Sunday, May 8, 2011

Sticky prices and the Keynesian Narrative

In sticky price Old Keynesian or New Keynesian economics, there are two key ideas. First, it is taken as given that some prices are more sticky than others. There is now a large body of empirical work that characterizes the size and frequency of price changes across a wide variety of goods and services. For example, we now know that the price of gasoline changes about once every three weeks, while prices for restaurant items change once every 11 months. Second, in Keynesian models with sticky prices, the quantity of output produced by a firm is demand-determined when the firm's price is stuck.

Suppose then that we accept the standard Keynesian narrative about our current predicament as truth. Correct me if I'm wrong, but the standard narrative is that it is irrelevant how we got into our current state. The relevant feature of the current state is that there is deficient aggregate demand. Deficient demand spreads itself across sectors of the economy and, according to the narrative, can be remedied with stimulative fiscal and monetary policy.

The standard Keynesian narrative, coupled with the sticky price mechanism, is very useful, as it seems that it could generate restrictions on what we should see in the data, with regard to price and quantity variation and dynamics across sectors of the economy. As far as I know no one explores these things, but I could be wrong. It would be helpful if readers who know this literature could direct us to the relevant research.

What am I thinking about here? First, in a Keynesian world that experiences a fall in aggregate demand, relative prices become misaligned. The relative prices of sticky-price goods rise and the relative prices of flexible-price goods fall. Second, employment will fall in the sticky-price sector relative to the flexible-price sector.

Now, my empirical work here is going to be very crude. This is just a blog, after all. Suppose we look only at the 8 primary components of the consumer price index. These are:

1. Food and beverages (weight 14.8%) This includes two very different items, food at home and food away from home (about half-and-half). The purchase of food away from home, is primarily the purchase of a service.
2. Housing (41.5%) This includes not only the cost of shelter, but energy, furniture, and appliances as well.
3. Apparel (3.6%)
4. Transportation (17.3%) This includes gasoline, cars, and public transportation.
5. Medical care (6.6%)
6. Recreation (6.3%) Includes consumer electronics.
7. Education and Communication (6.4%) Computers are included in communication.
8. Other (3.5%) Includes tobacco and miscellaneous services.

Now, we'll take January 2005 as our base period, and look at the time series for the 8 components, displayed in the first chart. Now according to Mike Bryan and Brent Meyer (Mike took his sticky-price index ideas to the Atlanta Fed when he moved), of these 8 items the flexible ones are: food and beverages (food-away-from-home is sticky, but let's put it in this category anyway), transportation (dominated by energy prices), and apparel (they put infants and toddlers in the sticky category, but it just makes it). Bryan and Meyer use Bils and Klenow to categorize prices as flexible or sticky, but Bils and Klenow do not address housing. Since more than half of the housing component is somehow tied to the price of housing, we'll say that housing prices are flexible, since I don't think anyone wants to argue that house prices are sticky. Otherwise, the prices of medical care, education, recreation, and "other" are sticky.

Now, during the recent recession, we certainly see marked changes in the trajectories of the flexible prices. There is a decline of more than 25% in the transportation component, and food prices (which had been increasing) level out, as does the housing component. This all seems more or less consistent with the Keynesian narrative. However, the price of apparel seems to be behaving like a sticky price which is impervious to the recession. Indeed the prices that are supposed to be sticky - medical services, education, recreation - are also impervious. Surely, in the face of this persistent deficient aggregate demand, the sticky prices should ultimately be adjusting downward.

Now, even more puzzling is what happens from early 2009 to the end of the sample. The decline in transportation prices reverses itself (energy prices up again) and food prices pick up again. According to the Keynesian paradigm, this now looks like an increase in aggregate demand, reflected in flexible price increases, and decreases in the relative prices of the sticky-price goods. Of course, this is not the standard Keynesian narrative we hear at all.

Finally, look at where we are in March of 2011 relative to January 2005. If the dominant force over this period was the period of persistent insufficient aggregate demand that began in late 2007, what we should see at the end of 2011 is lower relative prices of flexible price goods and higher relative prices of sticky price goods. The relative price increases were for transportation (flexible), "other" (sticky), medical services (sticky), food (flexible), education (sticky), and housing (flexible). Relative prices declined for recreation (sticky) and apparel (flexible). Thus, the medium-term relative price movements seem to have nothing to do with stickiness/flexibility as it is spelled out in the Keynesian narrative.

The second chart shows some components of establishment employment, chosen to correspond (pretty rough, I know, but the best I could do) to the CPI components. These are medical and education services (sticky), leisure services (sticky), energy (petroleum and coal products - flexible), food (manufacturing - flexible), construction (flexible), durables (flexible), and apparel (flexible). In the chart, the big employment declines are in fact in the flexible price sector. Sectors that are doing relatively well - medical and education services, leisure, energy, and food - are a mix of flexible/sticky price goods and services.

Conclusion: I can't square what I'm seeing in the data with the standard narrative. Can anyone help me out?

15 comments:

  1. I think that data is impossible to explain, at least using the baseline New Keynesian sticky price model. Maybe there are more elaborate New Keynesian models I'm unfamiliar with that could rationalize this, but I think they would have to move beyond a simple sticky price distortion.

    Personally, I've never thought sticky prices were a plausible explanation for this recession (or really, for most recessions). Intuitively, I think sticky price distortions probably matter in the very short-run (e.g. a few months). But we are now running into the third year of this recession. Intuition should tell you there is something else going on besides price rigidity. The first year into this thing, New Keynesians could plausibly cite nominal wage rigidity. I know Krugman has cited downward nominal wage rigidity in the past, but, again, this is the third year. Eventually somebody is going to be willing to take a wage cut.

    I also think it's worth mentioning that Head et al recently devised a search theory model where you can get nominal price stickiness, while still maintaining monetary neutrality. So the simple observation of price stickiness is not enough to conclude the New Keynesian model and its policy implications are valid. In the Head model when the money supply increases, some firms keep prices fixed, but consumption of the good produced by that firm rises, while others move their prices to maintain their profit. Maybe there is some sector-specific characteristics that make different sectors differ in whether they are going to keep their prices fixed, but have an increase in volume consumption, or let there prices rise or fall.

    Here's Head et al's paper is anyone is interested:

    http://www.bu.edu/econ/files/2011/04/HLMW-oct-2010.pdf

    ReplyDelete
  2. Good question. Here's one possible way for us sticky price people to weasel our way out:

    Think back to your good post one month ago on core inflation.

    The theoretically correct way to define price stickiness would be something like the variance of P relative to P*, where P* is the flexible price equilibrium. But if we measure stickiness as the variance in P, we will get a biased measure.

    Those goods which have a stable demand and supply will have a stable P*, and will look like sticky price goods. Those goods whose demand is very cyclical will have a high variance of P*, and will look like flexible price goods.

    For example, under the gold standard monetary policy will keep P* for gold fixed, so gold will look like a very sticky price good.

    I have always thought of (used) houses as having sticky prices. When demand falls (rises) it seems to take sellers a long time to cut (raise) their asking prices. So first we see inventories of unsold houses rise (fall), and only later do prices start to move. Presumably there's some sort of search/matching model, but it seems there's a lag before the U/V ratio affects P. I don't know why there's a lag, but it might be some sort of Lucasian signal processing imperfect information story. "There's no offers at my asking price; does that mean my asking price is too high, or might it be that I've just been unlucky and the potential buyer who really likes my particular house hasn't come along yet?"

    BTW: I just did a post in response to your post on core: http://worthwhile.typepad.com/worthwhile_canadian_initi/2011/05/is-core-inflation-an-artefact.html

    ReplyDelete
  3. Back in the 1980s there was some work done on the relation between inflation and relative price dispersion, eg by Barro and also by
    Cukierman
    http://www.jstor.org/pss/1807379

    Might be interesting to revisit in the context of modern New Keynesian models

    ReplyDelete
  4. "Sticky wages" is a Marshall idea from the 19th century.

    Economists of all kinds recognized the phenomena without any help from Keynes ...

    ReplyDelete
  5. Steve,

    I think the behavior of the transportation & food components makes perfect sense if you think of global AD. As you mention, most of the rise in the former is due to oil prices, and I suspect a lot of the latter is due to similar behavior of ag commodities. Both are more sensitive at the moment to relatively strong emerging-mkt growth.

    Also, maybe I missed it but when you talk about relative price changes, what are you measuring them relative to? Each other? Overall CPI?

    Nick, good point about used houses. In addition to Lucasian effects, I suspect that there's also a fair bit of "those fools don't know what a great buy this house is at $x!" I recall seeing evidence for that at some point, but don't remember any references. BTW, have you been to Adelaide since we were there?

    PS

    ReplyDelete
  6. Well this all pretty irrelevant anyway. We can finally put those foolish over-aggregated sticky price models in the grave where they belong.
    As you may know, a couple of MIT economists finally solved the problem of modeling the business cycle in a paper last week entitled "Decentralization and the origins of fluctuations". Its got something for everyone - Hayek for Greg Ransom and coordination failures for the Keynesians

    ReplyDelete
  7. psummers: Wow! I had seen your comments around the blogosphere, but never put 2+2 together. I haven't been back to Adelaide. Too far away. But a lovely city. All the best!

    ReplyDelete
  8. Yes, I heard about the Angeletos paper:

    http://econ-www.mit.edu/files/4042

    No opinion on that one yet.

    Nick,

    It sounds like you have something else in mind than what people have thought about in the New Keynesian literature. The housing market certainly involves friction. People with heterogeneous preferences are buying and selling heterogeneous objects under private information. But I would not want to think about that market the way a New Keynesian does, as involving some sellers who are constrained to satisfy demand at a price they cannot change.

    Pete,

    Yes, exactly, if you look at those time series you can think of all kinds of reasons for what you see that have nothing to do with stickiness or the lack of it. On the relative prices, yes, I was thinking relative to the CPI. I thought you could more or less figure that out from what you see in the picture, but I should have shown you a chart.

    ReplyDelete
  9. Steve: I cherish the fond hope that someday, when we are all very wise, we will all look back and see that sticky price models and flexible price models with enough trading frictions are at root the same thing. They just operate at two different levels of abstraction.

    ReplyDelete
  10. Though right now, I have no good reason to think that these are the "same thing." I could as easily think that the frictions the sticky price people seem to have in mind (though it's not clear what they have in mind, as the frictions are not explicit) have very different implications from what you see in a search paradigm, or in mechanism design, for example.

    ReplyDelete
  11. Steve:

    What about the sticky-information models? Do you see any merit in them?

    ReplyDelete
  12. A few observations:

    1. The housing data can't be trusted, as it doesn't measure current market prices. Does anyone seriously believe the relative price of housing rose in recent years in the US?

    2. The increase in the relative price of food and oil is driven by international factors.

    3. I also have some trouble with the new Keynesian focus on price stickiness, but you might not like my answer any better. I believe wages are fairly sticky in most industries (flexible price or not.) During a recession output tends to fall in those industries where demand is cyclical (construction, cars, etc) and not so much in industries where demand is not very cyclical (health care, etc) The degree of price flexibility in each industry might have a slight impact, but not enough to offset the more powerful influence of relative cyclicality of demand for various products. Thus output declines are not particularly closely correlated with the degree of price flexibility in each industry.

    The sticky wage theory is often criticized because real wages aren't particularly countercyclical, but that's probably not the right way to test the theory. The better test is to look at how wages respond to NGDP shocks.

    ReplyDelete
  13. "Stickiness" isn't the issue that jumps to mind when I look at the sector employment graph. Rather, it looks like those sectors with low elasticity of demand do well and those sectors that are more discretionary do poorly. That would be in keeping with a general depression of aggregate demand.

    ReplyDelete
  14. Yep…..do follow link is the way to get quality backlinks, you have done superb job.
    Thanx for this useful posts. Glass Offices

    ReplyDelete
  15. Looking at the number of people joining social networking sites these days, I feel it is the best platform to present or share anything with the world. But the problem is I don't really know how to get started. I want to advertise my company and I don't understand which platform to use and how to place my ad as these sites have many guidelines and norms to be followed or you might be blacklisted. It would be a great pleasure if you can help me out with it.
    credit card help
    chase credit card
    best buy credit card
    walmart credit card
    hsbc credit card
    citibank credit card
    capital one credit card
    reward zone mastercard

    ReplyDelete