How Frequent are Small Price Changes?
Recently, Eichenbaum et al got this article published in the AEJ: Macroeconomics journal. The question "How frequent are small price changes?" Their answer: not very.
Let's ask the bigger question first: so what? For that, we need to do a little history of economic thought.
Traditionally, microeconomic theory suggests that prices are very flexible - so small price changes should be frequent, since demand for most goods isn't particularly stable. But, then JM Keynes came along in the 1930s and suggested that prices are actually quite "sticky" - that is, inflexible. The result is that firms respond to changes in demand not by altering prices, but by changing quantity. This played a huge role in Keynes's theory of depressions being caused by a lack of demand. (Standard story: Demand falls => lower prices => lower wages. No need for unemployment. Keynes's story: Demand falls => lower quantity => lay people off => DEPRESSION)
Here's the problem: sticky prices don't necessarily make sense. After all, if demand for my product falls, it probably falls for my competitors too. In that case, I can gain a competitive advantage if I am willing to change my price before they do. So, there's a natural incentive for businesses to keep their prices flexible. So, a lot of work post-Keynes has been trying to justify when sticky prices might make sense. One of the big reasons? If menu costs are significant.
Menu costs are a metaphor for any cost involved with changing prices. They are named, obviously, after restaurants that have to print new menus every time they want to change prices. This is costly, so restaurants won't change prices very often. And, importantly, when they change prices, they're going to change them A LOT. I won't print new menus to add a penny or two to the roast beef sandwich. But, I might if I need to add a couple dollars to it.
So, the question "How Frequent are Small Price Changes?" relates to the question "How important are menu costs?" If small prices changes are frequent, then menu costs must be small. If small price changes are infrequent, menu costs might be large.
The interesting thing: it wasn't until recently that we had reasonable data to answer this question. Typically, gathering price data is awful, unless it's for a good that is highly commoditized (like wheat). But, in commoditized markets, we already know that menu costs are minimal, and that competition is so fierce that there's a strong incentive for flexible pricing strategies. What about less commoditized markets? What about, for example, grocery stores?
And that's where the data comes from. Thanks to modern scanners, we have loads of data about prices of individual goods. And that data suggests that small price changes are frequent.
But not really, according to Eichenbaum & Co. There's a problem with the data. Here's what it is: it's aggregated at the good level. So, we don't observe actual prices. We just see the number of the good sold and the volume of sales measured in dollars. Prices are computed from that. Why is this an issue? Because grocery stores often have multiple prices for a good.
Think, for example, of ice cream. Say that there are two prices: $2 for a loyalty card holder and $4 for someone without a loyalty card. If, say, 99 loyalty card holders buy ice cream and 1 non-loyalty card holder does, we'd observe: $203 in sales, and 100 units sold. Per unit price = $2.03. But, if the next day, 98 loyalty card holders and 2 non-loyalty card holders buy ice cream we'd observe: $204 in sales and 100 units sold. Price per unit = $2.04. So, we have a small price change. Except we don't, in the menu cost sense. No price tag actually changed, just the composition of buyers changed. So, menu costs might be large (leading to stable price tags), even though the average price changed.
So, Eichembaum & Friends got their hand on a dataset that had individual price tags listed in it. Using that, they find that small price changes are extremely rare. So, it might be that menu costs are large.
My take:
Personally, I'm an agnostic on the importance of sticky prices. To some degree, I like the idea that menu costs matter - mostly because I think it's typically true - but I think people follow Keynes in taking it too far. It's one thing to say that it's hard for prices to change for some reason or another. It's another to say that, because of this, the economy can be thrown into a Depression. Think of it: if given the choice between throwing an economy into a depression and changing your price on a good, what do you think most people would do? I suspect very few would choose the depression.
Now, we can bring out some game-theoretic arguments (no individual person believes they impact the odds of depression - though we all do collectively), but I find these arguments unconvincing. If my employer said to me "I can lay you off, or I can give you a sharp-ish pay cut.", it's not obvious that the pay cut is the *wrong* choice. Nor is it obvious that the employer should refuse to offer the pay cut. There are additional assumptions needed to get to that point - and I'm not confident that those assumptions are necessarily true. I think we should reach the same conclusion that Joe Salerno did several years ago: price stickiness is an entrepreneurial decision. Sometimes it makes sense to have sticky prices. Sometimes, it doesn't. So, while price stickiness is interesting as a factoid, I wouldn't give it the same business-cycle significance many macroeconomists do. I think they do, however, have important Cantillon-effect/distributive implications. (If prices are sticky in your market, monetary expansions will hurt you. If prices are flexible, they'll probably help you.)
(There are lots of arguments I could appeal to regarding the unimportance of price-stickiness in business cycles. I am particularly fond of Golosov & Lucas's argument that even if prices are generally sticky, the fact that the prices that change change a lot is enough to offset the majority of macroeconomic effects. So, we need something other than price-stickiness to explain business cycles.)