A Statistical Voting System

Abstract: a new voting system is proposed in which, rather than 100 million people voting, a statistically sampled random fraction, say 1 million or 100,000 people, vote. These people, realizing how much their votes count, take more care in their voting.

Imagine a perfect democracy. For every election, every voter would spend time learning about all of the issues, learning the stands of all of the candidates, and then cast their carefully considered ballot on election day. Of course, we don't live in a perfect democracy, but in an imperfect one where few voters have time to learn all of the issues, and election turnouts, especially for primaries or local elections, are small. Not only that, but consider the costs of such a perfect democracy: it takes time to vote, but even more time to study issues and candidates: days or more to do it well, over the course of a year, for all of the elections from local to national.

What if instead of a perfect democracy, we could figure out what the results would be in a perfect democracy, but without all of the effort: the results of an ideal world, without the costs. Finding these results actually isn't very difficult or costly. If we were to randomly select 1000 voters, give them a few days to study the issues, and then have them vote, we would get a very good idea of the results in our perfect democracy. Unless the results are very close, it will be highly probable that our randomly selected voters found the same winner as would be found in our perfect democracy. If the results are too close to be reliable, we could hold another election, with more voters, hoping for a statistically dependable result.

This simple basic idea is almost workable as it stands, but a few modifications will make it better. First of all, depending on the size and importance of an election, 1000 voters probably isn't safe. One certainly wouldn't want to put something like the election of the president of the United States into so few hands (although our electoral college system does just that). But a larger number of voters, perhaps 100,000 or 1,000,000 would be enough, still getting the advantages of most people not needing to vote.

One might object that this system is superfluous, given that we already have a system of proportional representation. While it is true that the current system saves considerable time and effort over one in which each person voted on every issue, it is also true that we could probably do even better with the system I propose. Even though no one spends every day working on legislation besides a relatively few legislators, lobbyists, and aides, many of us still spend substantial time trying to figure out who those legislators should be, time that could be saved in a statistical voting system.

A statistical voting system has other advantages (advantages which will certainly be its downfall). The first advantage is everyone will be equally represented, in a way they are not today. The second advantage is that it will be feasible to educate voters in a way which would otherwise not be possible. The final advantage is a campaign finance reform advantage: holding an election and educating voters becomes so much less expensive that we can easily afford to move to completely state-financed elections.

In the current system, not everyone votes, and the people who vote are not represented evenly throughout the population. People who have more time, or vested interests, are more likely to vote. Thus, senior citizens vote, in droves, while busy working parents with children are less likely to have the time. This has impacted our society adversely: social security reform is considered the kiss of death in Congress. Similarly any attempt to require elderly drivers to prove their continued ability to drive is hopeless. Other groups that benefit are municipal and state employees, who are more likely than most to attend town meetings, or to pay close attention to elections, and thus to influence them for the benefit of government employees over average taxpayers. In a statistical system, one would hope that everyone selected to vote would understand the importance of their task and take it more seriously. Given the rarity of the obligation, they would at least be more likely to do so. If you are only asked to vote once every 20 years, or less, and your vote counts significantly more, it is an opportunity you are much more likely to use than if the opportunity occurs twice annually.

The next advantage is that it will be feasible to educate voters in a statistical democracy in a way which would not even be possible in our ideal democracy. Our relatively small number of voters would have opportunities, such as meeting in medium sized groups with the candidate, or speaking personally to people who knew the candidate's views, or receiving books detailing the candidate's positions, that would be infeasible were everyone to be an educated voter.

The third advantage of this system is a campaign financing advantage. It will be much less expensive to be a candidate: a much smaller number of people will need to be reached, and so running for office will become much cheaper. In such a system, we could have complete state financing of elections, eliminating the disproportionate influence of those who can currently contribute heavily to elections, or the advantage of rich candidates who can extensively finance themselves.

Unfortunately, these advantages will probably be the undoing of a statistical voting system. Vested interests will not want to allow the system to be changed in ways that lessen their current undue influence. I assume that both the AARP and big business interests (the NRA) will vehemently oppose this plan (appealing to tradition, democracy, and the founding fathers, rather than finding problems with the proposal itself.)

One can imagine some variations on this system. One alternative which might satisfy some with objections is to make the system optional. Each person can either participate in the system statistically, or normally. A statistical participant has perhaps a one in one hundred chance of being chosen to vote, but his vote counts for one hundred. Others can elect to participate conventionally; they can always vote, and their vote counts as one. Conventional voters might argue that such a system is unfair, since they get only one vote versus one hundred for the statistical voters, but the statistical voters only have one one hundredth chance of voting: one vote on average.

A combined system offers other advantages. If we allow people to join a group, from which one person is randomly selected, it may ameliorate fears and decrease the chance of an incorrect election output. For instance, a husband and wife with similar views could join a two person group; one of the two would be chosen at random, and given two votes in the election. Alternatively, a staunch democrat could join a group of one hundred staunch democrats. Only one of them would be randomly picked to vote. Segmenting the system like this decreases the chance that a bad sample of the population will accidentally be chosen.

There are some other options to consider. We can either have those selected to vote vote on every ballot issue and every candidate in an election; or we can have each selected person vote on only one or two issues. Personally, I favor having each selected person vote on just one issue. This will increase the number of people selected to vote, and allow a greater feeling of enfranchisement.

There are some potential objections to this system. The most important is that with too few people voting, it might be possible to intimidate them or otherwise unduly influence them. We need to give the set of voters some way to remain anonymous so that they cannot be intimidated. The combined system also presents a good escape valve: should the system be corrupted, individuals could elect to vote, and switch the system back to a non-statistical format.

The big question then is, could such a system ever become widely used. If the theory of the system becomes known, there is no reason that some newly formed town of busy young people should not select such a system for their self-government, at least experimentally. If experiments at the local level are successful, then perhaps the system will spread, both geographically, and to higher government, such as County or State. If it is finally seen by all that the system is more convenient, fairer, and leads to better government, then it could one day be adopted even at the federal level.

Send me email: goodman@eecs.harvard.edu
Back to my home page: http://www.eecs.harvard.edu/~goodman

Append your personal comments to the end of this page: