Friday, August 27, 2010

New Keynesian Models

Mark Thoma tells us here why he thinks "economists rejected flexible price models." Of course, the models were never "rejected," but New Keynsians have been very successful with central bankers, and in shaping the research program in macroeconomics for the last 10 years or more. This really has nothing to do with empirical evidence. Smets-Wouters and Christiano-Eichenbaum-Evans medium-scale models fit the data, but there are so many bells and whistles and shocks in these models that this seems no more than data description. New Keynesian economics is succesful, because Mike Woodford has been a brilliant salesman. The approach is sold as a synthesis, meant to be inoffensive to the hard-core Prescott RBC people and to old-school Keynesians alike. Nobody's human capital depreciates with the adoption of the New Keynesian synthesis, and policymakers can keep doing what they have been doing, with Mike's seal of approval.

An expansion of these ideas, and a description of the alternative can be found here, in a piece I wrote with Randy Wright for the St. Louis Fed Review, and here, in our forthcoming Handbook of Monetary Economics chapter. Both are a little on the long side, but you can pick and choose.

15 comments:

  1. Interesting. But what are the empirical victories of the so called `new monetarists'. New Keynesian models have been successful because they allow economists and policymakers to interpret the data in useful ways and in terms of objects that we recognize. The `new monetarists' only response is that the new Keynesians aren't deep. Who knows what that mean? It is deeper to assume that agents are distributed across islands and there are no boats or radios? Build a model, estimate it / calibrate it and show that you do even vaguely as well as the new Keynesians. Otherwise get used to be ignored.

    ReplyDelete
  2. This is an ongoing research program. More to come.

    ReplyDelete
  3. Anonymous:

    You make a legitimate point, one that NMs will have to work on.

    However, you are being a bit unfair as well. It's not like their are legions of NMs out there able to undertake the program you suggest. And to ignore the paradigm on that account alone is likely to hinder the speed at which empircal testing proceeds.

    I should like to point out too that Woodford and his disciples have recently abandoned their cashless, financial friction free view of the world. They now explicitly model money, limited commitment, balance sheets, etc. In short, the NM is not being ignored (except to the extent that they are rarely cited).

    ReplyDelete
  4. I agree that there are more NK than NMs. But surely that is endogenous. The NKs are trying to non corporate financial market frictions into their model. But it's not clear that they need money per se to do that. In any event, their flexibility is a good thing. The data spoke - in the form of the recent recession - and they are responding. That's what good social scientists do. It would be nice if the NMs were less religious in addressing the evidence about nominal price and wage rigidities.

    ReplyDelete
  5. Which evidence do you mean? The Klenow et. al. regularities?

    ReplyDelete
  6. This topic could fill volumes. No point in trying to be comprehensive.

    I suspect you won't find aggrergate US time series evidence convincing. Or evidence on how wages and prices (for non-tradeable goods) don't respond to massive devaluations. So let me just cite a few recent micro studies. Nakamura and Steinsson's analysis on the CPI gods is very interesting: leave out sales and goods prices change roughly one every 8 months to 11 months. Also Midigran and Kehoe as well as Eichenbaum, Jaimovic, and Rebelo analyses were pretty convincing. See also Weiderholdt and Maciowak's recent industry level study. In addition the ECB had a huge study of micro data from each of the Western European countries. The picture that emerges is one of lots of `stickiness'.
    You may not like the NK explanation. That's fine. But it seems wrong to ignore the data and write down models where prices and wages are flexible. Heck even Randy Wright now takes this to be a `fact' worth theorizing about. I don't like his model but at least he's paying attention.

    ReplyDelete
  7. Anonymous:

    With respect to your comment "its not clear they need money to do that." The key friction in the NM framework is limited commitment. The focus is on "liquidity," and not "money" narrowly defined (as you seem to imply). Many assets appear to facilitate payments (like MBS used as collateral in repo). It is the NM view that these payment structures need to be undertood better (rather than ignored, or simply assumed in a reduced form way).

    With respect to your thoughts on the need to incorporate "price stickiness" into models, I refer you to this:
    http://andolfatto.blogspot.com/2010/07/sticky-price-hypothesis-critique.html

    ReplyDelete
  8. The observed pricing behavior is very interesting, and may well have something to do with how we think about the costs of inflation. The big leap though is from that to, say, the output effects from a monetary policy intervention. These models always rely on firms passively satisfying demand at a stuck price. Why do they do that?

    ReplyDelete
  9. Stephen Williamson said...
    > These models always rely on firms passively satisfying demand at a stuck price. Why do they do that?

    Not sure what you mean. They're maximizing profits, taking into account the fact that once they set the price it will be stuck there for a while.

    ReplyDelete
  10. Take the underlying model that Woodford works from in Interest and Prices. It's Dixit-Stiglitz monopolistic competition with Calvo pricing. I'm a firm in this environment. Either the Calvo fairy says I can change my price or I'm not allowed to. If I can't change my price, then, by assumption, I have to sell whatever output consumers demand at the predetermined price. That firm is not necessarily optimizing. (i) It's possible the price is so low it would prefer to shut down. (ii) Given market wages, it may prefer to sell less output than what consumers want to buy from it. I think if you actually let the firm optimize, given the Calvo pricing, you will get "funny" results - i.e. results that are inconsistent with the usual Keynesian narrative.

    ReplyDelete
  11. I'm looking at Ch. 3 of Woodford. Are you saying the profit function that Woodford derives -- where a firm's profits only depend on its price, prices in its industry, and the aggregate price -- is incorrect? Or maybe he ignores some corner solutions? I guess most of the analysis presumes profits are nonnegative, so there won't be shutdown (absence of fixed costs of course helps).

    ReplyDelete
  12. My book is in the office, and I think the explicit model is in Rotemberg and Woodford. Let's see if we can do it from scratch with an example. Suppose there is no capital, and a firm produces output according to y = an, where y is output, a is productivity, and n is the labor input. It's a monopolistically competitive firm, so with flexible prices it would always be earning positive profits. However, suppose that this is a firm that cannot change its price. Thus, let its relative price be p, which is given, let y* be output demanded at the price p, and let w be the competitively determined real wage. Now, because p is fixed from the firm's point of view, the firm is a price taker. It is possible that a lot of unexpected stuff has happened since the last time the firm changed its price, so it is possible that in fact ap < w, and the firm will shut down. if ap >= w, then the firm is happy to produce y*. Now, if you get a positive money surprise, output could actually go down, due to firms shutting down - not sure exactly what happens in general equilibrium.

    Now suppose decreasing returns, so you won't get the shut-down effect. Suppose each firm has a fixed amount of capital k, and the technology is now y = af(k,n). f has the usual properties. Now go through the same exercise. The firm faced with the fixed price p will produce min (y*,y**), where y** is optimal unconstrained output. That is, if the firm chose n optimally given p then its labor demand is n** determined by marginal product equals real wage, and we have y**=af(k,n**). Given a large money surprise, there will be a lot of firms who produce y** rather than y*, as the money surprise shifts the whole distribution of p's down. In the standard New Keynesian model, you don't allow for either the shutdown effect or firms optimally producing less than what consumers want to buy at the sticky price. The sticky prices definitely gum things up, and there is an inefficiency that monetary policy can correct. However, my conjecture is that a positive money surprise can make output go down - the Phillips curve is going to go the "wrong" way.

    ReplyDelete
  13. The point of Keynesian economics was always to support the progressive political agenda.

    The problem was that the political goals of perpetual debt and money creation conflicted with the known laws of "reality". Keynesian economics filled this void with technical sounding jargon that convinced people to ignore some of the more obvious aspects of reality, such as exponential numbers...

    ReplyDelete
  14. I'm not sure. Certainly Keynesianism comes with the idea that the government can and should intervene to fix things. There is an element of social insurance to it. However, narrowly defined it doesn't seem to say anything about the long-run quantity of resources allocated to the government.

    ReplyDelete
  15. It's silly to assume that firms can't change their price in a given period. It all depends on the size of the shocks hitting the firm, industry, and economy. Given the large negative shocks, the US and World economy have faced during the last few years (the end of the great moderation), I would think that downward price stickyness has if not vanished then become neglible (even when you exclude sales).

    Looking forward to the next Klenow et. al study covering 2007-

    KP

    ReplyDelete