Wednesday, May 11, 2011

More on Sticky Prices

This is an update on this previous post. There are a few papers in the literature that explore the consequences, in sticky-price Keynesian models, of having more price flexibility in some sectors of the economy relative to others. Here are a couple of early papers by Lee Ohanian and coauthors, in the JMCB and the Richmond Fed EQ. Also see a more recent paper by Barsky, House, and Kimball, "Sticky Price Models and Durable Goods." The upshot of this work is that it is important to think in general equilibrium terms. Sometimes prices that are actually flexible can behave like sticky prices, and sometimes there are counter-intuitive effects of monetary policy shocks, or effects which appear inconsistent with evidence. I'll leave it to others to sort out whether this literature has anything to say about current observations.

A recent paper that has some bearing on these issues is this one, by Head, Liu, Menzio,and Wright. Basically, this takes a Burdett/Judd search model of price dispersion, and uses it to think about the implications of monetary intervention for the distribution of prices across sellers. The idea is that, when the equilibrium object is a distribution of prices, then shocks that change this distribution do not require that all sellers change their prices. Indeed, we could observe prices changing very infrequently in a world where all prices are flexible and money is neutral in the short run. The first time I saw this paper, I thought of it as just a clever example, but now I think the idea is quite deep.

While it is useful to have the characterization of the price data that we get from work by Bils/Klenow, etc., as macroeconomists we should not be too focused on the behavior of individual prices. We observe price dispersion because of search and information frictions, and this can help explain observable price behavior, and yet have no implications for the short run effects of monetary policy. Further, most sellers are sellers of multiple goods, and the prices of those goods are interdependent. For example, the Target store prices particular items to bring shoppers into the store, and then uses physical placement of goods in the store and the prices of those goods to extract as much revenue from the buyer as possible. Online shopping is obviously different, as physical location and travel costs are irrelevant but, for example, Amazon uses a set of techniques to bring people to its site, and then uses information and prices to direct you to things they hope you will buy. None of this looks much like incurring a menu cost for each individual price posted, and then meeting whatever demand arises at that price.


  1. so the gist of the paper is that sticky prices can arise without menu costs. is that the right interpretation? and why is this an important result in practice, rather than "just" in theory?

  2. It's hard to say exactly what people mean when they say "menu costs." This last model is one where there are some frictions - search/information costs - that imply that sellers with identical production technologies can be selling goods at different prices. Basically firms selling at high prices are getting some rents from people who are "desperate" in some sense, and the low-price firms are making it up in volume. The Fed comes along and increases the money supply. This implies that the distribution of prices has to change, but that does not require that all firms change their prices. That seems interesting to me. I think about all the frictions there are in the world that make it costly for consumers to find particular goods and services at low prices, and the model suggests to me that this type of phenomenon could be quite general in practice, leading me to think that, even if see prices changing infrequently, that does not tell me anything about the effects of monetary policy. Thus, if I want to think about why, if at all, the Fed can affect things like real GDP and the unemployment rate, I might be wasting my time thinking about the pricing decisions of firms.

  3. thanks, steve. your last sentence seems to me to be the important one.

  4. I think it's also important to point out that in the Wright et al. paper, money is neutral: the nominal price distribution changes in response to a monetary expansion but the real price distribution is constant (even though some firms don't change their prices!). At least that's how it worked the last time I saw Randy present the paper.

  5. anonymous,

    Yes, I keep trying to give these sticky prices a chance, but that is where I seem to end up.


    Yes, exactly.

  6. I think with the internet and other tech advances prices are less sticky. But what about sticky wages?

    There are certainly huge personal risk and security reasons, human psychology, and negotiations costs to have wages stuck (at least downwardly) for substantial periods. And then of course there's well documented downward nominal wage resistance.

  7. I think trying to come up with a realistic description of how businesses set their prices is largely irrelevant to macroeconomics.

    The relevant question is: does assuming prices are sticky lead to a model which gives more accurate predictions about business cycles, the effects of monetary policy, etc.?

    If the answer is yes, then the question becomes how to model the phenomenon of sticky prices in a way that's consistent with rational, optimizing agents. Menu costs are obviously a leading candidate. The realism of this type of model is irrelevant. All that matters is whether assuming menu costs (or whatever) at the micro level leads to a more accurate description of how the aggregate economy behaves.

    If sticky prices do not lead to more accurate predictions (if, say, the evidence shows that money is neutral) then we should discard them from the model. The fact that some firms may change prices infrequently (whether due to menu costs or anything else) is beside the point.

  8. The first post is missing the point. If you assume menu costs, but the real mechanism is search, then your policy intervention measurements will be wrong. Prediction is of course important, but it cannot be the end of the story. The Lucas critique applies and matters.

  9. Not sure what you are getting at. Of course the Lucas critique matters. I'm not sure what you think "the point" is. The point of what?