Coordination Problem

Boiling it down.

|Peter Lewin|

Keynesian (and some Monetarist) versus Austrian views on the business cycle revolve around two basic empirical issues: the nature of capital and production and the nature of the national budget constraint.

Concerning capital: Ironically, Keynesians appear to assume a kind of “equilibrium always” story in which productive (human and physical) resources can, at low cost, be simply shifted to and plugged in to the production of whatever product or service happens to be the object of demand stimulus; and, further, that if enough of this is done, income, employment and tax-payments will rise precipitously. In this way economic-policy can produce economic growth that will expand the budget constraint sufficiently to be able to pay off the large past deficits (the accumulated national debt).

Austrians maintain that both of these (connected) assertions are false. Productive resources are not fungible. They cannot be easily moved and reassembled to produce whatever has been arbitrarily stimulated. The formation of productive resource combinations is the spontaneous result of idiosyncratic entrepreneurial judgments about future demand and revenue. There is nothing mechanical about it. Amorphous stimulus spending is not likely to produce an environment which entrepreneurs are willing to bet on. Rather, the result will be a malaise of uncertainty and reluctance to borrow and lend. The intertemporal budget constraint is real, not only for private citizens, but for the nation as a whole.

We cannot spend our way out of debt by hoping that the spending will lead to growth, especially when those doing the spending have no idea how to plant the seeds.

Keynesians stories do not imply Keynesian policies.

|Peter Lewin|

In a recent WSJ opinion piece, Jeffrey Liebman, a professor of public policy at Harvard’s Kennedy School of Government, articulated the current establishment understanding of the economic situation and the policies needed to deal with it. It is unabashed, one might say, simplistic, Keynesianism. Yet it represents a large and powerful consensus position. All the bad things that have happened to the economy are the spontaneous result of malfunctions in the market system – a system that has been insufficiently regulated and supported. The President proposed a reasonable and necessary jobs program and the Congress blocked it. Sensible people, Republicans and Democrats, know in their hearts this is true and need to get behind the president and support his spending-stimulus programs. (This is a very liberal interpretation – you can read the original for details).

The interesting thing for me is not so much the extent of the buy-in to this nonsense, as it is how they get there. I have always maintained that there is a sophisticated Keynesian story to be told, but it doesn’t get you Keynesian policy prescriptions unless you make some very unreasonable assumptions.

The essential element on which this turns is behavior in disequilibrium. There is no good theory of this. Yet all real behavior occurs in disequilibrium, in a situation in which different people have different expectations (including expectations about what things are worth), a situation in which their actions are bound to be inconsistent. Errors are inevitable. Do they on balance tend to get eliminated by feedback in the form of price (and other)  signals? What does “tend to get eliminated” even mean? What does this mean for the employment of resources?

To be more specifically Keynesian, imagine a situation in which, because of irrational exuberance, there is overinvestment in some sector that goes bad. The result is a precipitous fall in consumer spending as people reassess their wealth situations – revalue their assets and their net worth. There is contagion. The decline in spending precipitates quantity adjustments. Instead of prices falling rapidly to counteract the decline in spending (which would mean people made swift decisions to reduce prices), producers cut back on production and employment and so it goes. As John Hicks put it, this is a fix-price rather than a flex-price world.

I have no doubt that there are elements of truth in this. The world is neither completely fix-price nor flex-price, but is a bit of both, and sometimes more of one than the other depending on the circumstances. So what? So, what does this mean for economic stability? Does it mean that the market system is, therefore, habitually unstable, poised on a cliff’s edge in danger of being tipped over by the barest of bad speculative bubbles? Or is it rather a robust system, embedded in a set of multiple complex institutional networks that can deal with, and profit from, the inevitable errors that are made? Theoretical models can be created to support both stories, and, therefore, the matter is at bottom not a theoretical (logical) one. It is fundamentally empirical. I believe that we “know” from experience that the market system is robust and stable. But if people behaved differently, more erratically and if institutions did not provide the anchors that they do, the market system would be more unstable than it is. So what, again?

That the market economy is inherently prone to cyclical swings is, at best, a necessary condition for policy-intervention to stabilize it. It is not sufficient, not nearly. For sufficiency one needs to show that somehow macroeconomic policy-makers know enough to mitigate the cycles, rather than exacerbating them; and further that they can be trusted to do so. That is to say, pretty much all of the Keynesian policy prescriptions fail to come to grips with the familiar knowledge and incentive problems – with the central ideas of modern Austrian and Public Choice approaches. While the Post-Keynesians and Austrians may share the conviction that economic agents face unavoidable uncertainty, the latter see no reason to exempt policy-makers from that uncertainty. Why should policy-makers know any more about the future than the agents in the field? And why should we trust them to want to mitigate the cycle, rather than exacerbate it to their advantage?

 

Same questions, new models?

So, I have been reading a lot of stuff about complexity, emergence, agent-based –modeling (ABM), and the like – getting by with a little help from my friends, HT: Roger Koppl. It’s all very intriguing. My underlying agenda is policy relevance, particularly in this crisis environment.

In my last post I raise the question: Can we be certain, as a logical matter, that if certain conditions obtain, certain definite types of outcomes will result from the free-market process? As I understand it, much of effort in ABM-type investigations is into this type of issue. Not so much into the “logic” as we usually understand it, as into the necessary outcomes of certain kinds of processes, such processes being modeled by the behavior of the virtual agents of the model. By investigating these process-outcome variations one hopes to get traction onto the kind of economic-policy one feels appropriate – for example, supporting less intervention, support for property-rights, etc. – or the opposite.

There are some things I find very attractive about ABM (understanding that this may sound naïve, being the reactions of an outsider to the methodology and someone arriving late to the party). In particular, I like two things:

One, the ability to handle large numbers of interactions (variables) without the need to for closure (equilibrium), save for sufficient information to compute the next iteration. The system is open-ended, in disequilibrium.

Second, it makes me think about the uncertainty conundrum/paradox – the problem of how to model uncertainty that is shared by the agent and the modeler. I think it leverages the distinction between real/radical uncertainty (which is of a categorical, open-ended, inarticulabe nature – things will happen that could not have been foreseen or even imagined) and the effects of this uncertainty, which are arguably limited to a set of known categories. The AB model models the effects of the uncertainty – not the uncertainty itself, right?

But I have some concerns that  I am struggling to think through. The exercise seems to start from the conclusion. We observe a particular reality. So we wonder what kind of world might be producing this outcome and we try and put the essentials into a computer model. How do we know we are right? We see what comes out. If it is close, we are happy. If it is not, we tweak it until it is close, then we are happy. Where does this really get us? There are probably a very large number of agent-institution-profiles that will produce outcomes close enough to make us happy.

Of course, this is not so different, maybe not different at all, from what really goes on with business-as-usual empirical work in economics, based on classical probability theory and the estimation of structural parameters. There is the pretense of randomness, but, in reality, the researchers select for desirable results, which they report, and discard the undesirable ones. The upshot is a competition of models – often based on the same, or very similar data, and certainly on the same reality. Model choice then becomes a matter of credibility, prior inclination or something else; the “facts” never speak for themselves, and they don’t speak with one voice. In this regard ABM is really no different. The model choice here will depend on the plausibility of the model setup – the realism of the agent-behavior and institutional-framework modeled. I can see some potential for progress, as long as the number of outcomes from a particular setup-type is not too varied (as it is with the standard macro-empirical story). I guess time will tell.

In my next post I want to consider the question of action in disequilibrium and plausible Keynesian stories (ABMs and others) built around that. My punchline is that it all comes down to the burden of proof.