Sunday, May 29, 2005

Economic Science

What I find odd about economists sometimes is their persistent need to declare their work scientific. Is it a sign of some deeper insecurity that economics may not in fact be science? Do economists hope to bolster their findings and gain more credibility? Case in point: See this page with a picture of Ed Prescott and Finn Kydland (2004 Nobelists) listed as "Scientist" and "Chief Scientist". Or see The Collected Scientific Papers of Paul Samuelson. The Nobel Prize in Economics is officially titled "The Bank of Sweden Prize in Economic Sciences in Memory of Alfred Nobel". The Nobel Prize in Physics is just that. Moreover, you don't often hear physicists and chemists referring to their work as physical science or chemical science. (Well, you do for the former, but it is used when a broader context is denoted.)

One of the most interesting parts of Guns, Germs, and Steel by Jared Diamond was the very last part in which he discusses "the future of human history as a science:
One cannot deny that it is more difficult to extract general principles from studying history than from studying planetary orbits. However, the difficulties seem to me not fatal. Similar ones apply to other historical subjects whose place among the natural sciences is nevertheless secure, including astronomy, climatology, ecology, evolutionary biology, geology, and paleontology.
I would add cosmology to that list as a caveat to his next point:
People's image of science is unfortunately based on physics and a few other fields with similar methodologies. Scientists in those fields tend to be ignorantly disdainful of fields to which those methods are inappropriate and which must therefore seek other methodologies--such as my own research areas of ecology and evolutionary biology. But recall that the word "science" means "knowledge" ... to be obtained by whatever methods are most appropriate to the particular field.
Typically these sorts of discussions lead down a dark and pointless parth about the definition of "science" but I think such arguments are disingenuous. We can define it however we want such that whatever fields we want meet that definition, there is no natural definition. Besides, no field should be taken more seriously simply because it is termed "science". All arguments should be evaluated based on their merits.

Diamond continues by pointing out some aspects of the divide between "historical sciences" and "nonhistorical sciences":
Historical sciences in the broad sense (including astronomy and the like) share many features that set them apart from nonhistorical sciences such as physics, chemistry, and molecular biology. I would single out four: methodology, causation, prediction, and complexity.
In a nutshell, historical sciences cannot run controlled experiments, historical sciences care about ultimate as well as proximate causes, a priori predictions about future events are problematic if not impossible, and the systems studied by historical sciences are often complex, chaotic, and probabilistic.

I think that these four properties pretty accurately describe what separates economics from a discipline like physics as well, with the exception of the third one. Predictions on an individual level are certainly problematic...in the aggregate, less so.

Thursday, May 26, 2005

Evolution and Probability

Here is a good New Yorker article explaining and debunking the few theories supporting "intelligent design" that have attempted to disguise themselves with a scientific face. This is via Red State Rabble which has good coverage of all the goings-on in Kansas.

One of the primary arguments that the creationism/"intelligent design" movement offers against evolution is its improbability. The chances that life would have emerged from a chemical soup and then evolved into the complex structures we see today is miniscule, they argue. Something I have not seen argued is that by doing so they commit the elementary fallacy of confusing conditional probability with direct probability. That is, we shouldn't be looking at the probability that humans evolved, but the probability that humans evolved GIVEN that humans exist today. We should include all available information in our probability calculation.

I tried to do the calculation using Bayes' Rule but there's a lot of estimation for which I have little basis. There are a lot of numbers bandied about for P(humans evolved) but few (or none!) for P(humans exist | humans evolved) or P(humans exist | humans did not evolve) which you would need. But given that P(humans exist | humans did not evolve) is very very very small (which I think is reasonable since how many other options are there? Alien transplantation? God?) then P(humans evolved | humans exist) is not tiny at all.

(Keep in mind also that all these probability calculations should be done from the perspective of Earth 6,000,000,000 years ago. So P(humans exist) should really be P(humans exist 6,000,000,000 years from now), etc.

Monday, May 23, 2005

Thinking About Coase

Isaac made a comment last Fall in Micro regarding the Coase Theorem. Sans transactions costs, the Coase Theorem basically states that if you have two agents, A and B, and B is taking an action that has a positive or negative effect on A (an externality), A and B have the incentive to privately come to an agreement about the externality as long as property rights are well-defined. For example, if A are a group of fisherfolk (thanks, Larry) and B is a creek polluter the efficient outcome will result if you give A the right to have a non-polluted creek or if you give B the right to pollute the creek. This comes about through bargaining. Isaac's point was that the situation essentially gives A and B a pie of surplus to split up and that the bargaining process will eat up the whole pie. The logic behind this is that the loser in the bargain will always want to hold out for a little bit more. (If I'm not explaining this well please clarify!)

But if we put some structure behind this, the argument doesn't hold. I modeled this as a chain of Proposer-Decider games. In one iteration of the game, one player proposes a division of the pie and the other player can either accept or reject the division. If rejected, there is another iteration of the game, but with a smaller pie. The game continues until a division has been accepted or the pie reaches 0. We can determine the unique subgame perfect equilibrium by unraveling from the back. The result depends on who you give the power to. If you let one player be the Proposer in every iteration of the game then in the subgame perfect equilibrium you have that player initially propose to take the whole pie himself and the other player accepts that plan. If you let the players alternate in the roles of proposer and decider then in the subgame perfect equilibrium the pie is divided nearly equally by the first player to propose and the other player accepts. But in both situations none of the pie is wasted on costly extra rounds of bargaining.

I suppose the intuition here is that when the bargaining is set up in this way, each player wants to prevent the game from going on to the next stage in which they do not get to propose. In order to do this, A has to give B as much as B would get if the game did move to the next stage and then A gets the remainder of the pie. So the first player to propose divides the whole pie up in such a way that this occurs in the first round. The result where one player proposes the whole time is simple, they don't have to give the other player anything so they just take it all in the first round (like the one-stage Proposer-Decider game.) Of course there are other ways you could set up the bargaining...

Sunday, May 22, 2005

Blaming the beggars and still caring about poverty

It is consistent to think that beggars as individuals don't deserve our money and to still care about poverty. Not because of skepticism about where the money goes, but because of personal responsibility. Unless you are in a population where say 80% of people are beggars, then everyone had a certain chance of not having to be a beggar. So a given beggar is entirely to blame for his station in life. But when you compare two cities or countries and see wildly divergent rates of begging, then you realize that there are systemic reasons and so you might be interested in ameliorating those conditions that lead one place to be far worse off.

This is the same reason which says that you can hold a given minority individual responsible for having problems with getting tenure say, or into grad school, but still thinking that you should worry about the global figures: if they really worked hard and were superstars they wouldn't have a problem, they wouldn't be marginal, yet we'd prefer not to have that happen to marginal figures.

Friday, May 13, 2005

Star Wars?

Of course the Republicans have decided to appropriate Star Wars for their own ends. The 3rd episode opens this weekend, and Joe Scarborough on MSNBC is reading its potential success as a cultural validation of the religious right's Good/Evil dichotomy. Yes, people love it when Good beats Evil (not the case in Episode 3, however) in a movie but that doesn't mean that in the real world it is useful or true to see things in the same terms.

Wednesday, May 11, 2005

Labor supply and status

This article notes that Morgan Stanley's chief executive is paid $22 million and his new number two will be paid $20 million. But after taxes they're each taking home probably $13 or $14 million. This is generally true: salaries are always described pre-tax. So when comparing income people don't actually look at what people have to spend. Perhaps this is done because it is simply easier to not figure out a person's tax liability, but there is something else at work. Using a pre-tax figure means that the status of income is derived from the headline number ($20 million or whatever), and not the actual income. Given that you can reduce a worker's motivation to three parts -- income, status, and other -- it means that at least one of those isn't sensitive to marginal tax rates (insofar as different marginal tax rates don't impact compensation packages).

Saturday, May 07, 2005

More on Risk

I agree with Henry Farrell (and with Henry) that risk is fundamental. I actually think that risk explains everything in the world. A few examples:

--Morality: Michael Walzer's theory of how to fight a just war (as opposed to what makes a war just) is basically a willingness to expose your soldiers to some risk in order to save civilian lives.
-- Rebellion: James C. Scott argues that what caused rebellion in Southeast Asia was a changed tax scheme which shifted risk on to the peasants (a head tax as opposed to a percentage).
--Contemporary politics: Jacob Hacker is, well, really smart.

An interesting thing to note about risk and politics is what Richard Epstein -- a libertarian turning towards utilitarianism -- writes about John Rawls:
[T]he issue [of income distribution] appeared in the eyes of many to dominate his entire thinking about social institutions in that the central task of politics was to implement in connection with the "difference principle" some "maximin" solution whereby social institutions were altered in a direction that maximized the position of the least well-off in society, and thus compressed the wealth across individuals, without altering the rank order. In large measure this view was driven by an assumption that is not strictly required by a veil-of-ignorance approach, namely, one of extreme risk aversion. That position requires an extensive welfare state to engage in the needed redistribution from which it is just a short step to claiming that the good Rawlsian must accept a universe that is rich in positive rights — the right to housing, health care, education, and minimum subsistence. The hard question for the future lies in the extent to which Rawls's general veil approach can be disentangled from the particular prescriptions that he defended or which were subsequently defended by his many followers on the left. My own admiration for Rawls's work stems from the strong conviction that this separation can in fact be made.
In this account, the split between "liberals" and "libertarians" is an assumption about attitude towards risk: liberals think people are basically risk-averse and libertarians think people are basically risk-loving. The corollary is that conservatives know that people are risk-averse, but don't care (why else would they want to shift risk from business and government to little people?).

How (Not) To Value Human Life

The following argument is deployed to value human life: take two equally qualified people working identical jobs except that one person is forced that year to play a round of russian roulette with a gun with a barrel of 1,000,000 places with one bullet. Our russian roulette friend is paid more than our friend in a completely safe job. Take the present value of the difference in pay between the two, say $5. For that extra risk he has been paid $5. We can then say that this human life is worth...1,000,000 x $5 = $5,000,000 because if one one-millionth of that human life is worth $5, then our best guess of the value of each fraction of that life must be identical.

But that would be patently wrong. Here's why. I may well pay $5 for a one in a million chance of winning $5,000,000, but I would never pay $2,500,000 for a one in two chance of winning $5,000,000 (I would much rather have that $2,500,000). If $2,500,000 was all my savings, my life say, then maybe I'd take a pay-out of $10,000,000 or $15,000,000, but even then probably not. Something happens when we scale up, the Friedman-Savage hypothesis: in general, we are risk-loving with respect to the possibility of small losses and risk-averse with respect to possible large losses (when the large losses represent a substantial portion of net wealth: Bill Gates would happily risk losing $2,500,000 for a one in two chance of winning $10,000,000).

If the Friedman-Savage hypothesis holds, which seems reasonable enough, it means that the "value" of a human life will depend solely on how large a fraction of it you are measuring. The larger the fraction (the bigger the risk), the more human life will seem to be "worth." Yet whatever number you come up with by simple multiplication will be total bull because it doesn't take risk into account. Even if you try to consider risk, you really can't do it in any meaningful way because you have no way of knowing what happens when you move from a 50 percent to a 75 percent to a 99 percent chance of death in terms of the risk premium involved. And even if you did find some sample of people being paid a premium for a 99 percent chance of death, you'd really have to wonder if you aren't working with a very adverse selection. Perhaps the value of human life is incalculable after all.

Tuesday, May 03, 2005

Conservatives and Moral Hazard

Henry Farrell writes in response to Matt Yglesias' call for Democratic elevator pitches:
I don’t have an answer for him; I’m not that good at snappy one liners. But I do have a strong feeling that the answer lies somewhere in the left and right’s attitudes towards risk. Modern conservatives tend to fetishize risk as being a good thing in itself – see John H’s extended interrogation of David Frum’s claims that the risk of hardship and privation are character-building. There’s something similar going on in the insistence of many conservatives that the welfare state destroys character, and that Social Security, universal health care, bankruptcy protection and so on are bad things in themselves. This points towards the same conclusion as the libertarian argument that markets are good, but for different reasons – it’s less concerned with increasing individuals’ ability to make choices, than ensuring that they’re exposed to the brisk winds of chance and market forces. The left wants to provide a safety net in case you fall from the tightrope, but for people like Frum, the risk that you’ll break your neck is a good thing. It concentrates your mind wonderfully on staying up there, and makes you bulk up your moral fibre.
But what I have seen no one mention is that this conservative argument for risk (or rather, against state-sponsored alleviation of risk) is just the usual moral hazard argument taken to a much more extreme level. Since the conservative modus operandi is often to take ideas from economics and apply them to inappropriate situations (efficiency of the market mechanism, Laffer curve, etc.) this should not be surprising.

The problem is that if someone can get welfare if they are jobless then their incentive to retain a job is diminished, and if Social Security is available if they don't save enough then their incentive to save is diminished. This is almost exactly like the moral hazard problem in insurance markets: once insured, people are more willing to take actions that result in cashing in their policy. The difference is that in the insurance story you get missing and incomplete markets and in the conservative story you just get more risky behavior. So there is some legitimacy to the argument.

But I think there are a few exceptions that escape this critique...very basic poverty relief is one, and universal catastrophic injury/damage coverage is another. There is so much disutility associated with these situations that no one will want to be put in them even if they are compensated for their material losses. I suppose as you move up the ladder of insuring people against less and less undesirable states incentives are distorted more and more. The conservative critique, as usual, doesn't recognize that there is a middle ground at which the socially optimal result is achieved.

Monday, May 02, 2005

Federalism

The argument for federalism would seem to be: people have a limited ability to know; good policy requires significant amounts of knowledge of the specifics of the community in which it is to be enacted; the larger the government unit, the less likely it is to know and be sensitive to those specifics; therefore it is best if policy/law happens at the more local level. There is the corollary argument about power: power should be exercised close to where it came from to maximize accountibility. In theory there is nothing objectionable about these two points -- the only question would be to what extent does local variation for different policy issues matter(for example, does the optimal curriculum really vary from state to state or school district to school district? And is the variation we observe reflective of these legitimate differences?). Taking this seriously would require that states be redistricted occasionally to keep them approximately the same size so that lawmakers are equally close to where policy is enacted (wouldn't want Wyoming to be better governed than California), though there would be the concern about loss of tacit knowledge upon doing so.