For me, “voting system” implies the wrong question

I recently expressed frustration with something I called “voting-system-speak.” This is my attempt to define (or at least engage) that and say why I find it frustrating. The crux, I think, is that many (but not all) who think about electoral reform are coming at the subject from a social-choice perspective. Another key point, which is related, is that we may be able to sort reform proposals based on their handling of the problem of social choice. One class of reforms turns to parties. Another class turns to voters. And the former strikes me as more realistic, based on insights from the field of political behavior.

My points here are not entirely new. Matthew Shugart has a blog post on his own views of “voting system,” which he takes to mean ballot type or format (Rae 1971). I agree with those views. From the voting-system perspective, IRV, the Bucklin system, ranked-pairs, and many other options are all the same thing. Yet any voting theorist can tell you they are not, even though they all aim to elect a single winner.

On the broader issue of social choice, see a classic essay by Iain McLean (1988) and a more recent encyclopedia entry by Josep Colomer (2010). Part of that is worth quoting:

A subsequent line of research promoted by social choice theory attempts to evaluate the relative performance of different institutional and decision rules in satisfying individuals’ preferences and producing acceptable social choices. The ‘impossibility’ theorems tell us that it is impossible to guarantee fair and stable social choices with any rule. But certain rules tend to produce inconsistent choices more frequently than others.

[…]

In contrast, in multi-party elections producing coalition cabinets, different issues, roughly corresponding to different government portfolios, can be dealt with separately on single-issue ‘spaces’. Also, in regimes of separation of powers, each separate election for a different office can focus on one or a few issues and favor the consistency and stability of social choices. In these institutional frameworks, each issue can be the subject of a broad multi-party or inter-institutional agreement around the median voter’s position, which can preclude drastic changes and induce policy stability in the mid- or long term.

Why “voting system”?

I think there are two kinds of people who gravitate to this term. I used to be one of them: a believer in Duverger’s Law as taught in Political Science 101. So, for a while, I described myself as a “voting systems” researcher simply because it has fewer syllables than “electoral system.”

The other kind seems to come from the impossibility tradition. How do we find a rule, in Colomer’s terms, that maximizes production of “consistent” social choices? Or, how should we suck information out of voters’ heads, then use that information to elect the best possible monarch? (The fact that the information can be used in different ways already starts taking us in the direction of “electoral system,” by introducing what Rae called the “allocation rule.”)

Note that the second perspective fixates on single-seat elections. The reformer Clarence Hoag (1914) appears to have dealt with a similar intellectual current during the Progressive Era (emphasis in original):

We now come to a distinction of the utmost importance in our political thinking. Voting, which is at the basis of our modem democratic governments, has two distinct objects to carry out, and they must not be confused with each other. One object of voting is to make decisions, for example as to which of several competing policies shall be carried out, or as to which of several candidates for an administrative position shall be selected […]

We now come to the second object of voting. This is not to make decisions at all, but to make up a body fit to make them on behalf of all the voters. Now, though the principles of democracy require that the decisions made in such a body, like those made directly at the polls, should be made by majority voting, they by no means require that the body itself should be made up by majority voting.

Hoag was trying to get people to think about legislatures, i.e., politics. I have not read anything by him that suggests understanding of coalitions. (That may have been part of the problem with real-life use of STV in the U.S.)

Two kinds of reform

I used to mention doing a paper on this in job applications, but here it is instead, banged out on a blog.

One set of reforms takes coalition politics as given. The legislature as an institution is front-and-center from this perspective. These reforms respond to the question: how do we produce a coalition most people can live with? Enter party-list proportional representation. I highly recommend this essay by van der Hout and McGann (2009), which makes the case in social-choice terms.

Another set is trying to solve the social-choice problem at the level of the voter. We are going to elect one thing, which will make decisions for us as a unitary actor. What reforms are in this set? Well, one answer could be: every single-seat reform on offer.

It is worth noting that Lewis Carroll seems to have been thinking about coalitions. His solution to the problem, in the context of a single-seat election, was: drop candidates and make people re-vote until a Condorcet winner emerges (Green-Armytage 2023).

Accounting for frustration

I am in the position of doing work that speaks to advocates. Some of them want a multiparty system. Others want to break up the two-party system and describe the alternative as some form of moderation. Still others just want to see enacted this-or-that invention.

So, when I recommend a system like party-list PR, it makes sense that some will not be happy. The recommendation ends a conversation about “the best voting system.” It also does not imply a clear solution to the “spoiler effect” (i.e., strategic desertion), which means I disappoint those who are looking to induce a multiparty system. For example, I often hear: “what is your single-seat solution?” I don’t have one, and I don’t see why we need one — at least at an abstract level.

Part of why I’m such a downer is the attention I pay to what I see as reality. Behaviorists have taught me several things about voting. These include the importance of social-group attachments, party identification as one such attachment, the fact that most people are too busy to weigh/array/assess undifferentiated candidates, and so on. Hence a line I wrote two years ago:

There’s a tendency in reform circles to ask too much of voters. I’m thinking here of elaborate schemes like RCV, STAR (Score Then Automatic Runoff), and approval. Ballot reforms like these basically ask voters to pick a better coalition. One-vote flips that around — give voters representation, then have their representatives form the coalition. It’s a lot more realistic.

And hence my abandonment of the term “voting system” for “electoral system.” We should take the single vote as given, then use other components of the rules to get us closer to whatever outcomes we might agree on.

I will not persuade everybody, and that’s okay. I also may be totally wrong in my understanding of some things above. My goal here was to lay out that thinking.