Blog: Economics

Getting deeper into The Wisdom of Crowds... Surowiecki discusses the "Ultimatum Game" where two people are placed in a room and one, randomly assigned to be player A, is given ten $1 bills and must split them in some proportion (equal, unequal, whatever) with player B. Player B then gets to choose to accept the deal, in which case each of them keep their amount of money, or reject the deal, in which case neither gets any money.

The economic game theory ideal, the perfectly rational actor, would as player A choose to give $1 to B and keep $9. B would then have to choose between keeping $1 and getting nothing, and therefore take the deal. (There's a classic game theory problem often assigned in economics classes involving five pirates).

In real life, most player B's will reject offers of $1 or $3 even though it is not in their rational best interest to do so, because they perceive it as unfair. And most player A's, even without knowing the player B, will offer a fifty-fifty split.

I already knew that stuff. What I didn't know, and is fascinating, is that this only holds when the roles of A and B are randomly selected. If the players are told that A performed better on a test, then the player A's offer less money and the B's overwhelmingly accept their offers. If people have any reason, even an irrelevant one, to believe that one person is "better" than the other, their insistence on fairness goes out the window.

In a college Economics class, the professor told a story of a dessert table he saw which had but one cookie left on what was once a heaping plate of cookies. He decided not to eat the cookie, and a friend asked why. "That cookie has an adverse selection problem," he replied. He didn't know whether the cookie was good or bad. It's possible that other people knew something he didn't - maybe a child touched it after wiping his nose, for example. Most likely it just happened to be the last cookie, but the chances of that cookie being good to eat are lower than it would have been if there were many other cookies.

Adverse Selection is a concept in economics that deals with asymmetric information. Say a car dealer has two kinds of cars, lemons (bad ones) and cherries (good ones). In the absence of lemon laws, the dealer is better off selling you the lemons since you can't tell which is which. Therefore, smart buyers will only be willing to pay the value of a lemon, even for a cherry. There's then no way to sell the cherries since nobody is willing to pay what they're worth, not being able to assure himself he's actually getting one.

Asymmetric information crops up all the time with room postings. Some rooms are good, others bad. Reading a posting gives some information, but not much. Looking at the room gives a lot more, but the people living in the apartment know much more, like whether there are mice in the walls and whether the neighbors are really noisy. And they don't have an incentive to really talk about these things unless you ask the right questions.

If a room has gone for weeks without being rented, a smart housing seeker might surmise that it's likely to be overpriced, as others before would have seen the place but decided not to rent it. It's possible that it just happened not to be the right room for the people who saw it, just like the cookie most likely was just the last cookie by random chance, but there's no way to know whether others had more information or not. There's asymmetric information. And so, an ad that has been up for weeks is less appealing.

Therefore, on boards like Craigslist, savvy posters with rooms to rent will make their ad sound like it's just been posted for the first time, no matter how long it's really been there. If successful, this strategy in effect hides a piece of information from the reader, removing his or her ability to judge the rental based on those who have come before.

Blogshares is pretty cute. It's a play market where you buy and sell shares in blogs, which are valued based on how the links pointing to them in a PageRank-like algorithm (links from more valuable blogs are themselves more valuable).

They have a top 100 which is a nice summary of top blog sites (though it includes stuff like the New York Times which isn't a blog per se).

There are a lot of ways it's not really like a market - the prices are influenced by trading activity, but aren't set by a bid and ask spread with market makers. Instead, the software artificially pushes it up or down when the price gets too far out of sync with the "actual value" based on the inbound links.

Anyway, markets (theoretically) reflect public attitudes and anticipate foreseeable trends, so it's an interesting exercise in predictive modeling.