Stephen Williamson weighed in recently on Keynesian Economics versus Regular Economics in his New Monetarist Economics blog. Responding to Paul Krugman’s post citing my criticism of Robert Barro’s column in the Wall Street Journal contrasting Keynesian economics and regular economics, Williamson took Krugman to task for not acknowledging that old-style Keynesian economics has actually been replaced by a newer version (New Keynesian economics) that actually tries to deduce standard Keynesian results from regular economics.
Although Williamson aimed his criticism at Krugman, not me, he did criticize this passage of Krugman’s blog in which I was mentioned
As Glasner says, there’s something deeply weird about asking “where’s the market failure?” in the face of massive unemployment, huge unused capacity, an economy producing less than it did and a half years ago despite population growth and advancing technology. Of course there’s some kind of market failure, which means that there’s nothing at all odd about asserting that better policy can yield free lunches.
Williamson comments:
We cannot observe a market failure. To deduce that a market failure exists, one needs a model. Given that we cannot observe market failure by looking at the state of the economy, we also can’t say what a “better policy” is. Again, for that we need a model.
Fair enough; we do need a model. But it is not clear what kind of model Williamson thinks is needed to infer the existence of a market failure. The standard model used in regular economics posits a process of price adjustment in which unsatisfied demanders bid up the price or frustrated suppliers drive it down until a price is established that clears the market, i.e., transactors can execute their offers to buy or sell as planned. In recessions and depressions, as opposed to “normal” periods, the standard model of price adjustment doesn’t seem to work. So there seems to be a prima facie case for saying that recessions and depressions involve some sort of market failure. The case may be rebuttable, but Williamson seems to acknowledge in responding to a comment 0n a subsequent post that at least one rationalization for cyclical unemployment is unacceptably implausible.
In the later post Williamson sends us to an essay published by the Richmond Fed in which Kartik B. Athreya and Renee Courtois argue that the first theorem of welfare economics teaches us that to justify intervention into the private market economy, it is necessary to establish that the array of existing markets is incomplete or not fully competitive. This is a remarkable assertion inasmuch as I am unaware that anyone ever asserted that the set of markets in existence is anywhere near the set required to satisfy the completeness conditions of the first theorem of welfare economics. (But maybe I am ill-informed). So what exactly must be shown to establish what is self-evidently true: that the necessary conditions specified by the first theorem of welfare economics do not obtain in the real world?
Here is how Athreya and Courtois characterize the practical implication of the first theorem of welfare economics for macroeconomic policy.
The preceding result implies that in a well-functioning market system [i.e., a perfectly competitive economy with a complete set of markets, a theoretical requirement with no real-world analogue], the adjustments made by individual and firms in response to a shock in fundamentals, even when they are contractionary, will be efficient. If something bad has happened to the ability of firms to produce output (say, a shock to the financial sector that hinders the allocation of labor and capital to their best uses), then each hour of labor becomes less productive. In this case, it make little sense for firms to continue producing, or for workers to work, at previous levels. After a storm, for example, a fisherman may find that each hour spent in the boat produces less fish; an efficient response is for him to spend less time fishing (hire less labor) and more hours consuming leisure (or, perhaps working on boat maintenance and repair), given that the cost of leisure or other activities, measure in terms of fish foregone, has fallen.
This example is meant to show that even if there is a shock causing a decline in output and employment, it does not follow that the adjustment – reduced output and employment — to the shock was inefficient. This is a basic proposition of real-business-cycle theory, regarded by some as quintessential regular economics. Negative productivity shocks imply reduced output and employment, so who is to say that observed output and employment reductions in business-cycle downturns are not characteristic of an optimal adjustment path?
Athreya and Courtois acknowledge that “frictions” can cause responses to a negative productivity shock to be inefficient, e.g., frictions in capital and labor markets, and in insurance markets. But in discussing inefficiencies in capital markets, they merely refer to banks’ difficulties in discovering good investment projects. Their discussion of labor-market inefficiencies is no more helpful, referring to problems of matching workers and employers, which can cause the search process to be long and drawn out. But just because matching workers and employers is costly and time-consuming doesn’t tell us how much search is optimal or whether the actual amount of search in response to a recessionary shock is optimal.
They do provide a somewhat more helpful discussion about the process of reducing labor input, suggesting that there is an inefficiency in laying off workers rather than evenly reducing hours across the entire work force. But this too is problematic, because, as they acknowledge, there are various reasons why letting some workers go and keeping the rest fully employed rather than cutting hours is less costly for employers than trying to reduce hours worked equally across their employees. The hardship is concentrated on a subset of workers rather than shared by all workers. But is there an inefficiency? We aren’t given a model from which to draw an inference.
Finally, there is the absence of a private market for unemployment insurance. In the absence of such a market, and because most of the reduction in labor hours is concentrated on a subset of workers, workers respond to the increased probability of losing their jobs by increasing precautionary saving. Athreya and Courtois admit that they cannot demonstrate any inefficiency in the increase in precautionary saving except insofar as it is related to frictions in labor and insurance markets, but they provide no proof of such inefficiencies, leaving the impression that the inefficiencies are not that significant. That impression is reinforced by the following remark.
The preceding suggests there may be a role for policy. However, it also suggests that productive policies will likely be those which directly address the specific frictions in capital, labor, and insurance markets that make the observed decline in aggregate consumption inefficiently large.
We are not told exactly what the specific inefficiencies are, but, presumably, if we can find them, then it’s fine to do something about them. But there is little if any justification for trying to increase aggregate demand through fiscal (or monetary?) policy.
[C]urrent spending-based stimulus, while it may marginally increase the probability of a laid-off worker regaining employment, will also likely draw employed workers away from other productive uses – further hindering efficient resource allocation.
Their most explicit recommendation is actually somewhat surprising (what would Robert Barro say?):
We think a more fruitful approach would be to leverage the existing unemployment insurance and social safety net infrastructure to better insure the unemployed while more strictly monitoring their job search efforts.
What amazes me about the real-business-cycle approach exemplified by this paper, apparently much to Williamson’s liking, is that it ignores any transmission mechanism amplifying a shock as its effects spread through the economy. Such transmission effects are completely suppressed in the representative-agent model, a characteristic real-business model seeking to deduce all the relevant properties of a business cycle from the analysis of a single agent supposedly representing everything that one needs to know about how an economy behaves. This is not just a retrogression from Keynesian economics, it is a retrogression from pre-Keynesian classical and neo-classical economics.
Despite Keynes’s misguided attack on Say’s Law, the notion that supply creates its own demand, an idea Keynes misconstrued to mean that classical economics held that there could be no lapses from full employment, or if there were, that they would be temporary and quickly rectified, Say’s Law actually explains precisely how shocks are transmitted and amplified as they spread through an economy. If supply creates its own demand, then a failure to supply entails a failure of demand. So a technology shock that causes some workers to lose their jobs, by preventing them and complementary factors of production from supplying their productive services, necessarily forces those workers, as well as owners of complementary factors unable to supply their services, to reduce their demands for products. That is precisely the idea of the Keynesian multiplier, except that Say’s Law views it from the supply side and the multiplier views it from the demand side. At bottom, these two supposedly contradictory ideas correspond to the same phenomenon, viewed from opposite angles. But the phenomenon is completely suppressed in the representative-agent model.
But where is there any inefficiency? Well, as we have just seen, it is really — I mean really – hard to find an inefficiency if you try to understand what happens in a recession in terms of a representative-agent model. In a representative-agent model, the entire analysis collapses to finding the equilibrium of the single representative agent. You have necessarily assumed that social and private costs are equal, because you have collapsed all of society into the representative agent. Private and social costs are not only equal they are identical, because that is how you set up your model. You have taken regular economics to its outer limit, and you have annihilated the multiplier. Good job!
But if you think of economics as the study of interactions between independent decision-makers rather than the study of a single decision-maker in isolation, and if you recognize that production and exchange between individuals may not just transmit, but amplify, disturbances across individuals, you may prefer to think of an economy as a network of mutually interdependent transactors whose decisions about how much to supply and demand reverberate back and forth across the network. Because it is so highly interconnected, private and social costs in the network need not be equal; indeed externalities are a characteristic feature of networks. The pervasiveness of externalities is well understood in payments networks in which the insolvency of a systemically important transactor can cause the entire network to collapse. The looser interconnections of the real economic network of production and exchange is less brittle than a specialized payments network, but that doesn’t mean that the former network is not also pervaded with externalities, a point beautifully articulated by the eminent Cambridge economist Frederick Lavington when, when referring to an economy at the bottom of a recession, he wrote in his book The Trade Cycle, “the inactivity of all is the cause of the inactivity of each.”
So is there any inefficiency in a recession? Of course. And is there an economic model that identifies the inefficiency? Absolutely.