Perspective on Peanut Butter: Price dynamics


A dominant class of economic theories is built on the assumption that prices respond only sluggishly to new economic conditions. It’s an interesting challenge to try to reconcile that premise with what we see in the data.

A number of research papers have now tried to describe the actual high-frequency dynamics of the prices of individual grocery items as reflected in scanner data. One of my favorites is a new paper by Yale professor Judith Chevalier and University of Chicago professor Anil Kashyap. Here’s a diagram from their paper of what one sees, for example, in the weekly price of an 18-ounce jar of Peter Pan Creamy Peanut Butter at a supermarket in northwest Chicago.

Figure 1. Source: Chevalier and Kashyap (2010).


That doesn’t look to me like a price that’s frozen regardless of economic conditions. Instead, one can count on periodic deep price discounts. The timing and magnitude of these is a bit hard to predict, and there’s the curious feature that, after a brief sale, the price usually goes back to exactly where it was before the sale. A paper by Eichenbaum, Jaimovich, and Rebelo suggests that perhaps the "regular" price is the object that responds sluggishly to economic conditions. It’s worth noting that in the Peter Pan graph above, the product was only on sale about one week out of five over this period. However, Chevalier and Kashyap find that these sales account for almost 40% of the ounces sold. For other products, the importance of sales is even more dramatic. For example, 12-ounce cans of Minute Maid frozen orange juice are on sale in 30% of the weeks, but those weeks account for 70% of the ounces sold.

And, even though Peter Pan might not be on sale at any given time you show up at the grocery store, its competitor Jif might be.

Figure 2. Source: Chevalier and Kashyap (2010).


I found Figure 3 below particularly interesting. The red, short-dashed line is the regular, non-sale price for Peter Pan that one would infer from Figure 1 above. The blue, long-dashed line is the regular price for Jif from Figure 2, and the green, short-dashed line is the regular price for a third brand. The solid black line is the average price consumers actually paid for peanut butter. Its dynamic behavior looks nothing like any of the three regular prices.

Figure 3. Source: Chevalier and Kashyap (2010).


So if we wanted to talk about "the" price of peanut butter, what price would we use? The diagram below gives 4 possible answers. The blue, long-dashed line is an average of the regular prices for the three brands. The red, short-dashed line is the average price you’d pay if you bought a fixed "basket" of the three brands each week, corresponding conceptually to what the consumer price index is trying to measure. Again, neither of these look much like the average price consumers actually paid for peanut butter (the solid green line).

Figure 4. Source: Chevalier and Kashyap (2010).


The brown, dashed-dotted line in Figure 4 is another concept that Chevalier and Kashyap suggest using, which they call the "best price." This corresponds to how much you’d pay for peanut butter if you bought whichever brand was on sale, and were willing to stock up on special discounts to store it for up to 5 weeks. Although that extreme bargain-hunting price would by construction be lower than the actual price paid by almost everybody, the dynamic behavior of the best price resembles more closely than any of the other measures the dynamics of the average price paid actually paid by consumers.

Although prices of peanut butter may exhibit dramatic week-to-week variability, most people’s wages behave nothing like the prices of items in a grocery store. An older Keynesian tradition held that product prices were in fact perfectly flexible and that sluggish wage adjustment is the key economic friction.

But there’s obviously something much richer going on with price dynamics than what is assumed in virtually all the models that macroeconomists are using at the moment. I’m quite excited to see papers like these that examine what’s actually in the data rather than try to build a theoretical edifice whose foundation is nothing more than the imagination of economists.

Price dynamics
James Hamilton
Sun, 20 Mar 2011 14:54:25 GMT


Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s