Stuck in the Middle With You

Are voters too evil and stupid for democracy?

it took me a minute to remember what the LEGALIZE one was about. not an issue you hear much talk about these days

Since November or so, we’ve all had the “pleasure” of debating brand new topics like identity politics, which cultural issues to move to the center versus which economic issues to move to the left on, and tedious post-2015 chattering class tedium. They even had a whole conference about this, and it also somehow tied into the even stupider and more tedious Abundance discourse, which is basically just 2020s Capital in the Twenty First Century discourse. Except, of course, “C21” was an extremely dense and technical 1,000 page economic history book filled with charts and Abundance is a 200 page airport bestseller about how drones can deliver space medicine to your house - so nobody has any excuse to not actually read it before firing off hot takes¹.

So like, what is actually happening in the millionth round of moderates versus the left?

Clowns to the left of me… jokers to the right…

Over the last three months or so, I’ve been entertained by the recurring Matt Yglesias versus Jake Grumbach beef on Twitter. The general jist of it started in February and the arguments made by each side are as follows: the “Yglesias” side basically believes that “moderate” politicians win more elections by persuading people who are sort of indifferent between the two candidates, or favorable to one but not as much, to flip sides for them. Grumbach generally thinks that the evidence is low quality and that the definitions used are frequently squirrelly and dishonest. Unsurprisingly, they’re both kinda right and kinda wrong.

The evidence on the “popularist” (which is what the Matt Y faction is called) side is, I think reasonably strong: if you look at how left versus right candidates are relative to the electorate, ones closer to the middle tend to perform better, and ones further tend to perform worse. Also (and this one is filled with endogeneity so you can’t count it that much), members of “more moderate” groups within parties tend to perform better than members of more extreme ones. At the issue level, taking less extreme stances tends to result in stronger performances. There’s also somewhat of a correlation between being more moderate and performing better, a reasonable case to be made that politicians taking positions closer to the ones the voters hold helps them win, and a relationship between which policies candidates support and how well they do. And, additionally, candidates tend to move to the center between the primaries (where they have to pander to their own side) and the general election, where they have to appeal to everyone. So, case closed?

Kinda. Let’s dig slightly deeper. This whole thing generally comes from what’s called the Median Voter Theorem. The general idea behind the Median Voter theorem is that people’s opinions on a given issue are distributed normally: basically, along a single line and with a single peak that has the most concentration of people around it, and where the amount of people decreases further and further from that peak. If voters choose based on how close each candidate is to them, then the candidate who’s closest to the “median voter” (the one that’s at the top of the peak, that is, the one with the most popular opinions) should vote, because that should represent everyone to one side of the median. The interesting thing is that the model comes from economics of all places: in particular, from Hotelling’s model of spatial competition (here’s a refresher, from moi). Hotelling’s idea was that, to explain differences in quality such as location, you can imagine a beach where customers are spread out evenly, and where walking from place to place is undesirable for them. Now imagine two competing ice cream vendors: where will they go? Well, they’ll go to the place where the most people have to walk the least- so, if they’re evenly spread out, it’ll be the middle. If there’s an uneven distribution of people, they’ll go both to the same place, which is where the most customers are.

This gives you, generally, three general benchmarks for voting: the first is that (in two-candidate systems), candidates are going to tend to be in the middle of the electorate; that the candidate closest to the median voter should win, and that changes in where the median is should change what positions are “winning”. The core assumption here, besides that people have “rational” preferences, is that there’s a single peak (or, at least one fewer peak that there are parties), which means that the supply of available voters for either party overlaps a lot on each issue. One particularly thorny example for the Median Voter Theorem is… the United States Senate. In the US, each state gets two Senators; it’s plausible, then, that closely divided states have one and one (currently, only three states are that way: Maine, Wisconsin and Pennsylvania). Under “medianism”, the two senators should be relatively politically close to each other, and relatively close to the average voter of their state. But in fact, they’re not - in most cases, they tend to be closer to a random other Senator than to the other Senator from the same state. In the current US Senate, it is true for Maine that they have a centrist Democrat and a centrist Republican, Maine is a liberal state that votes around 6 points to the left of the rest of the United States on average. Meanwhile, Pennsylvania has a “normal” right wing Republican and… something else entirely as Democrat. And Wisconsin, the closest state of the 2024 election, is represented by far-right lunatic Ron Johnson, and progressive Democrat Tammy Baldwin - not exactly Joe and Jane centrist. The reason for this is ideology: unsurprisingly, politicians… care about politics! The “different politicians can care about different things” idea also has another surprising backer, the fact that when more women are elected, government outcomes change in pro-women directions, even though women could already vote in elections.

Twin peaks

So this opens up “objections” to the “popularist” views. There are a couple of objections. To get the tedious one out of the way, there’s a lot of empirical and measurement issues that are kind of maddening to discuss (they really, really are - most of the debate here is just threads of hundreds of tweets about minute measurement differences). In general, all the big questions relate to what’s called the Stable Unit Treatment Value Assumption or SUTVA - basically, that there’s only one version of a “treatment” (read: moderation on an issue) and that it can only be applied one time if there’s multiple ones. But like, that’s obviously not true, because “do you support policy X” has different versions even of a yes and no answer, which colors how people take it - “I support first trimester abortion because I think the White Race should be exterminated” is obviously not identical to “I support first trimester abortion because the Pope says it’s okay to save the mother’s life” or whatever.

The first (actual) issue to consider, which is what’s usually brought up, is turnout (i.e. how many of your voters show up): it’s possible that voters might “throw away” their vote if they think the median isn’t close to their preferences (for example, by voting for the Green Party because the mainstream center left is not left wing enough on the environment). In very bland empirical terms, running a more moderate candidate tends to depress turnout and running a more extreme one tends to increase it. This is true, but it’s also true that more extreme candidates also increase the turnout of the other party, so the effect is kind of a wash on net and depends a lot on the specifics of each race - so it’s somewhat conceivable that, on net, running overly moderate candidates lowers turnout of your party’s voters and could cost you an election. But also I think that it’s clearly true that “people don’t show up to vote for you if they don’t like you” has to do with how much people of both sides like you.

Another important issue, is that in the current moment we’re seeing intense political polarization violating the core “single peak” idea, because then there’s just not a lot of voters in the middle compared to on each side, drastically reducing margins of victory, gains from centrism, and opportunities for expansion. This is also important given that the US political parties have moved asymmetrically on the issues - the Democrats are pretty mainstream for a developed world political party on the left-right axis, and the Republicans are closer to like, Putin and Erdogan than even to someone like Marine Le Pen. Another manifestation of this phenomenon is a weird trend that the US didn’t have (given structural factors that make it a stable two-party system) is Pasokification: the idea that for a number of reasons, the European center left (think social democratic, moderately progressive parties) absolutely tanked politically relative to the "progressive left" after the Great Recession or so. This trend is named after PASOK, Greece’s version of the Social Democratic Party, which completely collapsed politically and was overtaken by Syriza in 2012. While some of this trend reversed in the 2020s (PASOK is a large party again and has overtaken Syriza in recent polling), it’s still plausible to just lose voters “to the left”, such as the aggressively centrist Sir Keir Starmer of the United Kingdom, who is also aggressively unpopular and appears to be getting “pasokified” by every single other political party in the isle of Great Britain (it’s also very much not what he was supposed to be doing in the “Blue Labour” space).

Secondly, and fairly obviously, the gains from moderating are generally small ceteris paribus, so political polarization makes the effects of the “moderation premium” much smaller (to the point of borderline statistical insignificance). The reason is, I think, fairly straightforward: if there’s fewer voters than before up for grabs, then there’s few voters that are ever going to change their mind on who to vote for regardless of what you say about gun control on your ad campaign. Likewise, this means that there’s fewer competitive elections than before, so in those elections the voters you can gain from moderating could be decisive - but they’re also a massive outlier compared to everyone else, kind of by definition. A similar related issue is that it’s not actually how moderate you are, but rather how moderate you’re perceived to be that matters - and that one’s kind of a black box, especially given the (pretty wide) range of information that counts as “a campaign”.

And one final issue is that “what does it mean to be a moderate” isn’t even a very clear-cut question. There’s, to my understanding, three positions here: the first is that it’s being kind of centrist on everything, the second is that it’s being close to what your average voter thinks, and the third is that it’s being in the middle of the two candidates running. The three are usually quite close, but they’re not necessarily the same. The first take of “just be kind of centrist” runs into the problem that most actual moderate voters have preferences that are kind of incoherent with each other - they’re somewhat economically populist, somewhat socially conservative, and fairly protectionist and nativist. In addition, the second definition is obviously the one that most people tend to mean by “moderate on X”, but it’s not exactly the same as number three, which is the one most studies use because it’s much easier to measure - but, mathematically, you can’t actually determine how moderate to be, or on what, by just averaging out your side’s positions with the other side because you’re trying to determine what your position is going to be. When I said the debate can get quite squirrelly and dishonest, it’s usually because people tend to slip in arguments about the evidence for or against different definitions in each point they’re making, and since it’s the same word it’s genuinely tricky to tell them apart without seeing whether it uses DIME or CW-Nominate or WAR or whatever in each specific case.

Leviathan versus Behemoth

My problem with “popularism” is that I think it’s generally a good and accurate theory of How To Be A Politician, but not really a good theory of “politics”. I think that, at its core, it has a step missing, not on how to react to voter preferences, but on how those preferences are shaped. Beyond whether moderates “do better”, even the critics of popularism tend to demarcate economics and culture in terms of voting: in terms of concrete policy questions you need to go with the expert consensus, and you need to pair an evidence-based “populist” vision on economics with centrist cultural positions on salient issues².

The obvious things to say here are, well, obvious. The first is that “the experts” don’t always agree with each other. The second is that what “the experts” recommend on “populist” questions of economics, and on issues such as trade, is exactly the opposite of what the voters (or at least the cross-pressured moderate ones) want. And third, the line between culture and policy isn’t even clear: some of the most hot-button issues in the culture wars, like trade (again), immigration, climate change, and transgender medicine are an explicit clash between “populism” and expertise.

But more generally, I think that the definition of “culture” being deployed is… not very useful. I think that a lot of the conception here is that voters are reaching their positions via some careful consideration of the arguments and not downstream of previously held “core” values. I’ve been on this for a while (I’m not going to do the Stronger joke again), but the values at play are three, they’re interrelated, and they’re pretty hard to get around.

The first of the key “value” distinctions between the American right and the American left is between moral universalism and moral particularism: quite simply, the left cares about people that the right does not even bother being reminded exist. This also shapes ideological positions on various issues, such as redistribution, trade, foreign aid, and the environment - and in the American context, areas with higher scores of moral universalism are more likely to elect liberal Democrats. The link between moral universalism and left wing politics is well established at this point.

The second source of values-based tension involves social trust as part of social capital. “Social capital”, which is kind of synonymous with social trust has four components: civic participation, political participation, trust in others, and trust in institutions. So if you imagine “high social capital” just imagine the Gilmore Girls. The main throughline between these components is how much people trust each other, aka social trust, which is related to moral universalism because both are determined by social structure and cultural factors - cultures with tight kinship norms score very low on both universalism and social trust. Social trust is highly determinant of many beliefs, such as those around gender norms, immigration, and trade.

And the third distinction is between zero and non zero sum beliefs - the idea of whether it’s possible for people to gain without someone else necessarily losing. Very obviously, zero-sum thinking is associated with lower support for, say, trade or immigration. But it’s also associated with higher support for government redistribution, and for affirmative action. In fact, individuals with zero-sum mentality are also less likely to emphasize the importance of hard work for personal success, which remains still a fairly “left wing” position.

In general, you can see that the particularist, low trust, zero sum group is anti trade and anti immigration, but also pro redistribution and somewhat socially liberal (or, at least, their views dont’ map that well on the liberal/conservative axis). Donald Trump cleaning up with the “Infowars” demographic is not especially surprising: a big chunk of his voters have very little trust in societal institutions, and instead rely on tightly knit social circles and their personal allegiance to Trump; what explains the “white working class” (and later voting class voters in general) making a big switch to the right is that they were generally centrist except on issues of trade and on racial attitudes (particularly, you guessed it, immigration). In general, while it’s not particularly surprising that the Democrats are hemorrhaging anti trade, anti immigration, nominally pro redistribution, and not particularly socially doctrinaire voters, the combination of positions to attract this group back is pretty surprising: protectionism, nativism, and economic populism, without a lot to say on what positions are important regarding, for example, gender or race.

And here’s the, I think, important part: while this values do indeed have a pretty substantial cultural and social component (for example, proximity to diverse groups tends to help social trust), probably the most important indicators for all three “value sets” are economic - particularly, education, employment, and other “positive” experiences like lack of discrimination. For example, people who grew up in times when economic growth was low tend to exhibit lower social trust, as do people who are unemployed or suffer from discrimination and unemployment. And something that I think is pretty striking is that trading “uneducated” voters for educated ones is pretty universal for center-left parties, which kind of puts into question that every single party in the developed world would have become excessively “woke” in 1995. In fact, the US is pretty unique for how early this started (back in… you guessed it… 1965), because the decline of class-based voting is pretty longstanding.

This happened, in large part, due to the disappearance of “good” jobs that do not require college degrees (a phenomenon called labor market polarization), driven by technological changes that boosted professional services over manual labor, technical substitution, and also (most infamously) growing international competition from trade. This affected men particularly disproportionately, while also benefitting women, which resulted in a pretty striking gap in values and therefore in a striking gap in political affiliation. Something that I think is pretty telling is that after NAFTA was implemented in the United States, working class voters abandoned the Democratic party after they lost their jobs, but only if they already scored highly on social conservatism - so they continued to vote “with their wallets” until the economic differences between the two parties became a less salient factor than cultural ones. This can also help explain why Republicans tend to benefit from a stronger economy - because voter’s class position is less, not more, relevant. And an interesting avenue to consider is that support for redistribution from the rich to the poor is directly tied to the fact that a desire for fairness is both associated with the “conspiracist” mindset and with the desire to take from slightly richer people to give to myself.

The part where this gets kind of worrying and not just “be a bit more populist” is that the low trust, zero sum, morally particularist group is also, well, not especially supportive of democracy. This follows a pretty strong pattern where conspiracy brain individuals tend to be worse off, and people who suffer economically are less supportive of democratic institutions, and also due to the fact that the values themselves imply a rejection of “all men are brothers” in favor of “fuck you, got mine”. I’ve already made this case in the post about the Pope, but one of the fundamental dividing lines is between gemeinschaft and gessellchaft, that is, between the rule of social bonds or the rule of impersonal institutions - or, in Carl Schmitt’s terms, the transition from Leviathan to Behemoth (the two are also somewhat important to the Dark Enlightenment morons you keep hearing about) One such example is, well, Nazi Germany: the organized, rational, bureaucratic state of the Weimar Republic was replaced with what Spanish speakers would call dedazo, the rule of arbitrary and capricious decision-making (“the finger”). One such example was economic policy, where discretion replacing rules meant that jockeying for political favor overtook any sort of practical considerations - an approach that could, at times, be reasonably popular provided it actually reached voters. Importantly, while participation in external social organizations could counteract the forces (the Great Depression) that brought Hitler to power, the Nazi regime very actively took care of dismantling them and replacing “good civil society” (churches, unions, etc) with “bad civil society” (the Hitler Youth, NSDAP party activities, and whatnot).

Time! And! Place!

If you’ve been paying attention, it’s really hard to overcome the rule of social norms for people whose lives are lead by social norms - you have to offer them the right carrots at the right time. This led to “deliverism”, the idea that the best way to win back voters is free stuff! Throw them a party! You know, like the Nazis did. This was the “political economy” approach of the Biden administration and I think it’s self-explanatory that it did not work and that Biden was as popular as diarrhea by the end of his term. In general it’s not been considered an effective way to run a government, and while people have used it to gas up Abundance, you should keep in mind that someone making use of that is Sir Keir Starmer, which is kind of the equivalent of being on the Forbes 30 Under 30 list for contemporary political philosophies.

But an actual, workable example is that, in US states where the Earned Income Tax Credit (a welfare program) is expanded, the governor’s party sees an increase in 2% of the vote - seems like not a lot, but if the Democrats did that, we’d be in President Harris’s term instead. Except a lot of it is driven by voters being surprised that conservative governors would expand it, and thus switching their vote, so it points to a big mediating factor being information and salience, where getting a check with the President’s face changes your opinion more than getting a deduction on your property tax payment in 5 years. Which is how you enter “vibecession” discourse where people just suddenly become completely irrational and respond to made-up economic conditions because of the “information environment”.

One example in the US specifically was the Biden stimulus package and the aggressive post-COVID response, which was supposed to avoid the backlash to the Obama administration by running the economy hot, especially since wages would beat prices. Except, it turned out that people fucking hated inflation and that prices and the ensuing high interest rates explained the entire vibecession, not because wages didn’t catch up, but rather because people somehow separate macroeconomic outputs between prices, which are the government’s fault, and income, which is my “fault”. And people’s dislike of inflation in general is not very factually based either, even if wages do rise to match it - in fact, what matters is their belief that wages didn’t catch up, not whether or not their wages actually caught up with prices.

So are people just too stupid to vote for good macroeconomic policy? Well, individuals do have a variety of behavioral and cognitive biases that prevent the proper use of information for economic decisions (no, I still don’t like Nudge). In The Myth of the Rational Voter, Bryan Caplan argues that, in fact, voters are just completely incapable of making informed decisions because of these biases - particularly, when it comes to economics. People care too much about their country and not enough about other countries, they don’t listen to experts, don’t consider facts, and instead willfully believe incorrect things and vote according to feelings and ideology. Other social scientists have their criticism of his book (I don’t know why I included a source because people always have criticism of books), but I think it also dovetails with some other, interesting questions like “why do poor, uneducated people vote for the right” - people’s “interests” can be, in fact, completely incoherent.

It is pretty ironic that a libertarian like Caplan wouldn’t recognize that the argument “people are not rational the way that experts describe them so this model can’t work”³ is also exactly what people say to argue against the free market. The reason why this criticism is wrong for markets is that that’s not how economists think markets work. The main belief on the role of markets comes from Friedrich Hayek, who says that their role is to aggregate information that is dispersed in an easily legible form. He gave this information the very snazzy name of knowledge of the particular circumstances of time and place. There’s a lot of specific, concrete information that people have that is really hard to summarize and use from the top down, but that’s fairly accessible from the bottom up because it shows up as prices to other people. So you don’t actually need to know how to make each part of a pencil to know whether to substitute one supplier for another - you can just look at their prices. I’ve made this comparison before, but it’s kind of like the theory of evolution: animals that adapt to, say, camouflage in a forest aren’t actually choosing how to adapt; instead, they’re randomly rewarded or punished by external changes, but you could say, if you wanted to make testable predictions, that they would “choose” to adapt. In fact, with fairly minimal stuff like “companies try to stop losing money when they’re losing money”, you can get to a model where the average company behaves as if it was following all these sophisticated rules, but in reality is just throwing shit at the wall.

But back to popularism, deliverism, and whatever aristocratic nonsense Caplan suggested, the very obvious question is what are the assumptions about the preferences themselves - particularly, that those preferences are “rational” or not - in particular, because there’s been significant discussion about, what if they aren’t. Let’s take an example: Woodrow Wilson, who was President of the United States in the early 20th century, was elected in 1912 after being Governor of New Jersey, and he carried his home (no I don’t care that he grew up in Virginia) state. In the 1916 election, Wilson was reelected comfortably, but somehow lost New Jersey despite being popular. Why? Well, because of transgender athletics shark attacks! In particular, shark attacks in the New Jersey shore, which depressed the state’s coastal and tourism economy, and thus caused voters to turn away from Wilson. While this is complete nonsense (especially considerign criticisms of the paper), it’s about as nonsensical as “blaming the President for gas prices going up because Russia invaded Ukraine”. In addition, voters are generally loyal to whatever party was successfully in power when they were young. So it’s all kind of random, really.

Two political scientists, called Christopher Achen and Larry Bartels, have a book titled Democracy For Realists that argues that, instead of voting “rationally”, people vote mostly on partisan loyalty, social affiliation, and “blind retrospection”, basically an up or down vote on whether they liked what was going on over the previous four years. In fact, their conclusion is criticized as too optimistic, because unlike Caplan they’re pretty hopeful that reforms to democracy can make social identity be better represented in politics and thus make group conflicts more tractable. And in fact the book that best captures American’s preference for political outcomes is called Stealth Democracy, and basically the argue the authors make is that voters don’t like when politics “happens”, punish politicians who make politics happen to them, and are generally averse to conflict. In fact, the authors argue that a lower-visibility, lower-participation, lower-stakes system of government that is less efficient would also be more popular and have higher direct support. There’s some evidence backing these views, but I think they’re all kind of similar with the “people stoopid” thesis.

So like, it’s pretty grim for the competence of voters. Except that “people are stupid and vote kind of randomly against people who either have bad things happen or kind of bother them” is not a new position. In fact, back in the 1920s, it was the type of thing that Walter Lippmann and John Dewey (who are the kind of people only Americans know, like Martha Stewart or Tom Brady) went to blows over.⁴ Lippmann argued that, wouldn’t you guess, the public is too emotional and irrational to make choices because they rely on stereotypes, media distortions, and simplified narratives. Instead, control over the state by civil society should be minimized and experts should take the wheel. However, Dewey’s counterpoint is pretty interesting: besides the fact that Lippmann’s case was, well, frankly immoral, and that there are many practical considerations (in particular, that the public could be educated and that experts can be fallible - case in point the COVID pandemic), he makes two far superior ones. First, that people who won’t voluntarily listen to experts also won’t necessarily obey them - take the case of COVID lockdowns, which were rendered ineffective by low compliance. But Dewey’s secondary point is actually far more interesting (according to my friend, anyways): he makes an epistemic case for democracy, where if you take the liberal position that individuals know what’s best for themselves, then the only way to know whether individuals think things are going in the right direction or not is by letting them have a say even if they’re basing their “thumb” on a completely asinine logic - because, if they truly are responding to the real world, then they’ll lead government in the generally correct direction. Which is why concerns over the “information environment” and how social media distorts public opinion are so important - because people would be responding to things that don’t exist.

Conclusion

So no, I don’t think people are too stupid for democracy, even if largely uneducated voters choose bad candidates and parties based on stupid reasons that largely reflect values that I find incorrect and morally reprehensible to hold. If you don’t give people a say, then you won’t know whether you’re going in the right or wrong direction. However, I think that reading the tea leaves on what, exactly, went wrong is a bit stupid - Americans did just press the “make things different” button. What you need to offer them is an alternative, vibrant vision of how to solve problems - not just an electoral strategy, but a way to build a coherent liberal state that is successful in its material pursuits. It’s also worth noting that Melvin Rogers, the biggest John Dewey scholar out there, thinks winning the culture war is more important than winning the political war, because culture is generally upstream, not downstream, of voter preferences. So I dunno.

A lot of this advice from election gurus and pundits is fairly tautological. But relying not just on politicians but also on private individuals and civil society groups only taking politically advantageous positions for given parties is… pretty bad by the standards of John Dewey - democracy is all about the exchange of information and if you encourage people to conceal what information they have about their preferences, you can’t have choices that reflect the real, existing public, and instead get one that reflects the consensus of politicians back at them. Which is fine… until it isn’t. Because when people think the system isn’t delivering for them, they lash out and want to burn everything to the ground.

1. I haven’t read it, so no takes here (I think they have a good policy case but a weak politics one, mainly because they’re trying to pitch a positive-sum agenda to an electorate of zero-sum morons), but I think Matt Bruenig has a good overview of the Takes on Abundance, Mike Konczal has a good overview of the ideas, Bruenig also has a good piece about its relation to redistribution, and the Niskanen Center does a good job at putting together a US partisan politics case. Also heard good things about this post by Brian Beutler about it. And Eric Levitz at Vox has written a more detailed review and his defense of the political case for it, and Ned Resnikoff wrote about the political theory behind it.

2. This is kind of beyond the point but there isn’t a good guide to what topics to moderate on, because the relative importance of topics to voter is not very easy to gauge from polling, and is a frequent source of disagreement between the popularists themselves.

3. Actually he does say that democracy has no feedback against bad outcomes and bad decisions, but like, politicians with bad economies frequently lose free and fair elections eventually. Inflation in particular is horrendously unpopular.

4. Completely coincidentally I went to my friend’s thesis presentation a few weeks ago and this was the topic of his thesis. So idk go yell at him if you disagree here.