Skip to main content

Open for Debate

How to vote well

26 August 2019

It seems elections are everywhere at the moment. Following on the heels of EU elections that for the first time saw turnout increase, the US is gearing up for a Democratic primary packed with candidates and, eventually, for a Presidential election. Meanwhile, the UK is bracing for the possibility of a general election, if the current government is unable to form a coherent view on Brexit that will command the support of Parliament. All of which raises the question: what’s a voter to do?

The first thing to do is appreciate the stakes: how each of us votes determines in aggregate the very rules we have to live by, and as such have far-reaching implications for voters and non-voters alike. So those who vote owe it to everyone else to vote well.

At a minimum, this means not voting on the basis ignorance about the nature and implications of the political policies involved. And if the past decades of research on public opinion has shown anything, it is that political ignorance is widespread. As Scott Althaus notes, ‘If ignorance is bliss, then the pursuit of happiness seems alive and well in American society’—and, unfortunately, the Americans aren’t outliers here.

Of course, in any relatively large electorate, any given vote won’t make that big of a difference. That’s why some political scientists have suggested that voters are rationally ignorant. But what’s individually rational might still be irrational for the group. My not recycling won’t ruin the environment, but it doesn’t follow that I’m not obligated to recycle—and something similar goes for ignorant voting. Such voting is a ‘collectively harmful activity,’ as Jason Brennan puts it.

So your obligation to others to vote well stands. But what I want to do now is point to an additional reason that you ought to vote well. You would let yourself down by engaging in ignorant voting, quite independently of whether (large-scale) ignorant voting might also be bad for others.

Here’s why. The choices you make depend on what you know. In particular, not knowing relevant information can have you endorse options you would have rejected, had you known all of the relevant facts. Consider an example. Say I get to pick between two environmental policies, A and B. A is great for the environment, B is terrible for the environment. But I have it the wrong way around so, while I care deeply about the environment, I end up picking B over A.

Here’s a question: do I prefer B over A? No. As Robert Goodin points out, under circumstances of ignorance, ‘we can serve a person’s “real” preferences only by censoring the misleading indication of his preferences that is revealed in his choices’. So, the fact that I chose B shouldn’t lead us to think I want it, given that I’m mistaken on the facts.

Indeed, there’s a general philosophical lesson in this: my true preferences are given by what I would have preferred, had I had all of the relevant information. Differently put, my true preferences are my fully informed preferences.

This also tells me something about how I should vote, in so far as I want to vote in accordance with my (true) preferences. I should vote the way I would’ve voted, were I fully informed. By doing that, I minimise the effect of ignorance on my voting, and to that extent vote well—or at least better than I otherwise likely would have.

Let’s take stock. You have two reasons to vote well—you owe it to others, and you owe it to yourself—and the way to do it is to vote the way you would’ve voted, were you fully informed. But how do I know how I would’ve voted under these circumstances?

It might be thought the answer is obvious: inform yourself! The problem is that it’s really difficult and time-consuming to do so, and trying by no means guarantees succeeding.

So this is where I have good news for you: you don’t need to. We can estimate how you likely would have voted, were you fully informed, through statistical modelling. If you’re a UK voter, you can try it yourself here:

How does this work? Models like these operate on large sets of survey data that enable us to identify the relationship between demographic information and voter preferences, as reported in a context where participants also answer a number of civic test questions about political matters. On the basis of that data, we can then estimate what preferences people of particular demographic profiles would likely have reported, were we to hold everything about them constant, but increase their score on the civic test to the maximum.

Why think that scoring high on such a test means being informed? Here are two reasons:

First, these tests typically concern matters such as the holders of key political positions, and the basics of the political platforms of the main political parties, all of which would seem crucial to making anything resembling informed political choices.

Second, to the extent that the tests concern more esoteric matters, knowing the answer to such questions is likely correlated with knowing a great many other things that are directly relevant to making wise political choices. For example, someone who knows the names of several foreign leaders, or of obscure legislative procedures, will likely know these things, not as a matter of isolated pieces of trivia, but on account of knowing a great deal about the political world.

How do we know that these models are accurate in terms of their estimates? As George Box famously pointed out, all models are wrong, but some are useful. And in the case of these models, their utility becomes clear when we survey the options we’re faced with when trying to vote well.

As already noted, we can try to inform ourselves on large amounts of political issues and facts—but in addition to being time-consuming, it also doesn’t guarantee that we succeed even if we invest the time. So for most of us, this is simply not a viable option.

By contrast, models of the kind introduced above in effect provide a low-cost tool for learning how an informed version of you likely would’ve voted, and you’d be wise to factor that into your voting choice. If I’m right, you not only owe it to others to do so; you owe it to yourself as well.

Photograph by Kristoffer Ahlstrom-Vij


  1. Ernest Bornheimer

    Utterly brilliant! I’ve thought for some time that each of us has an obligation to know the truth, and to act accordingly. I never imagined there would be a way for an ignorant person to act as if they were well-informed. We need something similar in the US, for every major election. Although……….now that I think about it, we know that reasonable people disagree on important issues, based not just on their knowledge, but also on their values. So, it might be good to include measuring the core values of each survey respondent, perhaps based on something like Jonathan Haidt’s moral foundations theory.

  2. Daniel

    This is a fascinating suggestion and proof-of-concept. It seems that the advice here, taken to its logical conclusion (if the model were tested/improved), is that each uninformed voter (anyone who does not score close to 100% on the test) enact a “voting rule”: Vote according to the largest weight outputted by the model. Thinking about a world where voters act this way, I have a few critiques:

    1) If a voter follows this advice, it reduces them from a rational agent to a representative of the plurality of the (most-informed) members of their cohort*. Yet a voter may have personal beliefs and values that differ from that of those who share my race, income level, or other markers of socio-economic status. The previous comment does seem to ameliorate some of this concern, by including “values” questions in the survey.

    (*I imagine that the model may not explicitly act this way, by assigning voters to “cohorts,” but a regression model of any kind would be essentially equivalent, assigning a voter to an idealized cohort constructed largely by people who answer questions similarly to them.)

    2) If this type of advice is taken seriously by voters, then it renders an election susceptible to “conceptual gerrymandering.” In other words, if uninformed voters are committed to adopting this rule, then they are voting the way that the plurality of the informed members of their cohort vote. This makes election results sensitive to how the cohort is defined, or, equivalently, to how the regression model is constructed. An unscrupulous survey-designer could work quite similarly to how modern gerrymandering map-drawers do, by simulating our thousands of conceptual “maps” and selecting the one with the preferred outcome. (Inclusion of “values” questions could introduce even more flexibility into this process, i.e., make it easier to gerrymander.)

    2a) A less vulnerable voting rule would be one that commits uninformed voters to vote according to a lottery based on weights outputted by the model, rather than one that selects the plurality. This rule assigns some chance that the voter will then actually “act like an uninformed voter” (if we accept the premise of the model). If no one voting choice receives a majority of the model-assigned weight, this rule would “condemn” a majority of voters in the cohort to “act uninformed.” This consideration seems to exacerbate issue #1, essentially that the uninformed voter is acting as automaton rather than agent.

    3) Informed-ness likely correlates with certain values and preferences, and it likely corresponds to various aspects of educational opportunities not captured by socio-economic survey. (Could a model-based voting rule correct for this problem?) While it is possible that encouraging voters to adopt the values/mindset of their more-informed neighbors would be a good thing, that encouragement departs from the stated purpose of this model, which was to empower voters to enact their *own* preferences more effectively.

    4) In a world where voters took this model and voting rule seriously, how would the electorate respond to new circumstances, new candidates, or new positions taken by parties? It seems we would need an “up-to-the-minute” panel of informed voters.

    Now here is something interesting: When I first read this post, I immediately recoiled, probably thinking about some combination of critiques 1 & 2, above. As I write this, imagining how a world with the proposed voting rule might operate, it actually seems not so far off from our own world (speaking from perspective of a U.S. voter): Early results (polls, fundraising, primary election outcomes, the preponderance of news coverage/editorials) are driven strongly by a small subset of the electorate, generally people who are very motivated, wealthy, or situated in special positions of influence. (Is this subset more informed than the average voter? Perhaps, if merely by virtue of spending a lot of time with early phases of campaigns.) Many voters with less time on their hands can and do base their preferences on these early results, “outsourcing” some aspects of their decision-making to this subset. And we do in fact have “conceptual gerrymander-ers”— poll-writers and folks who watch polls and try to determine the “momentum” of a race, or the strategic way to vote. I tend to trust (or at least read charitably) certain reporters who seem to share my values and patterns of thinking; this is one type of outsourcing I do.

    What your model seems to do, then, is less to present a radical way of thinking, but more to idealize a part of the way things actually work…

Comments are closed.