Big Thinker - Charles Plott
All bets were on as Caltech Professor of Economics and Political Science Charles Plott gave a talk to a full audience at Yahoo!’s Mission College Campus – the fifth in the 2008 Yahoo! Research Big Thinkers series. His talk focused on how information aggregation mechanisms work and whether or not they are effective- specifically, in pari-mutuel betting systems.
Plott began with a scenario about fishing and questioned the audience about how a fisherman would find fish. The answers varied but were similar: look for the boats, look for the birds. Plott explained that by “watching a bird watching another bird, watching another bird, watching another bird, watching a small fish, watching a big fish…” you can conclude that there’s fish from a distance without any communication involved. Information aggregation happens when an event unfolds and signals are set off to people about the event. Everyone sees information around the event, but not the actual event.
Similarly, when people are hungry and in search of a restaurant, they turn to guides, ask people they know, or seek out crowded venues. In doing so, they are assuming that the folks in these crowded restaurants know something they don’t, and that they are rational. According to Plott, this concept is seen all the time, through the “wisdom of crowds and body language.”
He further explained that information in the signals differs from the information in humans. “Filtered by human observation, it exists subjectively as vague ideas, intuition and hunches,” he said. You can deduce, from their actions alone, what it is they know. However, is this really possible?
Plott turned the audience’s attention to the pari-mutuel betting system that is used to collect information that is widely distributed in small amounts. Recent field experiments have been conducted to test the extent to which: a) systems can aggregate information b) there is a process by which this happens c) it can be wrong, but signals are given when it is not reliable.
The pari-mutuel betting system is the accepted betting procedure at horse racing tracks, where investments on the winning horse yield returns that decrease with the proportion of bettors who bet on the same horse. The proportion of the money that is bet on any given horse can be interpreted as the subjective probability that this horse will win the race. The odds are determined by dividing the total amount of money invested by the amount bet on the winning horse.
His talk led to the conclusion that it is possible to detect when the information was good or reliable, through the way that the information was released. When the information was poor, people reduced their bets, further crimping the release of information.
People with good information want to wait before participating to reduce information leakage and make more money. By forcing participation – for example, increasing the price of tickets purchased later - Plott overcomes this problem and induces the markets to converge to accurate aggregate information. Plott is not just eliciting a point estimate but the entire distribution of possible outcomes.
