Voting by the Dumb and Ill-Informed

Beyond a certain point of negative understanding, of course, it is better for someone not to vote.

Joseph C. McMurray discusses the Marquis de Condorcet’s mathematical analysis favoring decision-making by larger numbers of people.

An interesting, if somewhat uncommon, lens through which to view politics is that of mathematics. One of the strongest arguments ever made in favor of democracy, for example, was in 1785 by the political philosopher-mathematician, Nicolas de Condorcet. Because different people possess different pieces of information about an issue, he reasoned, they predict different outcomes from the same policy proposals, and will thus favor different policies, even when they actually share a common goal. Ultimately, however, if the future were perfectly known, some of these predictions would prove more accurate than others. From a present vantage point, then, each voter has some probability of actually favoring an inferior policy. Individually, this probability may be rather high, but collective decisions draw information from large numbers of sources, mistaking mistakes less likely.

To clarify Condorcet’s argument, note that an individual who knows nothing can identify the more effective of two policies with 50% probability; if she knows a lot about an issue, her odds are higher. For the sake of argument, suppose that a citizen correctly identifies the better alternative 51% of the time. On any given issue, then, many will erroneously support the inferior policy, but (assuming that voters form opinions independently, in a statistical sense) a 51% majority will favor whichever policy is actually superior. More formally, the probability of a collective mistake approaches zero as the number of voters grows large.

Condorcet’s mathematical analysis assumes that voters’ opinions are equally reliable, but in reality, expertise varies widely on any issue, which raises the question of who should be voting? One conventional view is that everyone should participate; in fact, this has a mathematical justification, since in Condorcet’s model, collective errors become less likely as the number of voters increases. On the other hand, another common view is that citizens with only limited information should abstain, leaving a decision to those who know the most about the issue. Ultimately, the question must be settled mathematically: assuming that different citizens have different probabilities of correctly identifying good policies, what configuration of voter participation maximizes the probability of making the right collective decision?

It turns out that, when voters differ in expertise, it is not optimal for all to vote, even when each citizen’s private accuracy exceeds 50%. In other words, a citizen with only limited expertise on an issue can best serve the electorate by ignoring her own opinion and abstaining, in deference to those who know more. …

This raises a new question, however, which is who should continue voting: if the least informed citizens all abstain, then a moderately informed citizen now becomes the least informed voter; should she abstain, as well?

Mathematically, it turns out that for any distribution of expertise, there is a threshold above which citizens should continue voting, no matter how large the electorate grows. A citizen right at this threshold is less knowledgeable than other voters, but nevertheless improves the collective electoral decision by bolstering the number of votes. The formula that derives this threshold is of limited practical use, since voter accuracies cannot readily be measured, but simple example distributions demonstrate that voting may well be optimal for a sizeable majority of the electorate.

The dual message that poorly informed votes reduce the quality of electoral decisions, but that moderately informed votes can improve even the decisions made even by more expert peers, may leave an individual feeling conflicted as to whether she should express her tentative opinions, or abstain in deference to those with better expertise. Assuming that her peers vote and abstain optimally, it may be useful to first predict voter turnout, and then participate (or not) accordingly: when half the electorate votes, it should be the better-informed half; when voter turnout is 75%, all but the least-informed quartile should participate. …

If Condorcet’s basic premise is right, an uninformed citizen’s highest contribution may actually be to abstain from voting, trusting her peers to make decisions on her behalf. At the same time, voters with only limited expertise can rest assured that a single, moderately-informed vote can improve upon the decision made by a large number of experts. One might say that this is the true essence of democracy.

His conclusion seems to accord with observed results. Ordinary people are surprisingly well able to correct the follies and delusions which too commonly afflict the experts and elites, but there are also people so clueless that they are always going to vote wrong.