Voters are mostly rational: a response to Keith Woods

Voters are only random within a small range of policies

Recently Keith Woods wrote a response to a critique of him I wrote. In his response, the thesis is the following:

Democracy does not produce responsive governments, and that all of the intellectual defences of Democracy fail when put up against what we know about voter behaviour.

I will argue that democracy actually does produce governments that are significantly more responsive than imaginable alternatives, such as random government selection or elite battle royale.

In general, the book (7 page article version here) Woods is supporting is like this: once upon a time there were massive differences in human intelligence due to random variance of genes. Eugenicists promised to reduce variance by making everyone’s genes the same. The mean IQ becomes 130 with a standard deviation of 2 points.

Larry Bartels and friend write a book called “Eugenics for Realists” where they argue for the “random IQ score.” They note that “In one study, only about 70 percent of people had an IQ of about 130. The rest randomly had IQs in between 124 and 136.” “The ideal of equal intelligence behavior is further undermined by accumulating evidence that people can be powerfully altered by lead exposure in the womb. Michael Hagen, and Kathleen Hall Jamieson tracked prospective humans’ responses to changes in the lead exposure and found that lead exposure can decrease IQ by as much as 0.7 points!”

Keith Woods reads this and concludes that “eugenics does not produce egalitarian IQ distributions, and that all intellectual defences [sic] of eugenics fail when put against what we know about the IQ distribution.”

This parable may be cryptic to some of you. But you will understand by the end of the article.

How rational are voters?

To understand, we need to define voter rationality. Bartels claims that voters are completely irrational.

A voter is rational if he can, under certainty, 1) rank candidates 2) these rankings are consistent and  3) his rankings are continuous. In practice, this means a voter is rational to the extent that, under informatic certainty, he is not an RNG. An irrational voter deficient in 1 cannot even vote, at least not non-randomly. If a voter is deficient in 2, he will vote randomly over time, displaying inconsistent preferences. If a voter is inconsistent in 3, there would be no parties or party preference, and his preferences would have no pattern, effectively being random between candidates. In other words, the irrational voter is a random number generator. A voter might be ignorant, petty, small minded, and naïve, and still be rational.  

Informatic uncertainty is not the same thing as essential irrationality. Some uncertainty is avoidable. Some is rational, because there is a cost to attaining information. This means that, in practice, if uncertainty is below a certain risk threshold, voters are essentially “rational” even though there is some randomness in how they vote.

But for now we can’t untangle rational irrationality and essential irrationality. Therefore, to be maximally charitable to Bartels and Keith, we will just count all randomness as essential irrationality. This will give us an upper bound for the amount of essential irrationality there really is in the voter-pool.

The key insight here is that some people might be irrational, and others might not be. Some might be more irrational than others. If there is a scale of ideology, someone who one day is on the far right and the next day is on the far left (horseshoe theory aside) is more irrational than someone who sometimes feels like a moderate liberal and sometimes feels like a milquetoast conservative.

Essentially, we can quantify rationality as political variance explained. If political beliefs are in between -3 and 3 in this society, and the average person or polity randomly votes somewhere in between -3 and 3, then there is total irrationality. If they vote on a smaller range, then we can explain some variance if we can predict what the mean of each person’s range is with, say, genetics. Total rationality means we can perfectly predict how someone votes.

Views are stable through lifetime

Political views and party preference are mostly stable throughout an average voter’s lifetime. The heritability of conservativeness is relatively high, in the range of 0.50 to 0.70. That doesn’t leave a lot of room for randomness especially considering that some fraction of the non-genetic variance is going to be accounted for by economic incentives. Thus voters are mostly maximizing  their genetic utility when they vote.

Here is what that looks like. What this means is an average person (genetics = 0) will not vote for someone with position = 2 (an extremist to the left or to the right). Thus they are only “random” within certain bounds. This makes sense because you can predict that if Trump said the n-word and started quoting Hitler approvingly that this would completely destroy his election odds. 

In other words, non-rational voter theory says a voter, come election day, is like this:

X ~ N(0, 1). 

They vote for any position on [-3, 3] randomly. Rational voter theory says they are actually maximizing something and therefore they are not random at all. They have no variance once you know what they maximize: 

X ~ N(g, 0). 

In between is something like this. They vote around their genetic and economic preference with a little bit of randomness: 

X ~ N(g, 0.2)

Where 

g ~ N(0, 1).

What this means is that there might be some randomness within bounds. Political scientists might find that two parties that take position .2 and position -.2 respectively randomly changed offices depending on who gets lucky. However, what they are missing is that anyone who takes position 4 will not win, because voters are not that random. In fact, people who take position 1, like libertarians, or position -1, like democratic socialists, also basically just don’t win. There has to be a lot of rationality if we can predict that, no, there will not be a sweeping Libertarian Party victory in 2024 by chance. Rather it will, as usual, be a coinflip between the guy who takes position -.2 and the guy that takes position .2.

As high heritability predicts, views are also stable through the lifespan.

Half of all people never change their views meaning they vote for the same party their whole life. This is not randomness. Also, as the gene pool has accumulated mutational load, shifting to the “left”, more than half of the view-changers report moving to conservatives. This is rational given the background. Thus only about 21% of people are pseudo-irrational, although their flip can probably be explained by extrinsic factors. And again, they are likely very average people genetically and therefore are only random within certain bounds. These people will not randomly join the Communist Party of America, which is what irrationality theorists predict.

Polity level preferences are also very consistent. In a two party system, a voter irrationalist predicts the following polity flip distribution:

In between 15 and 36 states will flip. Mostly, 21 to 30 will flip in any given election.

But how many flipped from 2012 to 2016? 6 states. What about from 2016 to 2020? 5 states. And  this is again for candidates that don’t actually vary as much as they could. Thus it seems that the makeup of most states means they will basically always vote on one side of 0 while in a few states the results will be in between 0.3 and -0.3, roughly speaking. But Georgia won’t suddenly vote for the American Nazi Party. There isn’t that much randomness. This is equivalent to about 66% of polity-level variance explained by past views, since the variance of the irrational distribution is about 12 and the observed variance is about 4.

The parable explained

At a minimum, about 66% of the variance of political views can be explained, indicating that voters are mostly not irrational. Now you have the tools you need to understand why this is a dumb argument from Bartels:

The ideal of rational voting behavior is further undermined by accumulating evidence that voters can be powerfully swayed by television advertising in the days just before an election. A major study of the 2000 presidential election by Richard Johnston, Michael Hagen, and Kathleen Hall Jamieson tracked prospective voters’ responses to changes in the volume and content of campaign ads as well as to news coverage and other aspects of the national campaign. Their analysis suggested that George W. Bush’s razor thin victory hinged crucially on the fact that he had more money to spend on television ads in battleground states in the final weeks of the campaign.

How different is Bush from Gore, actually? If the range of political opinions is -3 to 3, and Bush is at 0.2, and Gore is at -0.2, and Bush got less than a percent of people to flip from -0.2 all the way to 0.2 with an ad, how much does that actually say about voter irrationality?

Can I get voters to vote for my plan of Total Economy Death in 2024 with an ad? That would be “powerfully swaying” voters. Bartels does not prove that “voters can be powerfully swayed by television advertising.” Rather, it’s clear that a few voters can only be weakly swayed by television advertising. Voters are mostly rational.

Keith Woods says

Democracy does not produce responsive governments, and that all of the intellectual defences of Democracy fail when put up against what we know about voter behaviour.

If preferences are X ~ N(0,1), that means that the majority view is 0. Therefore, a maximally responsive government is always at 0. A random government is in between -2 and 2 95% of the time. If democracy causes the government to be between -0.20 and 0.20 95% of the time, does democracy not produce responsive governments?

Bartels argues that the government randomly varies between -0.20 and 0.20. Woods says this means democracy does not produce responsive governments. But isn’t it much more responsive than alternatives? Just like how in the eugenics parable, there is much less variance under eugenics, even if it isn’t exactly 0?

What we know about voting behavior implies a democratic government will be at most 1/3 as irresponsible as alternatives. How then does democracy “fail” by this metric, compared to alternatives?

Now you can understand what Keith misses here. Actually, 75% of voters know who their PM is. I can’t name the PM of Australia. Furthermore, Australians do not vote directly for their PM like Americans almost do for their President. De jure, the PM is actually appointed by the Governor-General who is appointed by the King of Britain. Australia is, on paper, not a democracy at all. In practice, the Parliament makes a recommendation to the Governor-General, who obeys the recommendation. The Parliament is like the US Congress and consists of local members elected by small constituencies. This would be like if the House of Representatives appointed the President in the US.

In real life, the House and Senate appoint Speakers which lead the chambers. A PM is perhaps most like that position — and I’m not sure who the Speaker of the House is currently because I don’t vote for that. I do know that I will vote Republican for my Representative next election, because I am against woke and other Democratic gayness.

You can also see why the Agrabah question is stupid. My life did not improve in any way after the war in Afghanistan ended. How did that war effect me? Moreover, why would I need to know the entire geography of Afghanistan to whether I support the US drone bombing random foreigners?

On Party leaders, I think you can decide whether you support Republicans or Democrats without being able to name the DNC chairman. Who cares?

On branches of government: dare I say that again this doesn’t matter. An American voter casts four votes that are relevant to the federal government: he votes for a Representative, two Senators, and the President. Say you genetically care a lot about abortion. The economy never seems to change much depending on who you vote for, probably because democracy reduces government variance.

One party supports baby murder and the other is against it.

Do you need to be a constitutional scholar to decide who to cast your 3 votes for?

Obviously not.

Same issue here as above. Maybe psy-ops work a little bit. But they don’t work that well. Imagine a flock of sheep that are frequently slaughtered by wolves. One day, a eugenicists modifies the sheep to have sharp horns that can impale a wolf. Wolves get impaled and stop attacking, but the sheep are still plagued by ticks that suck their blood. A blank slatist says “look, the horns don’t matter because the wolves disappeared, and the sheep still let the ticks suck their blood. What retards!” The sheep know when a wolf is going to rip their organs out. They don’t need to be zoologists to do this. Public judgment is never “massively distorted” — recall the Bush and Gore example.

This one is a little different. Keith is making a logic error here. He says because the masses sometimes have false positives, that this means that don’t respond to bad economic performance. In other words, the masses don’t vote out politicians that totally loot the coffers, because sometimes they vote out politicians who merely oversaw, and did not cause, an economic downturn. This is fallacious reasoning. If the masses vote politicians out when the economy goes bad, then if a politician destroys the economy, he will be voted out of office. Therefore, even if the other guy would have done worse, a politician knows he will be punished if he ruins the economy.

Conclusion

Contrary to Keith, Democracy does produce mostly responsive governments, and the intellectual defense of democracy which argues that democracy increases government stability is compelling. In fact, countries with strong records of democracy are far less likely to experience civil wars than hybrid regimes. This is what voters being mostly rational predicts: the likelihood of an illegitimate government coming to power is greatly reduced. Thus, there is no cause for a civil war, which by definition is the violent removal of an illegitimate (i.e. an unpopular and insecure) state.