical and quite dangerous.

    Finally, I have to agree with Harald that your arguments in favor of your proposal are unconvincing. How will having decisions be made by non-representative decision-makers result in better policy making? In such a situation, even if the random voter is fully informed (something that is very doubtful in itself), the decisions that he or she makes would reflect his ideas and interests rather than those of the average person. Constraining the decision to a menu created by a competitive qualification process can mitigate the worst of the imaginable consequences of this process, but leaves us, at best, in a situation that is quite similar to the present one.


  5. “Constraining the decision to a menu created by a competitive qualification process can mitigate the worst of the imaginable consequences of this process, but leaves us, at best, in a situation that is quite similar to the present one.”

    Actually it’s a lot worse, because Claudio’s concerns that the franchise lottery would lead to political instability would pale into insignificance in comparison with the dangers of this approach. And if the theoretical justification is awkward, I can’t imagine what sort of rhetoric would be required to sell it to the demos and the existing power elite as a viable political alternative to our present arrangements. Get real.


  6. I was considering recently two implementations of sortition and trying to work through some questions I had with each. To my suprise when I had worked through them, the two options very nearly converged into one.

    First I thought about blanket sortition like proposed by Phillips and Callenbach. The question I was trying to answer was: What if a selected individual doesn’t want the job? I had the idea that maybe this person could put forward an alternative. So, I’m picked, but I would prefer if my brother took the position instead. He roughly has the same values and interests as me anyway. I was also wondering if parties would still exist. Probably they would as less formal organizations with members of common political leanings.

    Second, I was considering the lottery voting of Amar in various districts in order to get an overall legilative assembly. Since he suggested an exclusion threshold, then only parties that get, say, five percent of the total vote would be eligible for his “twirling drum”. But in my mind, there is no point to the randomness then. We might as well go for proportional representation voting. So I considered what if there was no threshold? Well, the ballot might be hundreds of names long since the barriers of organization and funding would be low. But such a ballot is cumbersome. So, then, I thought instead of hundreds of names, we have a ballot with the few big names from the big parties, and then we have a line at the bottom: “Other: _____” You just enter someone’s name and contact. Enter your own name if you want. This ballot then goes into the draw. The selected congress would then have a mix of party people and a number of independent Others.

    With these changes, now these two schemes resemble each other. But the latter pays homage to our tradition of parties and suffrage. It provides an avenue of participation and may appeal to those who typically look at the ballot and think “None of the above”, or “I could do a better job”. Lastly, every election would essentially be a referendum on the party system (or on sortition for that matter.) If the public was to view professional politicians with distate, more people would vote “Other”. If the public disagreed with most proponents of sortition, that positions should be staffed randomly from the general populace, then they would turn out in droves to elect party men and women.


  7. Is this about random balloting and/or Probability-Proportional-To-Size Sampling?


  8. Hi Mike,

    The question of whether the allotted slots should be transferable is rather interesting. Having a single chamber occupied by both professional politicians and amateurs is problematic in my mind. In this situation the professionals have an inherent tactical advantage that may be used to maintain disproportional power in the chamber and even to discredit the entire idea of the citizen legislator.


  9. Agree with Yoram on this. And I think we need to remember that a sortitive representation system would not (unlike elections) focus on persons, so would sit uneasily with an assembly including elected persons (who would naturally claim a mandate, that would be lacking in randomly-selected members). If parties are to continue to have a role it would be a different one from a descriptively-representative chamber. I’m all for pragmatic compromise, but we need to keep a clear focus on the reasons behind different systems for appointing political office-holders.


  10. In case it was not clear, my point was not about “focus on persons”, “mandates” or any such metaphysical entities. The professionals politicians – like any professionals – would simply have the advantage of experience and organization over the amateurs.


  11. It would take an unusual metaphysical system for persons and mandates to be anything other than empirical entities. Persons are (for example) Yoram Gat, Jacob Richter, Keith Sutherland etc. Mandates (in electoral systems) refer to the legal right of the winning party/parties to rule, and are usually (but not necessarily) operationalised in terms of manifesto commitments. A stronger candidate for “metaphysical” (ie non-empirical) status would be an individual “allotted representative”, as there is no such creature in the singular case — it’s a category mistake (an inversion of Ryle’s example regarding the ontological status of “the University”). In addition to experience and organisational advantages, professional politicians would benefit from the right to be present as mandated individuals, whereas descriptive “representatives” have no such right. This has nothing to do with metaphysics, merely unpicking the meaning of the linguistic term “representation”. Whatever sins linguistic philosophy may be guilty of, it’s certainly not idle speculation on how many angels can dance on the head of a pin.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: