Game Theory and Dissent

A while back I posted a link to a critique of the misuse of game theory. Here's another example of what I mean.

There's this notion that it's "unpatriotic" to criticize the government's conduct of a war. Of course, one can simply reject this version of patriotism, but it's important to understand why it exists in the first place.

One reason is that it has some logical coherence. Imagine war as a game of chicken. Each nation drives a car at full speed toward the other nation, and if one swerves it loses. If neither swerves, both lose (and the outcome is worse than losing by swerving). This obviously doesn't capture the intricacy of war, but it describes one salient feature.

Now, one sure way to win is not to have a steering wheel (so long as everyone knows you don't have one). Your opponent would rather lose than die, so he swerves. Democracy is like a steering wheel: it is the choice of doing one thing or the other. You can see that this is a weakness of democracy, in a sense. In some cases, it is possible that we would be better off if we could bind ourselves to a policy that we couldn't overturn with a vote.

All of this is fine, but at this point certain hawks stop the analysis. Democracy is a disadvantage, they proclaim, or at least dissent during wartime is a sure way to lose. The only way to win is to remove the steering wheel or make it very clear to everyone that you have no intention of using it.

So you can see some obvious problems cropping up. One is that, in the general run of things, you want a steering wheel. Another is that, in this country, the steering wheel can't be removed. The hawks' theory is an example of one that is correct only if it is completely accepted. Which is to say, if they could truly remove the steering wheel, the strategy could in theory work (the informational demands are still unrealistic, and there are countless other problems, but ignore that for now). Partial acceptance of their arguments is the worst outcome.

A great example of this is the Church's HIV policy. Antipathy to condoms causes the deaths of lots of people. Yet some Catholics don't see any tradeoff between doctrine and humanity: after all, if everyone would just adhere to Catholic sexual doctrine, HIV would be a minor problem (you would have to clean up transfusion practices, but that's a fairly uncommon means of transmission I believe). The problem is that the real-world Catholic doctrine is being defended with a model of ideal human behavior. It's idiotic and cruel to base policy on the assumption that humans won't sin or make mistakes. So we get the worst of both worlds: Church officials are unable to get 100% compliance with Catholic sexual doctrine. Nevertheless, they insist on pursuing their anti-condom strategy, and resort to all kinds of unethical behavior (for instance, telling Africans that condoms won't protect them from HIV). So imagine a man about to have sex with an HIV-positive woman. He asks a priest if, given his decision to have sex, he should use a condom. The priest says no, and goes home satisfied in his knowledge that Catholic doctrine is great at stopping HIV. This is not a hypothetical: the Church tells married couples not to use condoms when one of them is HIV positive.

The same thing is going on with hawkish "patriotism." Of course, if we could behave in the "ideal" way, it still wouldn't be worth it (as I pointed out earlier, a steering wheel is a useful thing to have). But even for the limited purposes of the game of chicken, it's simply not in the cards for Americans to acquiesce in a misguided war. Given that the steering wheel remains firmly in place, calls for us not to use it are insanely dangerous.

All of this leaves aside that most of the hawks are not in the car, that chicken isn't a very good game to begin with, and that our enemy is unlikely to think in these terms. My point is just that hawkish patriotism has a theory behind it, and that theory is correct on its own terms. We need to be aware that this isn't enough, that a perfectly good theory can lead to horrible outcomes when people use it stupidly.

A Salt on Liberty

I like this article in the New York Times about salt iodization in Kazakhstan. Apparently iodine deficiency is serious, and iodizing salt is extremely cheap. Kazakhstan made iodization of table salt mandatory in 2003.

A quick reader survey: how many of you think mandatory iodization is a good idea in these circumstances? How many of you think it's improper for the government to intervene at all in cases like this? You are in a unique position to confirm or disconfirm my suspicions about the libertarian instincts of my readers. Don't take this as an attack, take this as an opportunity to articulate the proper role of government in circumstances like these. And don't worry, I don't plan to rehash the terminological debate, I just want to know where you sit.

[UPDATE: Judge Posner considers the merits of a ban on trans fats. Many of his points would be valid in the argument about iodization, except that I take it that the arguments for iodization are much stronger than the arguments against trans fat. Incidentally, Judge Posner considers the trans fat ban to be something libertarians are likely to disagree with, which is a data point in favor of my definition of libertarian.]

Simpathy for the Devil

So I've posted recently about society's use and misuse of economics. Another example, one that is particularly disappointing to me, is the set of algorithms used to run SimCity 4.

If you never played the SimCity games, they put you in the position of a mayor running a city. You zone land and provide city services, and if you do a good job your city will grow and thrive. The original game was quite simple (though, realistically, simple rules and components could yield complex systems), but by SimCity 4 you had a wide variety of power plant and mass transit options, three different densities of each zone, water pipes, schools, libraries, farms, and plenty of other buildings I can't remember. The complexity had increased greatly. You had to worry about waste disposal, sewage, education, and plenty of other factors.

It... was... awesome! Or at least it seemed awesome. You could design a fairly complex working city, and in doing so you had to solve countless problems that cropped up. You revised your strategies, you learned new ways of structuring your city, you generally had a blast.

If you were inventive enough (I usually wasn't), you discovered weaknesses in the algorithm used by the game. So for instance, raising taxes reduces "demand" for certain types of buildings, but "demand" only affects new growth, not pre-existing buildings. So you could always raise your taxes to the maximum, raise some money, and then lower them again when it was time to grow your city. I haven't actually tried this, but I assume a post toward the end of this thread is reliable.

It also turns out that the traffic algorithm used by SimCity is unrealistic. Residents don't choose the shortest commute in terms of time. I'm not sure what they do exactly, but I think they simply minimize geographical distance traveled. So for instance they would rather drive on congested streets directly to work than walk a block out of their way to take a subway.

Now, admittedly the system is still complex and challenging. You could play the game and just pretend you live in a universe in which residents have a bunch of irratinal behavioral quirks. This would be difficult for me, since I would much rather play a game with rational citizens. More importantly, though, I get the sense that the rules have been jury-rigged to achieve some semblance of rationality, but have created all kinds of unintended consequences and perverse effects. So in other words, the developers wanted to simulate actual city dynamics but for whatever reason didn't use basic economic principles to write the algorithms.

The result is that you have layers of meaningless crap. "Demand" isn't really demand, so they layered on the concept of "desirability." Planting trees raises desirability, but it's a temporary blip (which is quite strange; trees should add more to desirability as they mature - think of the magnificent trees on many college campuses). The list could go on forever - the more I read, the more examples I find.

I'm not insisting on perfect economic realism. I don't mind if things are simplified or if residents sometimes behave irrationally. I also understand if things have to be smoothed over for ideological reasons (what are the consequences of a tax hike, or increased spending on education?). The basic pattern should be a reasonable one, though.

As a quick side note, it would be pretty awesome if developers came to you with plans, asking for help with eminent domain or tax breaks. That's highly realistic, and figuring out which plans to approve could be quite a challenge.

It's especially frustrating because I don't think it would be much harder to do it the right way. The developers got so many things right - it really is an exquisite game in some ways - that it is especially galling for the game to be ruined by this one factor. Now admittedly, some things are computationally difficult. Traffic algorithms in particular might be tricky. Still, it's hard to escape the feeling that ultimately the game's algorithms were designed largely in ignorance of basic economic principles.

Finally, imagine how awesome it would be if more sophisticated economics were brought to bear. There was a hint of this in SimCity 2000, if I remember correctly. You could lower taxes on car manufacturing in anticipation of the car boom, then raise them when the boom happened. This is a crude version of what could be a much more thorough use of path dependency. Imagine having to choose whether to go into hog slaughtering or steel. It would be hard to predict that one would lead to Chicago and the other to Pittsburgh (and if the game were well-designed, neither would be automatic, it would truly be path-dependent).

Such wasted potential! I have hopes, though. First, users can design their own "mods" that alter the algorithms of the game. In theory, someone might have already designed reasonable mods. More importantly, a new SimCity may be in the works (see the end of this entry). Hope lives on!

Tarun Menon Society is Up

Check out the Tarun Menon Society The rumor going around is that the first discussion will be on libertarianism economic geography.

Fin

So look, terminological debates are usually pointless. I gave my reasons for my definition of libertarian, but I'm not married to it. To some extent there's not a word for what people want to say about themselves, and libertarian has come to mean something pretty close.

I am more concerned that people seem to associate libertarianism with economic sophistication. Dave seems to be using the term in that way. The problem is that the term is too inclusive for this purpose. You end up attributing economic acumen to the extremists, or at least seeming to. You also stretch "libertarian" to include redistribution, Keynesianism, regulation, etc., but again, that discussion is pointless.

A final point: there is now a big struggle within the Democratic Party over its economic policy. Nothing you could do within the Libertarian Party could matter a tenth as much. Those of you who call yourselves libertarian because you like markets and don't hate gays, consider working within the better major party for now. Robert Rubin needs you.

p.s. if you actually think the Libertarian Party is better than the Democrats on the merits then you're crazy

Amateur Economics: Uses and Abuses

Two recent blog posts get at an idea I've been thinking about. Steven Levitt has a post about interesting ideas that are misleading in terms of policy choices. Matt Yglesias has a post about the necessity of sophisticated economic analysis.

Ultimately both posts are about the same thing: theory can be misleading when it is insulated from data (Levitt) or when it ignores more sophisticated economics (Yglesias). The posts converge even more when you understand that sophisticated economics is often very empirical.

This consideration should feed into economics education, but too often it doesn't. I can't count the times I've encountered obnoxious college students who know the right policy because of something they learned in introductory economics. On occasion I've been such a student. The problem is that on some level this is inevitable: you can't teach people a method of reasoning and then expect them not to use it. You can't give them a powerful tool and then tell them only grown-ups can use it safely.

Luckily, I think there's a way out of this dilemma. It's a matter of putting economics in perspective. The whole point of economics is to abstract away from the overwhelming richness of the world to the manageable austerity of mathematical models. To take the conclusions back into the world, intensive empirical work is needed. Students should see examples of economic theory failing in its application because of the facts on the ground. Good economics education already does this, but obviously plenty of introductory economics doesn't instill quite the sense of humility we should aim for.

Another way to put this is that someone who has taken introductory economics should be able to poke holes in bad arguments about minimum wages, for instance, but shouldn't try to reach a conclusion on the basis of theory alone.

Part of the problem is that this is not an apolitical subject. At various times in history various ideologies have benefited from the misuse of amateur economics. Marxism is one example, but today I think conservatives and libertarians have the most to gain. Treating introductory economics as a conclusive indicator of good policy tends to favor right-wing economics. That actual economics often points in the opposite direction need not bother conservatives. Most students won't progress to the higher levels of economics, and many students don't understand the limitations of low-level theory.

So conservatives and libertarians lead by example, proclaiming the manifest rightness of their policies and the ignorance of their ideological enemies while largely ignoring the actual implications of higher-level and empirical economics. For economics education to be useful, it has to alert its students to the dangers of this attitude. Otherwise economics will become something it need not be: a discipline useful to the experts but misleading to the casual student.

Amateur Game Theory Has Consequences

I really like this post about the misuse of amateur game theory, which Brad DeLong linked to. I've been meaning to post a similar criticism of the various ways game theory is abused, particularly in hawkish foreign policy circles.

My one quibble is that signaling doesn't have to be costly to be credible. This is often the case, but sometimes you are a type that can send the signal costlessly. An example is the use of language tests to root out spies. I can't find it, but somewhere in Postwar Judt gives an example of a dessert that Germans could never learn to pronounce (rodgrod med flode?). You could signal that you weren't a German by pronouncing it correctly.

Also, check out the Klein bottle.

Slate's Dumb Obsession with Contrarianism

[UPDATE: The Onion says it all]

So there's a Slate article arguing that the BCS system shouldn't necessarily set up a championship game between the #1 team and the #2 team. The article argues that we only care about who the #1 team is, so the game shouldn't necessarily feature a matchup between the #1 team and the #2 team.

I wrote a long post examining what it means to be #1, but this passage demonstrates the fallacy pretty clearly:

"Here's what we do know: Michigan is not the best. How do we know that? By the traditional criterion: They scored fewer points in a football game than Ohio State did."

But then, we also know that Florida isn't the best. How do we know that? By the traditional criterion: They scored fewer points in a football game than Auburn did.

That's not all we know. We know that Auburn isn't as good as Arkansas. How do we know that? By the traditional criterion: They scored fewer points in a football game than Arkansas did.

This rendered the SEC champtionship game, held last Saturday, a mere formality. Arkansas is better than Auburn, which is better than Florida. Florida confirmed its inferior status by the traditional criterion: They scored fewer more points in a football game than Arkansas did. Wait, what? How did that happen?

It's simply not true that the better team wins every matchup. We know this from the baseball postseason, in which many series are not sweeps. True, the better team is more likely to win a particular game, but we can't conclude from a single close game, won by the hometeam, that the winner is definitely better than the loser.

What we can do is, bearing in mind all evidence, come up with the two teams most likely to be the "real" #1 and let them play each other for the title. This is why we would normally match the #2 ranked team against the #1 ranked team: they are likely to be the two best teams, and one of them is therefore likely the #1 team. In this case I suspect that Michigan would be ranked #2 except that people are more interested in a Florida-Ohio State matchup.

Admittedly it would be different if Ohio State had blown Michigan away. That would be much more conclusive evidence that Ohio State really is better than Michigan. As it is, though, I suspect people are punishing Michigan based on the same flawed Slate-style logic that would lead us to believe that Arkansas is simultaneously better than and worse than Florida. The BCS is indeed flawed, and the best evidence is that Slate agrees with it.

The Perverse Libertarian Urge

So I've had this discussion several times. Why does it matter if non-libertarians call themselves libertarians? What else should they call themselves?

The truth is, I don't care too much, I just find it irritating. Libertarianism gets a free ride because people are lazy. I also think it would be premature to say that it doesn't matter - the psychology of party/ideological affiliation is complicated. I've heard, for instance, that most people identify with a political party before forming opinions on most policy issues. Causality doesn't always run the direction we might like it to.

Mostly, though, I just think it's silly. Stephen Colbert often asks Democrats and/or liberals whether George Bush is a great president or the greatest president. If they try to give another answer, he says, "Sorry, those are the only two options." When they refuse to choose one, he says, "I'll put you down for 'great.'" This is essentially what pseudo-libertarians are doing, except that they don't seem to get the joke.

More importantly, "libertarian" is a pretty bad alternative. It's ludicrous that people like Alan criticize liberals for being bad at economics, then turn around and call themselves libertarians. This is an error that is very much of the "out of the frying pan, into the fire" sort. Read some libertarian blogs and see how good their grasp of economics is. You'll rapidly come to appreciate that worshipping the market is different from understanding it. Among my favorites was an argument for privatization of the police force. Now, I understand that reasonable libertarians don't believe this. Partly, though, that's because reasonable libertarians aren't libertarians.

In fact, ultimately the real problem with calling yourself a libertarian is that it associates you with libertarians. It's the same reason I would never call myself a Green even though I agree with many of their positions (I assume they favor higher gas taxes, gay marriage, etc.). I'm not really arguing in favor of the term "liberal," I just think it's vastly better, because more reasonable, than the term libertarian. Ultimately I think people are just being lazy, and if they delved deeper they would recoil from libertarianism.

[update: fixed typo]