Price Dispersion in Competitive Markets

During the past couple of weeks, gasoline prices in our area have ranged from 97 cents per litre up to $1.35 per litre on different days in different, but neighbouring, towns. On one single day last week, in the space of under half an hour, I saw prices between $1.11 and $1.29 per litre. Two days later, the price range was between $1.05 and $1.15 per litre

Bryan Caplan noticed the same thing in the D.C. area:

With the whole country bemoaning the rise in the average price of gas, a far more economically surprising change has been almost overlooked: The massive rise in the variance of the price of gas. Before the hurricane, the spread between the highest and lowest price of gas in Northern Virginia was about 10 cents. Now a half hour drive to Manassas revealed a four-fold increase in the spread, with prices ranging from $2.99 to $3.39.

Many suggested explanations for the increased price dispersion were offered in comments to Bryan's posting, but as I indicated there, I didn't find the increased price spread at all surprising. Here is why:

Nearly every year, I do some market simulation studies (a game involving buying and selling) with my students. They have no information about the pre-determined market-clearing price. We call out and post the price of each transaction as it occurs. And after just one or two rounds, the players pretty much converge on the market-clearing price. Initially, however, when they have no idea what the equilbrium is, there are some transactions that deviate substantially from the market-clearing price.

How fast they converge, I have discovered, depends on several things:

1. How many are playing the game. The more there are in the game, the more information they receive as contract prices are called out. [increasing the sample size reduces the variance of the estimated mean for them].
2. How much they have at stake. If the prize is big, or if they have some pride invested in doing well, they converge on the market price quickly. E.g., pre-business students converge much more quickly than the surfers I once taught in Hawaii.

In other words, when there are strong incentives at work, price dispersions narrow quickly, which seems to be what happened with gasoline prices within the two-day time span mentioned above.

In my classes, after we are well into the simulations, I randomly remove about half of either the sellers or the buyers. As a result the students have no idea what the new market-clearing price will be. In these rounds of the simulations, the price dispersion increases back to where it had been initially, when the students had no price information. But it very quickly converges to a new equilibrium.

My guess is that's what happened with gasoline pricing during the past two weeks. There were large supply shocks (followed by demand shocks based on altered expectations). As a result, the price searchers [there's a reason we use that term!] weren't quite sure what the new equilibrium price would be. Some guessed quite a bit higher than others.

At the same time, consumers had little idea about whether the price at one station was above or below what others were charging, and so a wide variety of prices could survive for several hours or days before market forces brought them back closer to equilibrium.