There are different ways voters can be rational or not

The rationality hypothesis works sufficiently well for some types of choices, terribly for others

Refusing to see, hear or speak of voter irrationality

Imagine you have a good friend Bob from business school with whom you have mostly lost touch. Then a mutual acquaintance, Alice, who has recently been with him, tells you Bob said:

1. That he ran into some pretty convoluted legal complications with his business and

   a) that he picked a lawyer and that the lawyer is handling it;

   b) that he’s handling the legal issues himself because he doesn’t trust lawyers

2. That he was going through some big health issue with multiple symptoms which gravely affected how he felt and

   a) that he picked a doctor who made a diagnosis and prescribed a treatment;

   b) that he’s treating everthing himself because he doesn’t trust doctors

3. That he decided his firm should build the next skyscraper in the city, full of innovative choices, and

   a) that he picked some big-shot engineer to design the project;

   b) that he’s designing the project by himself because he doesn’t trust engineers.

In any of these cases, you would be more concerned if your friend picked (b) instead of (a). The fact that we trust experts to make choices in our interest, choices we would not be able to make as well ourselves, is such an obvious trait of our lives that it feels weird to even point it out. This is the case even in the examples above where many aspects about the quality of the service are often unclear to the client even after the service was provided. By extension, we tend to be worried about people who, being so distrustful of experts, are willing to make all choices in their lives without talking to specialists. In other words, we think that, when faced with some types of problems, people are generally rational enough to pick an expert to help them achieve their objectives, but are far from rational enough to solve those problems directly (unless they are experts in the specific issue themselves, of course).

In political discourse, however, this distinction is almost never made. Either we treat voters as presumably rational on everything (which is the overwhelmingly prevalent view) or we treat them as all-around irrational. But we really should distinguish between these two types of rationalities. Voters are hopelessly irrational when choosing specific public policies, but are rational enough to pick representatives tasked with making public policy decisions. To clarify the terms, I should specify that when people use the word “rational” in this context they almost always imply a rational, self-interested person who casts their vote in order to increase the probabilities of their candidate or policy winning the election and with that realizing their goals. If they vote because they like to, or they feel it is their duty, independently of the consequences, they are of course still rational in a sense, but not in the usual meaning. A good deal of the current crisis of democracy - and of some past ones - derives from failing to realize this distinction.

Two of my favorite books are “The Myth of the Rational Voter”, by Bryan Caplan, and “The Myth of Democratic Failure” by Donald Wittman. You may find this confusing because the books seem to contradict each other. But in fact the books, to a great extent, talk about two largely independent phenomena - how voters can be irrational, and how democracy can work. Rational self-interested voters would be of great help for democracy to work. But it turns out it is not a necessary condition, both theoretically and empirically.

The failure to distinguish between the two is widespread. On the one hand, democracy enthusiasts will come up with ever more implausible explanations for why voters are indeed rational. On the other hand, the literature on voter irrationality is hardly enthusiastic about democracy. Caplan’s “Myth of the Rational Voter” states in its introduction: “This book develops an alternative story of how democracy fails.” Jason Brennan’s book, in turn, requires only knowing the title to reveal its skepticism: “Against Democracy.”

I do think highly of democracy - of the parliamentary kind, of course. In fact, even Brennan himself is fond enough of it that he wrote a kind blurb for my book. I believe we can square these differences.

The (ir)rationality debate

Everyone wants to save democracy - for good reason. Consistently democratic countries provide a kind of abundance, peace, security, and freedom without precedent in history. If democracy is “government of the people, by the people, for the people”, then many believe it is simple logic that it presupposes the people always know best. Hence, any type of suggestion that they might not know best on any circumstance must be an illusion, or the works of those who deliberately want to undermine democracy for their own benefit or pernicious ideologies. The problem is, the evidence for voter irrationality is abundant, as expected by theory. So we come up with ever more elaborate justifications for the evidence of irrationality we see and explain how, really, the evidence actually shows voters are being perfectly rational.

The view that voters are irrational is not hard to acquire. Indeed, it is the default position of people who only start to think about the issue. Our institutions then educate citizens that they are mistaken and that voters are perfectly rational. The hypothesis becomes an article of faith, so much so that trying to refute becomes much harder than the theory and evidence would suggest.

Let’s start with the theory. Regular democratic discourse, as we will find in reputed newspaper coverage, op-eds and the equivalent of the old “Twitter blue checks”, treats voter behavior as an application of the homo economicus hypothesis for public policy. A voter will examine their interests, see which policies will best tend to them, and accordingly vote for the policy itself (if it is an iniative, referendum) or for the candidate committed to implementing it. “Sure” - these thought leaders may think - “it is only an approximation, jus like homo economicus itself is an approximation. But it works as a model.”

However, a rational self-interested voter theory can’t work, even as an approximation. We have known this for ages. A rational self-interested person wouldn’t vote at all if what they want is to influence policy in the direction of their interests, because the probability that any one’s person vote affects the outcome is extremely small. While most people will concede this point, they will immediately retort - “yeah, but if everyone thinks their vote matters (or doesn’t matter), the outcomes do change.” This is true, but doesn’t alter in the least the fact that, whatever everyone else is doing, a purely self-interested rational person would not vote.

Considering that the theory posits that the rational self-interested voter is a paradox, we could very well skip dealing with the evidence, since no amount of evidence can make a contradiction true. But let’s concede them a first tweak. Let’s check the hypothesis - prima facie unlikely - that voters are not only self-interested, but also rational about everything, except the expected value of voting. That is, they assess policies pretty well and can see which ones would benefit them. But they somehow forget they are rational and that their vote is not really going to make those beneficial policies more likely in any meaningful sense.

First, on self-interest: is it true that voters mostly support candidates and policies they perceive to be in their interests? The answer is no. That this empirical finding surprises people is puzzling. I will bet that in the circle of people you walk in, no one chooses a presidential candidate based on how much they will personally benefit, and they would probably be reprimanded if they did so. But we assume (our assumptions always favor the rational self-interested voter hypothesis, you will see) that a significant majority of the voters we do not know, however, are doing that. Fair, judging general attitudes from the behavior of one’s close circle can be misleading sometimes. But the evidence shows that, in this case, what is true of your circle is true of the general populationt.

On rationality, the evidence is also negative. As Garret Jones shows in 10% Less Democracy (a good book with an unfortunate name, as I’ve said before), among countries with good institutions (good enough that the government “experts” really are experts), those which implement policies preferred by the experts instead of by “the people” fare better in those areas. The evidence can also be seen with the naked eye. As I write this, it is becoming consensual that the cumulative inflation during the Biden administration was decisive for the Democrat’s loss and Republican win. Any mainstream economist will recognize inflation does have some economic costs (tax distortions, infefficient wealth redistribution, menu costs, shoe-leather costs, etc). Those costs, however, are relatively small in the grand scheme of the economy, particularly in light of a less than two-digit annual inflation even at its maximum level. The reason people are so upset about inflation is that they believe the increase in the prices they face - for groceries, gas, etc - is due to bad policy, but that the increase in their own earnings is due to their own merit. In other words, because they are irrational. Many economists have pointed this out on Twitter, but instead of acknowledging the irrationality, many accounts are mocking them for failing to perceive that the people know better.

That should be it. The rational, self-interested voter hypothesis implies a contradiction, hence it cannot be strictly true. Even an ad hoc, implausible, modification of the hypothesis fails to be supported by evidence. It should be simple, but it is really hard to get people to accept that. Voter rationality seems to be “unfalsifiable”, not in the Popperian sense that a test for its validity cannot be devised, but in the sense that any evidence of irrationality will be justified with an ad hoc hypothesis.

Aggregation - the miracle that didn’t happen

You might know this story already. The statistician Francis Galton went to a country fair. The people there were playing a game in which the best guesses on the weight of an ox in display would win prizes. While there were some people who presumably were experts, such as butchers, many had no particular knowledge on the issue. Galton wanted to test how competent the people were on average, so after the game was finished he collected all the tickets and found that the average guess was 1,197 pounds, while the measured weight was 1,198. Almost perfect.

The reasoning, hence, is that maybe voters aren’t particularly rational as individuals, but if their mistakes are independent from one another, they cancel each other out and the median vote - the average of the voters - is a wise one. But the hypothesis has problems.

It still assumes self-interested voters. But even voters who are minimally rational - say, who think their vote is ten times more potentially decisive than it actually is - would still have no self-interest motive to vote. And we have evidence that they are not, as mentioned above. One could argue: “fine, so voters aren’t self-interested. The fact that they do vote shows they care, though, so we can assume they are trying to maximize their society’s total welfare. Aggregation still works”. I would reply that even if that was correct, we still needed to stop analyzing politics as if voters were self-interested. I would also point out that, unfortunately, giving up self-interest to save rationality doesn’t work either, which I will discuss below. Even if it did save rationality, aggregation would still fail, though.

People’s errors about policies are nowhere close to independent. As Bryan Caplan compellingly describes, the average person has very important biases about policies. Compared to experts, regular people are too suspicious of the benefits of trade, too suspicious of the benefits of interacting with foreigners, don’t understand that you want to save as much labor as possible instead of increasing it, and are too pessimistic about how the economy fares. Maybe you think that Bryan Caplan is just a neoliberal and of course he would suggest the people are wrong on these issues, but in fact he is the wrong one. But whatever your worldview, I’m sure I can find topics which you think should not merely reflect public opinion of the day - climate policy, human rights, health policy (such as vaccination), death penalty, funding of the arts, national defense. No one consistently holds that the majority opinion of the public should be followed at all times. What I see is many people rationalizing that for whatever topic they happen to agree with public opinion, voters should be sovereign, and if they disagree, they find some reason for why, on that specific issue, public opinion is not a good guide.

This means that the resulting policy from the aggregation of all preferences will be far from optimal. The miracle was an illusion. Still, academics have been trying to provide several lifeboats. I’ll go through a few. My point is far from being exhaustive, but to illustrate how much we bend to try to save the hypothesis.

“Actually, giving lots of power to populist presidents is good for the majority of people”

Daron Acemoglu, James Robinson and Ragnar Torvik (ART) are no fans of uncheked executives. In “Why Nations Fail”, Acemoglu and Robinson exhalt the virtues of the 1688 Settlement: "The Glorious Revolution limited the power of the king and the executive, and relocated to Parliament the power to determine economic institutions." Robinson and Torvik, in turn, have a paper arguing that “politicians in general and particularly political leaders, capture more rents and provide fewer public goods under a presidential system compared to a parliamentary one” and, consequently, “political leaders prefer to be presidents rather than prime ministers”.

As adherents of rational voter theory, they had a puzzle to explain, because voters often approve removals of checks and balances - notably, parliamentary oversight of the executive, the very essence of the Glorious Revolution so praised by the 2024 Nobel Prize winners. Unfortunately for the hypothesis that parliamentary oversight of the executive is central to what “good institutions” are, dogma has it that the rational voter can’t be challenged. Parliamentary oversight has to walk the plank (at least “under certain circumstances”, as the paper states).

ART build a model to explain that voters are acting in their own interest removing those checks. The reasoning is that systems with checks on the executive have lower opportunities for rent. Presidents are hypothesized to have a “political bliss point” with the right balance of personal rents and redistribution to the poor. If a president does not have to face checks on his power, he can capture all the rents he needs to be satisfied in that dimension, and then prioritize redistribution. With checks and balances, the rents captured by the president are too small to achieve his bliss point. So powerful lobbies seeking to prevent redistribution can bribe the president and get him closer to the bliss point. In ART’s model, there is one president (naturally) but also one legislator. The electorate - aware of all of this - votes to remove the checks on presidential powers so that he can achieve his bliss point and implement redistribution.

I find these assumptions simply wild. I can’t unfortunately have hard data on whether when voters approve more powers to a president they do it thinking that this way he will not be limited in his rent seeking opportunities and will be satisfied enough that he won’t be as easy to bribe. I don’t think anyone ever formulated such an hypothesis before ART. If anyone did, it certainly isn’t widespread. I have followed presidential politics all my life, though. And I am confident that voters who want to give more power to a president are almost always convinced he is the kind of president that is not in politics for rent seeking, or at the very least is less of a rent-seeker than the alternative. In the rare cases where voters admit they are voting for a corrupt candidate, what is emphasized is not that while the president may pocket a lot of money he also redistributes it, what is emphasized is that he pockets the money but “gets things done”.

The evidence provided by the authors in favor of their model - in the form of selected quotes - only shows that populists run with an anti-elite agenda. But that is unsurprising as can be, given that being anti-elite is often the defintion of populism. The cited speeches do not argue that unless the president’s power is unchecked, corruption will be too tempting for the president. They argue that unless their power is unchecked, they will not be able to implement the measures they are absolutely committed to implement if they can.

“Give me unlimited powers to pusue your interests, otherwise I will have no choice but to take money from powerful lobbies” said no politician ever

Another strange assumption is that there is only one legislator. Even when they “relax” this assumption, they don’t work with the assumption that a majority of legislators must approve a measure, but that all legislators have some probability of being the one legislator deciding by himself. This matters a lot, because a division of powers - having multiple politicians with the same legal powers capable of approving something by a majority of them - is a central mechanism for why parliaments can reduce rent-seeking. Having to bribe tens or hundreds of people instead of just one is either much more expensive (meaning there will be fewer profitable bribe situations) or implies much reduced bribes. As per ART’s model, though, politicians have rent-seeking but also altruistic motives. Facing a diminished bribe for approving the measure which goes against their altruistic motive, there is a good chance at least a majority - any majority - will choose the altruistic choice.

But probably the weirdest assumption is that populist policies are good for the majority of the population. When Tanya Singh reviews the literature on the issue, she provides this brief conclusion: “To conclude, the few studies conducted so far have highlighted populism as a damaging policy that has had lasting negative impacts on economies the world over. Though economic and political theory have long predicted the consequences of populism, there is a clear lack of empirical evidence when it comes to studying the phenomenon. Given that populism, both left-wing and right-wing, is on a rise, it is absolutely essential that more research is devoted to studying its result and consequences.” I believe the empirical effects of populism should be more studied, but maybe the reason they aren’t is that economists widely regard them as clearly harmful, even harmul by definition sometimes. “Looks good, but is actually not” is pretty close to what we understand populism to be. Directly related to ART’s hypothesis, one specific paper worth mentioning: "The Political and Economic Consequences of Populist Rule in Latin America" by Houle and Kenny (2016). Besides identifying several negative consequences for populism, they find that redistribution does not increase, even under left-wing populists, contradicting ART’s hypothesis.

In sum, for their model to work they use an unrealistic assumption for why people vote the way they do, an unrealistic assumption for how governments would work with more or fewer checks on the executive, and an unrealistic assumption for the benefits of the outcome.

How would the rational irrationality framework explain voters preferences for removing checks and balances? By noting that a stated preference for increasing the power of presidents is very common all over the world, particularly in presidential countries - in any situation, not merely “under certain circumstances”. If a president becomes dominant enough that he gets to ask this type of question in a public consultation, it always has a good chance of being approved. The countries that avoid populism consistently are those who do not have this type of consultation, do not have strong presidents, and have strong parties. They implicitly reject voter rationality on issues while keeping the assumption of rationality on selection of representatives.

Selective Bayesians

Ashworth, Bueno de Mesquita and Friedenberg write: “An important empirical literature evaluates whether voters are rational by examining how electoral outcomes respond to events outside the control of politicians, e.g., natural disasters or economic shocks. The argument is that rational voters should not base electoral decisions on such events, so evidence that these events affect electoral outcomes is evidence of voter irrationality.” That is, some scholars found that incumbents often fail to be reelected when some negative event beyond the politician controls happens. Those scholars conclude straigthforwardly - this is evidence of voter irrationality. If voters are rational, why are they on average more likely not to reelect a politician if a hurricane strikes?

ABF don’t think we can accept this evidence. They point out that a hurricane may be totally beyond the control of a politician, but a response to a hurricane is not, nor is preparedness for one. According to ABF, the hurricane provides new information on the quality of the politician. All this seems reasonable enough. But if voters are rational, shouldn’t they assume some baseline level of quality, and then update their priors upwards if the response is good and downwards if it is bad? Then average reelection rates should be the same.

The authors build a model for why this is not necessarily the case. “Suppose voters believe that the incumbent is, ex ante, more likely to be high quality than is a future electoral challenger. Then, if there is no hurricane, they reelect the incumbent. If, however, there is a hurricane, reelection depends on preparedness. With high preparedness, voters learn the incumbent is high quality and reelect her. With low preparedness, voters learn the incumbent is low quality and replace her. By giving the voters new information, the hurricane creates the possibility that the incumbent will lose—something that does not happen absent a hurricane.”

They argue, then, that an empiricist looking at data generated according to their model would find exactly the same pattern as described above - on average, incumbents lose more often when a hurricane happens.

As it is, we have two theories which explain the data. But ABF’s hypothesis has a problem. If a shock reveals more information about the incumbent which allows voter to tell with more precision whether the politician was high quality (re-elect) or low quality (don’t) then it should not matter whether it was a negative shock or a positive one, reelection rates should always drop, unless voters are selective Bayesians and only update on negative news. If not, the average mayor should be hoping that his city was not surprised by some really positive news. I find this implausible.

“Oh dear, this will be terrible for my poll numbers”

We also have some evidence against it. Brazil found large reserves of oil in the early 2000s. National law established that municipalities close to where oil was going to be explored receive substantial royalty money. As it turns out, cities with royalties had a higher reelection rate than those without.

The rational irrationality hypothesis would suggest, instead, that voters will often be affected by information unrelated to the quality of politicians frequently. Some will be lucky, some not. But - I would add - on average representatives and parties would be of reasonable quality.

Stuggling Spocks

Ali, Mihm and Siga investigate a puzzle: “why does zero-sum thinking prevail even in ‘positive-sum’ settings in which voters’ interests are largely aligned?” That is: why is it so common for voters to perceive a situation where the benefits can be very widely distributed as one in which someone’s gain must be compensated by another person’s loss? This is a particularly interesting defense of the rational voter theory because it does not assume that the collective decision is optimal; it instead attempts to explain suboptimal decisions.

They explain their mechanism through an example. You have three voters, Ann, Bella and Carol. They have to decide on a policy, using majority rule, p1 or p2. If p1 is chosen, each one gets a payoff of 0. If p2 is chosen, then two of them will get a payoff of 2, and one of them will get -3. Social surplus is 0 in the first case, and 1 in the second. Under total uncertainty, they would all choose p2, which has an expected value of 1/3, while p1 is zero. Under full certainty, the winners would choose p2, which will get approved. But if the voters know that only sometimes any of them will know whether they would be a winner or a loser under p2, then p1 would win, even though it is suboptimal.

They find that in cases similar to the above, it is rational for an uninformed voter to choose p1. Suppose Ann is making a decision. Her vote will only matter if one voter has chosen p1 and the other p2 . It is stipulated (I confess my reading doesn’t allow me to say for sure whether or not this stipulation is central to their results, I assume it isn’t so my critique won’t rely on that) that only a person who has learned they will be a winner will choose p2. This means the other person is either a certain loser, or uninformed. Since, also by stipulation, the probability of learning one’s condition is close to zero, Ann will assume there’s a 50% chance she will be a winner or a loser. But now her expected value of voting p1 is still 0, but the expected value of choosing p2 is 0.5*2 + 0.5*(-3) = -0.5

Their model generalizes this. Their main result is:

“Theorem 1. The inferior policy p[1] wins in a strict equilibrium with scarce information if the collective choice problem is adversely correlated. Otherwise the optimal policy p[2] wins in every equilibrium with scarce information.” (bold in original)

The problem I see is that to get to those results they keep two conditions from the example above which are debatable. The first problem is that the collective choice problem must be adversely correlated - as the authors say, in all other conditions, the optimal policy wins. Being adversely correlated means that the more people have already benefitted from the policy, the fewer people are likely to benefit from it. In the example above, the number of winners and loses is fixed. If two people already know they are winners, you are sure to lose. But that comes really close to a zero-sum situation, even if technically it is not really zero sum because the optimal policy still delivers overall greater benefits. So what this theorem says is that according to those assumptions, voters who face a quasi-zero-sum situation behave as if they are in a strictly zero-sum situation. I don’t know that this is particularly noteworthy. But more importanly, what reason do we have to think such adverse correlations are particularly common in real world situations where voters fail to approve positive sum choices? None of the examples they present suggest that, if the positive-sum arguments are correct, there would be adverse correlation.

A second problem with the model is that it depends on the voter assuming they are the pivotal voter. But we have seen above that a rational voter would never assume that, because the chances that such a thing would happen are vanishingly small.

And third, it is odd to assume that voters would not be able to acquire the relevant information to make a sound decision based on the merits because that is cognitively demanding, but also that the are smart enough to unconsciously solve the theorems from the paper. Which is it, calculating Spocks or distracted average Joe? If they would - most of them - solve the theorem unconsciously, how exactly?

Smart for the hard thing, clueless for the easier ones

There is a twist to their model, however. Unlike in the example above, voters might not know whether they are in an adversely correlated or an advantegeously correlated situation. If they receive information that suggests the situation is adversely correlated, and no information they are in the beneficiary group, they are more likely to reject the proposal, even if in reality the situation is advantegeously correlated.

But is this surprising? It is tantamount to saying that if voters act according to incorrect beliefs regarding the benefits of a policy, they will fail to choose the optimal policy. We have always known this. They argue that if your beliefs are incorrect, then it is rational to choose the suboptimal policy. That is one view of rationality, and it is well-established in the academic literature. However, that is not how most people view rationality. In regular usage, rationality implies making the best decision considering publicly available information. If a whole group of people makes the suboptimal decision because they are entangled in a poor information-gathering system, most people would not deem their decisions rational. Using an extreme example, most people don’t consider that those who drank poisoned Kool-aid in Jonestown were rational, but by the authors’ criteria they would have to be.

This is made clearer with their examples. Regarding the discussion on approving or not the Emergency Economic Stabilization Act during the financial crisis of 2008, the authors write:

While it was clear that the financial industry would profit from the proposed $700 billion bailout, there was considerable aggregate and distributional uncertainty about the broader general equilibrium effects of the deteriorating credit market conditions. Would a bailout be solely distributional, transferring wealth from taxpayers to CEOs, or would there be significant beneficial aggregate effects of bailing out financial institutions? Information on this score was scarce given the sheer complexity of predicting the effects of approving or rejecting the bill, and much of the information in the public sphere emphasized distributional considerations. In line with Propositions 7 and 8, our analysis suggests that the political debate was ripe for the kind of zero-sum thinking that initially doomed the EESA bill to fail. Our analysis also speaks to why, following the subsequent market crash, the EESA bill passed on its second vote; that crash produced clear information about the aggregate benefits of stabilizing the financial system, rendering the collective choice problem more advantageously correlated during the second vote.

In other words, decision-makers had reasons to believe the situation was zero-sum, so approving the policy was suboptimal. They then gathered elements pointing to it being positive-sum, which made then approve the policy. I believe the authors are correct in arguing that the uncertainty regarding the benefits was greater before than after the crash, and that it could even be considered rational not to approve the bill until it was clearer it would be positive-sum (even though maybe even after the crash that was not the best bill which could have been approved). I will note one incidental thing - they are talking about Congress, not about the population. Congress is the entity that was convinced by the supervening events. Were the bill subjected to popular consultation, the chances that it would be rejected seem very high to me - indeed, the Occupy Movement claiming to speak for the “99%” originated exactly because of a similar perception.

In sum, if we stick to the first, more stringent premises, their theory is conceptually interesting but of little practical application because of how unrealistic their critical assumptions are, and increasing the realism would significantly alter the results. If we expand the premises to allow for voters not to know the overall effects of the policy, it is realistic, but the results are trivial and they would not make voters rational in the meaning used in public discourse.

Ineffective Altruists

In “Voting as a Rational Choice”, Edlin, Gelman and Kaplan recognize that it would never make sense for a rational self-interested person to vote in large elections, for the standard reason - as the number of voters increases, the probability that a single vote would change the outcome goes to zero. To the extent that voting requires any effort, that effort is better spent on something with greater impact. They argue, however, that it can be perfectly rational to vote for the benefit of others.

Here’s the logic - suppose that if candidate A wins, all citizens will make $250 more. Considering how small the chances of a voter to change the outcome, that personal $250 would never justify him casting his vote. But if he cares about his fellow citizens, even if he cares less than for himself, it can still be rational to vote. Assume that they care 1/10 about the benefit for others as he cares for himself. This means that if 10 people get $10, he’s just as well off as if he had received the $10 himself. The authors show that because the probability of being decisive decreases with population size at the same rate as the benefit derived from seeing others better off increases, then it can be rational to cast a vote for the better candidate with a population of any size. Elegantly, the expected personal benefits are approximately the size of the individual benefit times the discount factor for the benefit not going to oneself (multiplied by a number which increases as the election is expected to be closer; without going into the math, their multiplier is 25). Hence, with an individual benefit of $10, a multiplier of 25, and a discount factor of 1/10, the benefit of voting to the voter would be $25, independently of population size.

The authors note, correctly, that many elections can involve much greater stakes than $10 per person. Personally, I would find numbers up to $10,000 per citizen plausible. They neglect, however, that a discount factor of 1/10 implies an unrealistic amount of altruism. The idea that if some ten random people happened to make $1000 more than expected the average person would be as happy as if they got $1000 themselves seems preposterous. Still, let’s be generous and stipulate 1/100. This would imply a benefit of $2500 for voting, which is not bad.

In my book, I accepted this reasoning and claimed that the rationality of voting was then an empirical issue. After reading Chris Freiman’s “Why It’s OK to Ignore Politics”, I now believe it was never an empirical issue to begin with. Here’s an explanation inspired by the book. We are assuming that, at least in expectation, the typical voter is more likely to effectively cast his vote for the truly more beneficial candidate than for the other one. Given that in a close election there are about as many voters for one candidate as for the other, the typical voter can’t be that much more likely to cast a vote for the best candidate, if the population is large enough. Suppose that each voter had a 60% chance of voting for the best candidate and a 40% chance of voting for the worse one. This seems like a pretty weak signal. Still, if there were only one million voters, the probability that the worse candidate ever won is merely 2%. As you increase the number of voters, it goes effectively to zero (by which I mean the software R doesn’t even compute it).

The amount of real information (as in signal, not noise) a typical voter will get regarding the candidate’s quality has to be extremely small if a race is close. In an election with 10 million voters, any real information that makes the average voter more than 0.01 percentage points more likely to cast the vote to the correct candidate would imply that the best candidate would already win in around 75% of the cases, and if the information would make the average voter more than 0.1 percentage points more likely to cast the correct vote, the best candidate would virtually always win, so the problem of close elections wouldn’t present itself.

This means that for the typical voter, the expected value would not be the $2500 suggested. The voter would enter a lottery where he would have 50.01% chance of making $2500 and 49.99% chance of losing $2500. The expected value of that lottery is 50 cents.

What cost should we assign to voting? Edlin et al. think it could be the case that voting costs nothing, because people find it enjoyable. That may be the case sometimes or even often, but then we have learned nothing about a rational altruistic voter. The interesting question to ask is not “Is it rational to vote if you like it?”. It is rational to do anything one likes. The question is “Is it rational to vote even if you don’t like it?”. We must examine the hypothesis that voting is work.

In that case, most people would bear the costs of the time it takes to vote. If median hourly wage is $28 and people on average take an hour in the whole process, that would imply a benefit-to-cost ratio of $0.5 to $28, or 1/56. Ineffective altruism indeed.

While I find the case against rational altruistic voters well-refuted, I still want to highlight that anyone unconvinced by my arguments and who wants to stick with Edlin et al.’s thesis should accept that the self-interested voter hypothesis cannot explain voter behavior, so whenever we are trying to understand electoral results we should not revert back to self-interest explanations because “Edlint et al. showed it can be rational to vote”.

Why does democracy even work, then?

We know why people vote, or at least we know why people think they vote, because in surveys they have told us. The problem for rational choice theory is that the answer is boring, and it is not clear that it makes people instrumentally rational […] In fact, models of expressive voting (Brennan and Lomasky, 1993; Schuessler, 2000, 2001) do produce predictions that are consistent with empirical evidence. The marginal considerations we saw above still operate with an expressive component. The expressive component could not be the whole story or voters should not be too upset if it turns out the ballot box where they cast their vote was compromised so their vote was not counted. But voters do get upset, which suggests expressing yourself is not the only factor, especially since many people want to keep their actual vote secret. Furthermore, if all one wanted was to express a preference then there would be no room for tactical voting and there is evidence that some vote strategically (Cox, 1997; Franklin, Niemi, and Whitten, 1994). Nevertheless the ‘D’ answer [N: the expressive voting answer], despite being simple, despite being empirically verified by stated preference evidence, consistent with aggregate data evidence, and, if not properly tested, corroborated by Barry’s and Knack’s evidence, does not find much favour among political scientists whether rational choice advocates or critics. Why? Because they want deeper reasons.

Keith Dowding, “Is It Rational to Vote? Five Types of Answer and a Suggestion.”

Before studying public opinion, many wonder why democracy does not work better. After one becomes familiar with the public’s systematic biases, however, one is struck by the opposite question: Why does democracy work as well as it does? How do the unpopular policies that sustain the prosperity of the West survive?

Bryan Caplan, “The Myth of the Rational Voter”

I could go on showing how attempts to save the rational self-interested (or altruistic) voter hypothesis fail, but it probably wouldn’t make much of a difference now. If a reader is unconvinced, another example won’t convince them. One thing which is likely working against them changing their mind is the fact that the most successful governments ever to exist are democracies. But the success of democracies does not require voters to be instrumentally rational. Within the right framework, voters need only apply some simple heuristics, much like they do in everything else.

There are three key characteristics of representatives which will make a parliament work better: virtue, competence, and legislator’s independent influence over society beyond their role in parliament. The first two are (or should be) pretty obvious. Virtue is the most straightforward - if the legislators care about achieving the goals of the society they represent, those goals are much more likely to be implemented. Nobody seems to dispute this, so I won’t belabor this point.

More competent legislators will be in a better position to evaluate the quality of proposed policies, as well as evaluate the reliability of whatever group is pushing for a policy. While I think this should be clear to everyone, there have been recently some arguments that the most, perhaps the only, important characteristic of a legislature is that it is representative of the population, and “representative” is understood in the statistical sense, meaning that any member of the population has an equal chance of being in the selected sample - which in this case would be congress. The proposed reform is called sortition or “lottocracy”.

One problem with the proposal is that policymaking is hard, even if goals are exactly aligned. Assume that every person in a country, and consequently every legislator, agreed that the role of government is to create the conditions to have the greatest possible value in the Social Progress Index, which attempts to measure a large number of progress indicators and produces a final number.

Should they adopt free trade? Many smart people disagree significantly about that. What material should be prioritized in schools? What should the crime prevention policy be? What about vaccination and health in general? How much money should be be allocated to each problem, and when should programs be killed and taxes lowered? Remember no values are in discussion here, the whole country wants to maximize its score in the Progress Index. Still, it is the case that smart, caring and dedicated people will often disagree on many of these issues. But those who are smarter, more caring and more dedicated will generally arrive at better answers.

If you find that implausible or elitist, I would point out that we apply the same logic to everything else. When a newspaper needs board members, not one will pick a random selection of its readers. No university picks a random selection of students, administrators and teachers to lead it. No football team picks their management from a random selection of fans (not to mention picking their players). When problems are hard, we use the best screening processes we have to put the most qualified in charge. Why would lawmaking be the one exception to this rule?

The misguided idea of sortition also provides a good contrasting point to explain why independent influence over society beyond their role as legislator is important. I will rephrase a point I made in this post.

The legitimacy of parliaments to make decisions for the entire society is not a god-given right. Anyone could perfectly well assemble a random selection of fellow citizens who would start approving “laws” which should be valid in the entire territory. Would anyone follow them? Doubtly.

Now imagine instead that sortition did get approved and random legislators “rule” the country. Parliament is discussing an issue and then it receives formal communication that the leadership of the military, civil service, police, states, cities, major companies, newspapers and the tv and movie industry all wanted a specific measure to be taken and that they would in fact act as if it had already been. Would they manage pull this off? Likely yes.

How good are voters - or electorates, more precisely - at picking virtuous, competent and independently influent representatives? I have already made the case in the book* that we have good evidence that representatives are likely more competent and virtuous than the average citizen, so I won’t repeat those arguments - I will take a different approach. One thing I did not cover is whether people are any good at selecting independently influential representatives. I’m unaware of statistical papers on this, so will have to rely on my own reasoning and observation. Still, I am pretty confident that this is where voters do best.

How in fact do people with very limited knowledge of politics, candidates, and positions decide who to vote for? The same way they choose a doctor, a lawyer or a civil engineer: relying on heuristics. If the candidate or party is already well-known, that is a good sign. The second-term Brazilian senator Romario de Souza Faria would almost certainly never have been elected had he not led the Brazilian soccer team to its fourth World Cup title in 1994. But while being famous helps, it is not always a sufficient condition. Being known as a person who knows politics is more helpful. It doesn’t hurt to look competent, so candidates always dress like very serious people. Impressive credentials are also good, all else equal, as well as being perceived as attentive to segments of the population the voter thinks are neglected. But the most important factor is what the people you trust think of the candidate or party.

People usually know someone who they perceive as more politically literate then themselves and will assign particular weight to that opinion. They will almost always be correct that the people they perceive as more knowledgeable than themselves is indeed more knowledgeable, even if they will often delude themselves into mistakenly think they are more knowledgeable than still others. That knowledgeable person recognized as such might be a family member (parents are particularly influential, so party affiliation is in large part hereditary), a university professor, a talk show radio host, a priest, a union leader, the people one follows online (increasingly so), neighbors, among others.

The job of a campaign is, hence, to inspire more trust than its rivals do. That is a very large operation. The first problem is becoming well-known. If one is not a movie or a rock star, it is extremely hard to be famous on a national, or even a local, base. The hardest thing, however, is to enter in the trust trickle-down network.

For earning the support on the higher echelons of the network, it helps a lot to be actually competent and have integrity. When people are empowering someone else, they will most often prefer that person to be honest instead of a crook - even if the person doing the endorsing is not the most honest themself. One knows what to expect from an honest person but not of a crook. (Another strategy is to appeal directly to some in the middle echelons of the trust network by saying things which sound good but that others are not willing to say, often because they sound good but are actually not. I will come back to this strategy later.)

As it turns out, this works reasonably well - which is to say, perhaps not quite as well as choosing a doctor, a lawyer or an engineer, but close. Doctors, lawyers and engineers will often disappoint us, yet we trust this recommendation-based system enough that we find those who ditch it in favor of making all these decisions on their own completely insane.

As for democracy selecting independently influential people, according to my rationale this is the very reason they are elected in the first place. If you think about it, one of the rarest complaints about parliaments is that the people there are not that influential in society.

Not quite a roof without foundations

Why do all these people vote - is it because it is (instrumentally) rational ? Surely not, we established it’s never rational to vote for accomplishing a specific goal, out of self-interest or otherwise. They vote to express something about themselves, to the people they admire, to the people they believe admire them, and even to themselves. They often find it pleasurable in itself, or they think fulfilling a perceived duty relieves them of guilty feelings. The process is in many ways similar to how people become a fan of some music. The network effects have a lot of noise and the process is influenced by outside factors. But the best tend to do well generally and the worst tend to do very badly. When we aggregate the results from this whole process, the representatives or their parties can be reasonably competent, honest, and, even more than reasonably, influential. Just like with music, though, your gain in prestige only happens if you do not come out and say “I voted for [X] because person [Y] said it was the best candidate”.

Some may object that they don’t find the level of competence and integrity in parliaments reasonable at all. My claim that it is reasonable rests on two points: 1) there is no other process which more reliably will achieve all three goals; and 2) the success democracies have achieved is unprecedented, as stated before.

Putting it all together - how does voting work well, how does it go astray, and what should we do about it?

The two sides of voting

I’ve talked about why people can be irrational while voting, and I have talked about why they can be rational. But what are the conditions for it to work well or not? Since the argument is that voting can produce reasonable outcomes when people are selecting professionals or “businesses” (parties, in the political case), then the first thing to dismiss is the idea of direct democracy. One may ask: but why can’t voters go through the same process when picking policies? Just favor the policy that those they trust more favor. There is some of that, but - because people vote to look good to themselves and to others - once the vote is on a specific issue, you may look like less of a fool giving your sincere reasons to vote for the inferior policy (particularly if your counterpart is no specialist either) than drawing a blank when someone asks you to explain your decision to vote for the superior one, or just admitting that person X said so.

But even when voting for politicians, sometimes appealing to rational irrationality can beat strategies based on network of most informed to less informed. Unsurprisingly to any of my readers, I will argue presidential politics has much more of this problem. For one thing, in a winner-take-all election, if a candidate who appeals to voters’ rational irrationality is elected, the harm can be much greater. In a parliamentary government, rational irrationality measures will depend on the concurrence of many more elected representatives. That is not easy to accomplish. An anti-vaccine legislator might have no problem with free trade; a proponent of doing away with the separation of Church and State might not be onboard with doubling the minimum wage; a degrowther would likely not agree with a climate change denialist. In a presidential race, there could be lots of disagreement during the campaign, but once elected, the irrationality of the president would have a special position. This argument has some similiarities with the “miracle of aggregation” mentioned above but also fundamental differences. In a presidential election all the biases affect votes in an independent form, and any candidate representating a combination of biases might be elected - hence, no miracle. In a parliament, legislators must negotiate and sustain a coalition. The process of negotiation contributes to weed out irrational aspects.

Further, the very fact that no individual legislator is that powerful by him or herself makes any promises to implement rationally irrational policies less credible. Because of that, a regular voter will often think that when casting a vote for legislator, the best thing is picking someone particularly competent, particularly committed with some interest group or cause the voter cares about, or who belongs to a party they trust. Which all conspire in favor of the trust network instead of the populist approach.

The greater importance of parties in parliamentary democracies - a widely recognized fact - is another contributing factor for fewer incentives for populism. A single person can rely on personal charisma to connect with large shares of the electorate. A party has to maintain a brand, and brands do not have personal charisma. The preferred strategy will involve cultivating long term ties with publications, think tanks, civil service, etc. All of this will make the candidate incorporate wisdom from several corners of life when making decisions.

Appeals to irrationality work better when those in the higher echelons of the network start failing to produce sensible recommendations in a way that anyone can see (even our rationally irrational electorate). A populist is often defined as someone who campaigns with a platform against out-of-touch elites. But sometimes elites are out of touch! When they are, the case that they are disconnected will be much easier to make.

Presidential politics will more often produce such out-of-touch elites. Because the risks of concentrating power at the hands of the president cannot be completely ignored, presidential countries come up with all sorts of legal (or practical) protections against political interference for the civil service, military, judiciary, academia. These corporations may acquire a self-sufficiency quality which increasingly will make them more likely to adopt stances which are either clearly self-serving, false, or both. I would note that, because elites can be out-of-touch, even some not usually inclined to populism will at times want to rely on “the wisdom of the crowds” for a course correction. It is a poor remedy. Irrationality is, well, irrational. Mobs marshaled for a cause can often just as easily be mobilized for a different one, with unwanted results.

Of course parliamentary democracies are not immune to all appeals to voter irrationality, and neither is being presidential the only contributing factor. Malapportionment, for example, will tend to distort how much an outcome in parliament reflects the negotiated opinion of the most influential sectors of society. This will more likely produce out-of-touch elites as well. Likewise, some technologies which make information spread more horizontal can favor those with a populist message.

Wrapping up

The “democracy works, so voters must be rational” meme seems to be hardwired in people’s brains - and when it isn’t, the hardwired meme is “voters are irrational, so democracy can’t work”. I will give it one more shot bringing back the examples from the beginning of the text. Are people not rational enough to define their own medical treatment but rational enough to choose a doctor? A lawyer, an engineer? If you answered yes, you probably don’t think that telling someone that he should trust an expert in some issues instead of deciding alone is diminishing to that person in the slightest. Why then can’t we enthusiastically embrace similar thinking applied to voting? For the benefit of democracy, I hope we do.