[sfnhltb]

Interesting that 87.5% of the players didn't cheat, and the one that did bid 0.3% of the supposed equilibrium price, but this is somehow supporting the theory put forward.

It reminds me of the RAND corporation experiment, an economist and a mathematician play the game and reach Nash Equilibrium in a handful of moves. Same game played by secretaries and they all co-operate all the time completely defying all of the theories they build up like a house of cards. So in an economy where everyone studies math for decades and has no social skills we have a great tool to predict the economy, but for most of the real world isn't much help.

Sure it can work in certain financial markets with no contact between participants outside the market mechanism and most of the players can be assumed reasonably expert, so it has its uses, but a lot of care needs to be used when trying to take it outside that sort of area because the assumptions break down very badly in most situations.

[jimmybot]

It only takes one cheater for the collusive agreement to collapse.

[pyre]

True, and we do see this happen. Look at RAMBUS and SDRAM though it was a standards working group not a collusive group of price fixers.

There is something to be said for the consequences of breaking the agreement in the real world. In the classroom, she gets boo'd, but if she had been colluding with say 'the Mob' she might have to worry about something else after breaking the agreement.

This also doesn't take into account:

* Industries where the people at the top are part of an 'old boys club'

* An Industry were the conservative move is to collude and let the marketplace stagnate.

* In the classroom example, she had nothing to lose. She probably didn't even really care about the $20 or not. Whereas in the real world, people like guarantees. If a collusion agreement with players that you feel you can trust gets you a better guarantee than striking out on your own it's an attractive prospect.

* It could put a 'black mark' on you with other key players in your industry to play on their trust like that. Which could be a bad thing.

* This excludes collusion to to exclude a player. (e.g. the rumor that companies were going to break the FCC spectrum auction rules in an attempt to make sure that Google didn't get anything because it seemed that their plans were commoditize the industry.)

[jimmybot]

Yup, repeat games are a different situation that may lead to different outcomes. If you can enforce the agreement, it's a different kind of game.

But then consider situations where information is unreliable:

- Cheating because you can't trust others not to cheat, i.e. to protect yourself, and everyone ends up cheating, as in Prisoner's Dilemma.

- In a lot of situations, you may not even know who is cheating and by how much, but you might be sure someone is cheating. This will feed back into your own incentives to cheat. I believe OPEC suffers from this a lot.

[pyre]

> In a lot of situations, you may not even know who is cheating and by how much, but you might be sure someone is cheating. This will feed back into your own incentives to cheat. I believe OPEC suffers from this a lot.

In situations like OPEC, it's also really different. You don't have everyone on an equal footing like in the classroom game.

If Venezuela can only produce 1 barrel/day and Saudia Arabia can produce 50 barrels/day, it doesn't matter how cheaply Venezuela prices their barrel. But oil is also a market where the appetite of the consumer severely outstrips availability. As long as you can produce large quantities of oil you can pretty much name your price (so long as that's large quantities in relation to other producers).

[jellicle]

This is an example of one of the assumptions that break down very badly. In the real world, it usually doesn't take just one cheater for the agreement to break down, and the cheating is usually obvious to all participants, and results happen slowly (i.e., it's a repeat game scenario) rather than being a one-time event.

In other words collusion in the real world is MUCH easier than the scenario presented, and yet the students almost managed it anyway.

Suppose that the colluders are three (instead of eight) dominant telecom companies, who want to collude to keep cell phone prices high. Their prices are public knowledge, widely disclosed, and competitors can check each other on a daily basis. Market share - the prize - changes very slowly with changes in pricing - if you drop your prices 5 cents under your competitors, you may eventually win more market share, but it will happen veeeeeeeery slowly.

In this environment price collusion is easy to maintain indefinitely. It would almost be surprising if they weren't colluding.