Monday, November 1, 2010

Sticky Prices and Keynesian Economics

Since Keynes wrote his General Theory, other economists have tried, in various ways, to formalize what Keynes appeared to have had in mind. Hicks constructed the IS-LM model, which is a static framework in which prices are fixed in nominal terms. In the 1980s, Mankiw and Blanchard and Kiyotaki, among others, thought about Keynesian menu-cost models in which it is costly to change prices. John Bryant,Peter Diamond, Cooper and John, and the authors represented in this JET volume studied equilibrium coordination failure models with Keynesian features. Mike Woodford, among others, modified the neoclassical growth model, using some of the insights from the menu cost literature, to develop sticky-price New Keynesian models.

The followers of Keynes are sometimes motivated strongly by belief. You can see this, for example, in Ball and Mankiw's Sticky Price Manifesto. Some Keynesians look at a contractionary episode - the Great Depression or the recent recession - and cannot see anything other than Keynesian economics to explain it. The "fundamentals" in 1933 looked about the same is in 1929, so how could real GDP have been 40% lower, except as the result of a "deficient demand" phenomenon? There is also empirical evidence, for example on the stickiness in prices. Prices of goods are certainly much smoother than the prices of assets, and this forthcoming chapter from the new Handbook of Monetary Economics summarizes what is known about the empirical regularities in the changes in prices over time and across goods. Another interesting regularity is that much of the variability in the real exchange rate between two countries, even in close proximity (e.g. the U.S. and Canada), is explained by variability in the nominal exchange rate. Thus, it seems to matter how prices are denominated - e.g. in U.S. or Canadian dollars.

What do we want from an economic model? First, we want it to explain something to us. Why are people unemployed? Why do economies grow? Why is there inflation? Second, we would like to ask the model questions. In macroeconomics, some of the key questions are policy questions. How, if at all, can fiscal and monetary policy be used to make us better off? The answers to the first set of questions can have a lot to do with the answers to the second set. Once I understand what causes inflation and why it is bad for us, this can tell me how monetary policy should be conducted to control it.

Now, the class of coordination failure models I mentioned above explains something. These models tell us that aggregate fluctuations can arise due to people coordinating on extraneous information. Further, these fluctuations are generally inefficient. What should we do about this problem? Typically, fiscal and monetary policy rules in these models can be set in ways that eliminate undesirable equilibria. Coordination failure models are certainly Keynesian. Indeed, one can find passages in the General Theory that are essentially coordination failure stories. However, in spite of the large coordination failure literature, and some success in fitting these models to the data, most Keynesians are currently uninterested in coordination failures. This may be because of the technical demands required to work with these models. However, full-blown New Keynesian models appear no less demanding. To me, it's a puzzle.

The New Keynesian models in wide use now typically rely on Calvo pricing (a form of time-dependent pricing), whereby monopolistically-competitive firms receive random opportunities to change prices. Upon receiving an opportunity to change its price, a firm does this optimally (taking into account that it may not be able to change it again for a while), but is otherwise stuck with the price it charged last period. Some New Keynesian models go deeper and use state-dependent pricing, under which there is a menu cost, and a firm changes its price when it is optimal to do so. Time-dependent pricing and state-dependent pricing are just special cases of the same model. With time-dependent pricing the menu cost is random - it is either zero or infinite. With state-dependent pricing the menu cost is constant.

Neither state-dependent or time-dependent pricing actually explains why prices are sticky. Why should it be costly to change a price? This can't literally be the cost of posting a price schedule - surely that is trivial relative to other costs of producing and selling a good or service. Keynesians must have something else in mind, if they have thought about it. But what is it? And surely what is causing the price stickiness has to matter for whether and how policy should fix any inefficiencies arising from sticky prices.

The typical argument seems to be that actually explaining the sources of price stickiness is too hard - otherwise someone would have solved the problem. As the argument goes, a useful shortcut is to just fix the prices, perhaps in a Calvo fashion, and go from there. With the world falling apart and large numbers of people unemployed, why should we have to wait while the theorists work out the details? Of course, the problem could be that, once we know all the details, we might change our minds about how the policies work.

What would a theory of price stickiness, or more appropriately a general theory of pricing, look like? First, we have to have a model in which sellers want to quote prices in units of "money." While we have models of media of exchange, models of the unit of account do not exist, to my knowledge. In any monetary model I know about, the numeraire could be anything. It is irrelevant whether we quote prices in units of money, goats, or soccer teams. Clearly though, how we write contracts and quote prices has a lot to do with what we use as a medium of exchange. We want an environment that will imply that contracts with few contingencies are optimal, and possibly the nominal contracts can then be derived as optimal risk-sharing arrangements. Clearly, though, you have to be a long way from complete Arrow-Debreu contracts. The frictions will be critical.

Now, once we are past the nominal-contract hurdle (a very serious one at that), we have to worry about the stickiness. Suppose that I am selling toothbrushes at a particular location. There are many ways to determine how many toothbrushes I sell at what price. I could haggle with each customer who comes in the door concerning how many toothbrushes I will exchange for how many dollars. I could specify a time each day when I will auction off so many toothbrushes, possibly selling them individually using an English or Dutch auction. I could also post a price and sell toothbrushes at that price to whoever comes in the door. Of course we all know that it is not efficient to haggle over a toothbrush, or to sell toothbrushes in an auction (unless it's Elvis's toothbrush). However, haggling may be efficient when I am buying a house, and an auction may be efficient when I want to sell an idiosyncratic object. Surely, part of what the theory needs to tell me are the characteristics of a good or service that determines how it is exchanged - using price posting, an auction, one-on-one haggling, or some other mechanism.

Now, in the case where prices are posted, why would firms want keep a posted price unchanged for months at a time. It could be that it just does not matter much. Maybe the firm has other more important things to think about (inventory control, shoplifting, shirking employees) than the price of toothbrushes. Maybe the firm is trying to minimize effort for repeat customers. If I know that firms (such as the grocery store, where I buy the same goods with some regularity) are changing prices frequently, as a buyer I will have to reoptimize more frequently to make sure that I am purchasing the optimal basket of goods. Maybe I would rather buy at a store where the prices change infrequently.

Now, what could go wrong here? How are firms going to get the prices wrong, so that the government needs to step in to fix things? Possibly with infrequent price setting, we get phenomena like those in Calvo-pricing setups, where firms make staggered price adjustments, and relative prices are distorted. This is far from clear, but suppose that relative price distortions are the difficulty. Now, if the government has all the information, this problem is easy to solve. The government knows what the correct prices are, and it forces firms to price correctly. Of course that seems silly, as it is unrealistic to think that the government has access to enough information to accomplish this. We all know that market economies solve very complex allocation and pricing problems that it would be impossible for a central planner to solve.

If the relative prices are distorted, the government must necessarily rely on some indirect mechanism for dealing with this. If the prices are distorted due to inflexibility, why not subsidize flexibility? But how would this be done? Do some sectors of the economy need larger subsidies than others? Could firms game the system by making a price change today and reversing it tomorrow? How does the government monitor all these price changes? The more I think about this, the more unworkable it seems.

Now, the solutions that Keynesians, new and old, have come up with for solving the relative price distortion problem seem odd. First, Old Keynesians tended to (and tend to) focus on fiscal policy remedies. Why should the provision of public goods and services be an important element in correcting a relative price distortion? I know how typical Keynesian models work, but when I think more deeply about the pricing problem, this does not make sense. Second, New Keynesians argue for the use of monetary policy (within the constraint determined by the zero lower bound on the nominal interest rate) to correct the problem, when it seems that indexation would be feasible, and in fact optimal for the firms involved.

We need a serious theory that explains the behavior of the prices of goods and services. Such a theory is necessary, as it is far from clear that the observed behavior of prices implies some market failure and, even if it did, it is also not clear that standard Keynesian prescriptions would solve the problem. In the absence of the theory, I think there are good reasons to be skeptical about what Old and New Keynesians have delivered thus far.


  1. Professor,

    First, you will have to handicap my question with the knowledge that I have only read through chapter 5 of Mas-Colell. Second, if we are thinking about sectors in which profits are zero, how do downward shifts in prices even come about? Finally, would it be feasible to think of sticky prices as a product of the natural time delay from producers of raw goods through producers of intermediate goods and then through producers of finished goods? But, I suppose then all prices would shift immediately once one subset changes them as they affect every other producer. Your thoughts?

  2. In their book about the history of modern macro, Snowdon and Vane include a third branch of modern keynesians in the GReenwald - Stiglitz models with credit rationing which rely on wage stickiness but not on price stickininess

    What's your view about them?

  3. I think I can anticipate your reaction to this, but I believe that you put too much weight on finding microfoundations for price stickiness.

    Akerlof and Yellen just make the reasonable assumption of quasi-rationality in price setting.

    For me, it is quite obvious that firms do not react continously (like in continous time I mean) to their environment, but take a bit to process information, understand what is really happening, etc... This is not very different from a slightly modified Lucas cycle model with incomplete information. Or you can understand that, as in the Greenwald-Stiglitz models, thinking of firms as moderatly risk averse. As I see it, what NKE shows is that if everyone does more or less the same thing, you can get blocked on the "wrong" price vector which is actually an equilibrium due to imperfect competition where profits are interdependent.

  4. Hamed,

    I don't think that profits or intermediate goods matter much here.


    1. I wrote about credit rationing back in the day:

    Stiglitz thought that credit rationing was important for how monetary policy worked, but he did not work out the general equilibrium implications. I did some of that here:

    The credit market friction and the rationing matter, but there is nothing particularly Keynesian about it. However, in general I think that sticky prices may be a dead end. Particularly in light of recent experience, it's much more productive to think about credit market frictions.

    2. There were a series of papers where people argued that the losses from suboptimal behavior are second-order in various contexts. Mankiw was one, and Akerlof, and Tony Smith I think. I'm not sure what to think of this. Suboptimality is suboptimality.

  5. In many fields, when a complete understanding of the condition treated and the mechanisms by which a prescription works cannot be gained, the prescription might nevertheless be accepted if it is shown to work. People would die if the mechanisms of all drugs and their diseases had to be fully understood before they were used, for example. Perhaps that is what is at play here.

    Now, you could argue that Keynesian solutions have not been shown to work. But you haven't argued that, only that the mechanisms they are intended to address aren't fully understood. You then leave as a totally unexamined assumption that it follows that that they don't work.

  6. What about the "Noisy Business Cycle" model of Angeletos and L'Ao ? Don't they generate cycles in a model with perfectly flexible prices? They make a big deal about heterogeneity of information about sector productivity shocks as the root cause of fluctuations. This "lack of common knowledge" channel is a separate channel from the uncertainty or news models of Beaudry, Jaimovich and Rebelo. Maybe we have more than enough to explain cycles without having to bother about sticky prices, which may just be an endogenous response to the fundamental factors driving cycles.


  7. Second to last anonymous:

    1. Medical science is certainly known to go wrong, perhaps more often than not. A drug "works" in the sense that it cures a particular condition, but it has a long-term side effect that actually makes you worse off. Sometimes doctors seem like mechanics. You try a bunch of things until the patient can get up and walk out of the hospital, and then you claim success. In general, even in medicine, we would like to know why a particular treatment appears to be working. Without knowing the underlying mechanism, you can't correctly weigh all the costs and benefits.

    2. "You then leave as a totally unexamined assumption that it follows that that they don't work."

    Yes, I'm saying in part that we really don't know much about it. Of course, even if you think these things "work," there are still plenty of qualifications, principally relating to policy lags of various sorts and quantitative uncertainty.


    Yes, I agree that there is plenty to think about. Why should we be primarily interested in Keynesian frictions? Keynesians seem to fear that if you abandon sticky prices and wages that there is no longer a role for government, which is false.

  8. To use Friedman’s parable: Does a pool player need to understand geometry to play pool?

    Perhaps not. But some rough rules of thumb may be useful.

    Now a deep understanding of the goemetry of pool are undoubdedly useful to play that game, but in the meantime, seems to me that it make sense to play doing the best we can without deep understanding of geometry.

    Now central bankers don't even have the option of deciding wether or not to play monetary policy. They have to play, thats their job, no matter what.

    So they have to come up with some workable hypothesis while economists figure out the best "deep" microfoundation for monetary policy.

    In the meantime, New Keynesianism is the only game in town. As Tom Sargent is known to says: It takes a model to beat a model.

    No workable alternative to the New Keynesian DSGE models is yet in sight. That's the sad truth of our profession. All the alternative so far are toy models where its typically optimal to set interest rate at zero (so I suppose Ben Bernanke just reached the bliss?)

  9. Isn't wage stickiness a lot more important than price stickiness? And aren't there more obvious reasons why wages should be sticky?

  10. Second to last anonymous:

    The New Keynesian model has been a pretty dramatic failure in terms of what it tells us about how to handle a financial crisis. How can this be the only game in town when it's not in the game at all? Randy Wright and I make the case that we're within reach of having good quanitative non-Keynesian alternatives. And they're not just about Friedman rules.

    Last anonymous:

    Why do you think wage stickiness is more important?

  11. "Now, the solutions that Keynesians, new and old, have come up with for solving the relative price distortion [...] Why should the provision of public goods and services be an important element in correcting a relative price distortion? "

    Isn't the keynesian problem about fixing allocations rather than correcting price distortions? Say, move closer to the natural unemployment rate... equilibrium allocations are what matter, right?

  12. Stephen,

    The problem with economics is lack of long-range data. Maybe in another 200 years and after another 4-5 great recessions we will have enough data to make solid conclusions about the macro economy (what frictions, shocks, market mechanism, etc. matter)

    In the mean time we are stuck with as many theories as we have variables; an environment where ideology dictates theory. NK economists like the government, so they focus on variables (e.g. sticky prices) / theories that create a major role for the government.

    Agree, information frictions and credit rationing are very important features of the macro economy. Not sure this is enough, and rational inattention probably should be added or at least Hansen-Sargent model averaging.


  13. Steven says
    "The New Keynesian model has been a pretty dramatic failure in terms of what it tells us about how to handle a financial crisis. How can this be the only game in town when it's not in the game at all?"

    Really? Policy all over the world has been about preventing "deflationary spirals", ranging from Obama speaches to central bank governors. This pops right out of New Keynesian models (see e.g. Krugman (1998) or Eggertsson and Woodford (2003) or Auerbach and Obstfeld (2006)).

    Moreover all central banks have been sending signals about keeping nominal interest rates low for "some time" or "considerable period" or even giving special calander dates as Bank of Canada.

    This straight out of the New Keynesian playbook. All about "managing expectations" in Woodford's terminology. What do the "New Monetarist" have to say. Nothing! In fact, zero interest rates are optimal!

    It is ridiculous to claim that the New Keynesian litarature has nothing to say about the current crisis.

    In fact, one of the main branches of the New Keynesian literature has been about how to conduct optimal policy at zero interst rate, dating back at least to Rotemberg and Woodford (1997).

    Meanwhile, the "New Monetarist" have been touting why zero interest rate are optimal! That simply does not pass the stinktest.

  14. 1. New Keynesian models are currently predicting substantially negative nominal interest rates. I've seen numbers in the range of -6%. Smells a bit, don't you think?
    2. In a New Keynesian model, monetary policy is about moving a nominal interest rate around. These models have nothing to say about any of the policies that the Fed has been engaged in lately - private asset purchases, central bank lending, purchases of long maturity vs. short maturity assets.

    You're not up to date. Look at this paper:

    It's got liquidity traps, financial crisis phenomena, private asset purchases, interest-bearing reserves, retail transactions involving currency and bank deposits, non-neutral open market operations. The Friedman rule is not optimal. All kinds of great stuff!

    I can understand why people invested in New Keynesian models don't want to expend effort learning something new, but I would think you would be really embarrassed given the last two years' experience. Get with the program! Get your mind off stuck prices and think about banks, credit, private information, quantities on the central bank's balance sheet, currency, etc.

  15. "when it seems that indexation would be feasible, and in fact optimal for the firms involved."

    Good point. It seems that network effects are a significant barrier to a wider adoption of indexation. A good example is TIPS. Theoretically, investors should prefer to buy TIPS rather than regular treasury securities as they are better for risk management purposes. In practice it appears that Treasury could save serious money by reducing issuance of TIPS and increasing issuance of regular treasury securities, as TIPS investors are earning a very sizable liquidity risk premium. We have a problem as indexation is very expensive in the most liquid market with very sophisticated participants.

  16. "Why do you think wage stickiness is more important?"

    Because the New Keynesians advocate targeting core inflation which is largely driven by wages.

  17. Money demand blog:

    Yes, even in times of moderately high inflation, indexation was not as popular as you would think, which is a puzzle. How do you measure the liquidity risk premium on TIPS?


    This one I have never understood. There are plenty of reasons for focusing on the broadest measure of the price level as a target for monetary policy. Why should I ignore some prices that are deemed to be volatile, and how do I know that today's set of volatile prices will not be tomorrow's set of non-volatile prices? What do you mean "core inflation is largely driven by wages?" Wages are not exogenous.

  18. "How do you measure the liquidity risk premium on TIPS?"

    By measuring relative trading volume of TIPS and non-indexed treasuries, and by measuring the difference of inflation expectations embedded in TIPS vs. derivatives (inflation swaps).

  19. I think what you're saying is that if I suppose that the break-even rate implied by TIPS yields vs. nominal bond yields is a measure of the implied expected inflation rate, and I compare that to the measure I get from inflation swaps, there is a downward bias in the inflation expectations measure from the TIPS. Therefore, TIPS holders are requiring a higher yield to hold these assets for some other reason, and you're saying that is that TIPS are somehow less liquid than nominal Treasuries. Further, note that there is a zero lower bound on the indexation payoff you get on TIPS - if the CPI goes down, the contingent payment is zero. This would bias the inflation expectations measure from TIPS upward, which goes in the other direction. But why should TIPS be any less liquid than nominal Treasuries? I may observe that TIPS are traded less, but why would one be less easy to sell than the other?

  20. Yes, zero lower bound on indexation payoff increases the estimate of inflation risk premium.

    If TIPS are traded less, large uninformed sale will move the market price more, this requires compensation in the form of liquidity risk premium. But the size of liquidity risk premium is very high, so some kind of market inefficiency might be involved.