Wednesday, April 20, 2005

Mammals for Better Economics

Fellow big mammal battlepanda has some reflections on the discipline of economics.

1. Economics is not like the other sciences. Acknowledge that.
Battlepanda points out that you really can't do the lab thing in many areas of economics (though there are experimental economists busy at work in some areas), and mostly she's right. It's certainly not like physics. But one might be able to successfully liken it to epidemiology. Epidemiologists don't go out and start epidemics to test their theories about how disease progresses. They learn from past cases of epidemics (and a whole bunch of other biological information) and use this information to give advice about how to deal with future cases. People who do macroeconomics are in a similar position.

We make broad stroke assumptions such as "each individual will act to maximize his utility" or "if the price is driven down to zero, demand is infinite" not because we are intellectually lazy, but because making those assumptions allow us to simplify a real-life situation the point where we can apply logic.
Enormous danger lurks here. Being able to "apply logic" -- or more precisely, being able to use sophisticated mathematical techniques -- on top of false simplifying assumptions can be worse than trading your Econ textbooks for whiskey and never thinking about demand curves again. When you confidently apply fancy math to screwed-up empirical assumptions, you come out with full-fledged false beliefs about how the world works. Then you walk zombie-like through the world, using your math-derived cred to impress and terrify people until they embrace your false and destructive beliefs. Ye Gods, let this not happen unto me! Give me rather the honest second-order knowledge that I have no first-order knowledge! (Or even better, some genuine first-order knowledge. And yes, this preference ordering is transitive.)

No doubt, simplifying assumptions that make the calculations easier are needed in lots of fields. If they don't change the results, it's all cool. But I'm worried that economists have a tendency to accept false conclusions derived false premises because the structure of the argument is so mathalicious. This tendency must be overcome. It's the same kind of tendency one sees in lots of other fields (evolutionary psychology, anyone?). Somebody falls in love with a cool mode of argument and suddenly they're applying it in the most dubious places.

The government has put a ceiling on the rent of properties in Jenny's town. Is Jenny better off? Do you even need to look at the graph? That would be a bit of a waste of time, wouldn't it, since this is blatently a "government-actions-have-unintended-consequences-that-end up-hurting-the-very-people-they-are-trying-to-help" question.
Somebody falls in love with a cool mode of argument...

As I've told libertarians before, ours is a messy world. We've got collective-action problems, the diminishing marginal utility of money, asymmetric information, monopolies, monopsonies, differences in bargaining power, and all sorts of wacky things that require massive empirical investigation before you can even think about how theory should be applied. Be careful out there.


Battlepanda said...

Gee. I never conceptualized werewolves as mammals. But I guess you'd know better than I.

Yes. Trying to overcome incorrect assumptions with lots of cool maths is a dangerous thing. Sigh. If only economists would accept their place in the world as social scientist who just so happens to deal in numbers.

Dennis said...

As the local mathematician, let me second Angelica's assertion (I actually want to write a post over at my own blog about this very thing, but haven't had the time). At the same time, let me speak up in defense of simplifying assumptions. We have to make these a lot even in straight math (as in, "well, I can't show x (which is true, or possibly even false), but I can show y, which is like x. Maybe that's good enough."), let alone when we interface with trying to do things in practice. Just an hour ago, I was typing up notes on factoring algorithms, one of which assumes that a very nonrandom function is random. This is completely false and nothing close to it can even be proved, but it works. In probabilstic modelling, it's very very common to assume that obviously nonindependent events are independent to get formulas and algorithms that work just fine.
Even the example that Neil gives about demand at price zero is a great example of a good simplifying assumption -- obviously false in the real world, but nothing actually costs nothing, and we can easily believe that prices *near* nothing will increase demand a lot. So meh.

The subtle thing is figuring out exactly what the world of actually doing things will accept -- this is exactly the realm for the three intimately related real-world disciplines of science, engineering, and art. And of course the subtlety is something that needs to be chased carefully, but this is of course what we have economists for...

Neil Sinhababu said...

Since humans and wolves are both mammals, and mammalhood isn't something that probably doesn't exclude dramatic changes in form, I'd say that werewolves are mammals.

All the italicized stuff, including the "if the price is driven down to zero" comes from Angelica, who also happens to be battlepanda.

The real thing I'm attacking here is a psychological tendency on the part of some economists whom I've talked to -- a tendency to hastily and carelessly accept unreasonable simplifying assumptions (about elasticity, for instance) in order to get to the mathy stuff that they're so proud of being able to do. It's as if they thought the purity of math would be sufficient to make up for the flaws in their assumptions, when really the math just seals the flaws into the conclusion.

Dennis said...

Neil: I apologize for misattributing the good Panda's words. Of course, I should also say that I completely agree with your comments about bad assumptions too (my favorite is in fact not elasticity but perfect information) -- that's why I believe in the science/art part of things (i.e. that it takes creative talent and empirical work to make things work). I just want to say that assuming blatantly false things is an important part of that game.

At the same time, misappropriations of mathematics bother me almost more than anything else in social science. At least economists who make bad assumptions prove theorems -- being the guy I am, I don't much care if the theorems relate to reality or not -- and it's just their inappropriate application that screws things up. What I really can't stand is the misuse of the mantle of mathematics (like the making of random two-by-two tables) by people not doing any math, which doesn't quite apply to all economists, but does sometimes seem to apply to enough.

Rousseau said...

I think I made my feelings on economists being pretentious pretty strongly a while ago.

They're an important social science, and most economists who are actually professors or analysts acknowledge that they are nothing more. (There's the old saw "all government professors want to be economists, and all economists want to be mathematicians". Numbers envy really).

It's those who do act like we're have proven orthodoxies that are bad, since the lack of science isn't the sin, but a claim to science and pointing out someone who lacks it. But that's pretty normal rhetoric for politicians and pundits. I hate people who say socialism is clearly bad because the USSR failed, but these are the same people who say sexual-freedom is bad because of the 60's. I don't take them to represent real economics.

The problem is panda's remarks seem largely anti-intellectual criticisms of any science. Physics, biology, math, CS, all often use simplifying assumptions. And "the man in the street test"? Are you kidding me? Etc.