Randall ([info]xkcd) wrote in [info]economics,
@ 2005-05-30 11:58:00
Previous Entry  Add to memories!  Tell a Friend!  Next Entry
When everyone has the option to kill the goose
Hi, I'm new here. Math/Physics/CS student; my understanding of economics is extremely hazy, but I'm interested in game theory.

I was thinking about what seems to me to be a particularly pointed example of what I think is called the Tragedy of the Commons.

Suppose you have a group of people (a company) with a pool of money they can all draw from. Every month, each person can decide to withdraw as much money as he wants (all on the same day -- say, the 15th).

But at the end of every month, a computer takes the total amount of money left in the account and adds a zero to the end. If there's $100 left, it goes up to $1,000.

Now, each person has the option to try to take ALL the money in the account (if two people try to do this, they'll just each get half. See comments for plausible mechanism.)

But, if everyone holds off on taking all the money one month, the next month they'll have vastly more money. And if everyone can wait for over a year, never withdrawing more than half of the account, they'll all be millionaires. A huge win/win situation for all.

But, with a large group of people, the individual incentive for each person to withdraw everything -- to kill the goose that lays the golden egg -- becomes huge. And the tremendous danger of this makes even the honest people think about cashing out. With a lot of people, you can get a million dollars versus a thousand this month -- you've got everything you want. By waiting for another week, everyone else can have all they want too, but that might not happen -- because someone else might take it all. It seems a lot like the prisoner's dilemma, in this sense. I know my first impulse in this situation would be to kidnap the children of all the participants and hold them hostage to control their behavior. For their own good, you see. Anyway, it seems like an interesting situation.

Any articles you can point me to written on this idea, or real-life examples, or thoughts in general?



(Post a new comment)


[info]roony
2005-05-30 04:33 pm UTC (link)
This could very well be a strategic game; it reminds me in fact of a type of dynamic Bertrand game in one of my Economics courses. If you include a discount factor and assign payoffs to each player under each combination of actions, then you can see which Nash equilibria are sustained under different strategies.

(Reply to this)


[info]benkilpatrick
2005-05-30 05:14 pm UTC (link)
Two thoughts.
1 - Totally impossible in real life.
2 - Extremely illumintated none the less.

(Reply to this) (Thread)


[info]roony
2005-05-30 11:18 pm UTC (link)
Well, such games (or variations thereof) have been proposed to model investor behaviour during a run on the banks, or a speculative attack in a currency crisis.

(Reply to this) (Parent)


[info]smurfette184
2005-05-30 05:27 pm UTC (link)
Have you read Garrett Hardin's article entitled the Tragedy of the Commons? If not, its a quick read. Also, most environmental economics texts address this issue when talking about open access problems when related to natural systems.

(Reply to this)

Game Theory
[info]twgiv
2005-05-30 05:27 pm UTC (link)
I would check out this book. It's Avinash Dixit's undergraduate game theory text. It's very readable and goes into the theory behind various kinds of games as well as the mathematical background. The two concepts are kept separate and it really helps to grasp the concept from both angles. The sell used for cheap, I'd give you mine, but I refer to it too much.

(Reply to this)


[info]dieseldawn
2005-05-30 08:59 pm UTC (link)
yeah the first place to look to is what people are doing to environment. because the earth' systems are unpriced, people tend to exploit them, despite their obvious value, even though they know that if they take a cut in their standard of living and waste less, the world will be a cleaner place.

(Reply to this)


[info]jayguevara
2005-05-30 09:16 pm UTC (link)
You need to know the utility of money as a function of time and amount for each player.

(Reply to this) (Thread)


[info]mackave
2005-05-30 10:22 pm UTC (link)
Perhaps if you want to do a more realistic analysis.

However, under classical assumptions the utility of money and the discount factor are practiacally irrelevent. As long as the utility of money and the discount factor are strictly positive, then a rational player will withdraw all the money in the first period.

(Reply to this) (Parent)(Thread)


[info]jayguevara
2005-05-31 02:39 am UTC (link)
Can you elaborate on your derivation of that?

(Reply to this) (Parent)(Thread)


[info]jayguevara
2005-05-31 02:55 am UTC (link)
I guess I should have checked for other comments before I said that! I never made the assumption of a known, finite ending date - hence the confusion.

(Reply to this) (Parent)


[info]puf_almighty
2005-05-30 09:47 pm UTC (link)
Are the players allowed to establish rules governing behavior among themselves?
And can they track who took out all the money, and punish him, if someone does?

(Reply to this) (Thread)


[info]roony
2005-05-30 11:15 pm UTC (link)
The conventional assumption is that of perfect information. Defining a strategy that prescribes cooperation or defection is then left to the reader.

(Reply to this) (Parent)(Thread)


[info]puf_almighty
2005-05-31 03:00 am UTC (link)
They can all maximize their gains by cooperating. So if a group got together and coerced cooperation, that group and whomever joined it would naturally have the most gains. And those who resisted it would be hurt by the group, which was as a whole stronger than they were.
So it seems like the system, and any system that has big common resources, would naturally evolve government and cooperation.

(Reply to this) (Parent)(Thread)


[info]mackave
2005-05-31 03:53 pm UTC (link)
"that group and whomever joined it would naturally have the most gains."

The group as a whole would maximize total surplus, however, an individual could maximize her own payoff by deviating from the cooperative strategy, given everyone else cooperated. Thus, cooperating is not a Nash equilibrium.

In noncooperative game theory we assume agents cannot make binding agreements beforehand. Contract enforcement must be endogenous, providing incentives. It's interesting that you suggest evolution of government and cooperation as a way around this problem, but I think the induvidual will always have some incentive to cheat.

(Reply to this) (Parent)(Thread)


[info]xkcd
2005-05-31 04:47 pm UTC (link)
Hence, my first impulse was to take hostage something belonging to the other players :)

(Reply to this) (Parent)(Thread)


[info]puf_almighty
2005-06-01 02:21 am UTC (link)
But you do have something hostage! The entire bank account!
The problem is that everyone is equally able to punish everyone else, so without any standard established (anarchy) it's only beneficial to withdraw on the first day to make sure nobody else gets anything.

(Reply to this) (Parent)


[info]puf_almighty
2005-06-01 02:20 am UTC (link)
however, an individual could maximize her own payoff by deviating from the cooperative strategy, given everyone else cooperated
But this (the group) would change the rules so that her payoff was minimized by them killing her for breaking the rules. The value of exploiting would be changed.

Contract enforcement must be endogenous, providing incentives.
You mean, they have to stick to a contract voluntarily? That system wouldn't work in the long run because of the problems described here. It would fail and be replaced by a superior system in which a certain level of external enforcement made most breaches of the rules prohibitive (and which also naturally rewarded successful exploitation of the rules). And the resources available for efficient enforcement of the rules would determine how strong this standard was, so that for any given scenario, a system would naturally develop which was most approriate to the resources involved.

I'm a biologist. I think in terms of evolution.

(Reply to this) (Parent)


[info]mackave
2005-05-30 10:15 pm UTC (link)
This problem is well understood within game theory. The relevant branch for you to refer to is the literature on repeated games.

Here's the basic idea: Under classical asaumptions this game can only sustain a pareto superior nash equlibrium if it is infinitely repeated. That is, if there is some round in which people know (or even if the believe) the game will terminate then they will wait until that round and withdraw all the money. However, knowing this, a rational, selfish player will withdraw all the money in the previous round. By backward induction, players will all withdraw in the first round.

There are a number of ways around this problem, one being infinite repetition, and another being an uncertain ending date. Another possibillity is a richer space of agent types. Having imperfect information can allow the existenece of pareto superior equilibria

A cannonical reference is Kreps and Wilson "Reputation and Imperfect Information." The paper examines the chain store paradox, but the logic is more general. They also deal with an equilibrium concept (sequential equilibrium) which may be difficult if you haven't had much exposure to game theory.

However, most decent game theory textbooks cover the issue (e.g. Gibbons "Game theory for Applied Economists"). Just see the sections on "repeated games" or "reputation." Also see the related "folk theorem."

(Reply to this) (Thread)

addendum
[info]mackave
2005-05-30 10:39 pm UTC (link)
I should have mentioned, imperfect information may allow the existence of pareto superior equilibria even if the game is finitely repeated with a certain finishing date.

And I case you didn't know, by "pareto superior" i mean an outcome where at least one person is better off, and no one else is worse off. Another way to put it would be "more efficient." In your example this corresponds to an outcome where people wait until later periods to withdraw the money.

I also should have mentioned that having boundedly rational players is another possible way out of the situation. However, this is not necesarilly so, since there are many ways to model bounded rationality, and not all of them need have this result.

Also, I misspoke when I said the problem is "well understood." It is well studied, in that there are literally hundreds of papers on this and related situations, but maybe not well understood. There is still controversy, and as you make the situation more complex, the analysis gets more complex as well.

(Reply to this) (Parent)


Create an Account
Forgot your login or password?
Login w/ OpenID
English • Español • Deutsch • Русский…