A Critique of the Lucas Critique

The blogosphere, it seems recently, has been particularly rich in blogoyakking (sp?) concerning microfoundations. Wren-Lewis, Noah Smith, Krugman, Rowe, Plosser, and others just in the week prior to this post.

And this has caused a persistent itch of mine to clamor for scratching. Here goes.

The Lucas critique is fairly widely acknowledged to have at least exacerbated the trend toward insisting on microfoundations in macro theory, and thus the rise of New Keynesian Dynamic Stochastic General Equilibrium models. You know, representative agents showing rational expectations over generations, and all such things, reacting to policy changes. Which individual behaviors we can, more or less simply, just add up in order to understand the effects on the macro economy.

For those needing a brief review on "the" critique, Wikipedia is actually not bad, which I summarize. Lucas said:

"Given that the structure of an econometric model consists of optimal decision rules of economic agents, and that optimal decision rules vary systematically with changes in the structure of series relevant to the decision maker, it follows that any change in policy will systematically alter the structure of econometric models."

This is from his 1976 paper.

And the Wiki article summarizes the implications:

"The Lucas critique suggests that if we want to predict the effect of a policy experiment, we should model the "deep parameters" (relating to preferences, technology and resource constraints) that govern individual behavior. We can then predict what individuals will do, taking into account the change in policy, and then aggregate the individual decisions to calculate the macroeconomic effects of the policy change."

Thus, Bortis-style classical-Keynesianism was pretty much denuded: at best macro policy is effectively toothless, since that pesky agent will simply react to neutralize the policy actions;  and/or react so that the empirical regularities we were observing and making policy on will change. So we need to focus on micro and get that "right," the best we can do. The domination of Methodological Individualism on economic practice was on the rise.

I think Lucas got part of the diagnosis correct: individuals do react to macro policy changes but only indirectly. And he got most of it wrong. For the most part individual behaviors do not  directly change the  structure of the time series, but it is the change in the structure of the time series that changes individual behavior. Aggregate demand falls due to some shock - asset deflation, price shocks, war, pestilence, loss of income due to technology or offshoring. Producers then decrease output, and fire their workers.

This shock mechanism does induce individually optimizing behavior on the part of the business owner, given that what he optimizes is his profit function into which his production output enters.

For the newly unemployed person, since s/he no longer has a source of income, one can stretch and say that reducing consumption is individually optimizing at least of negative deviations from the budget. Of course this mechanism has it's own brutal zero lower bound - subsistence.  And does nothing to maximize utility.

Notice, no macro policy change is required here to start a recession or the recent depression; all we need is a sufficient shock to the system to start the negative feedback cycles rolling along.

We classical-Keynesian macroeconomists characterize these mechanisms as showing the fallacy of composition; the canonical principles of economics example is the paradox of thrift. You cannot use added-up individually optimizing behaviors to determine what is going to happen to the macro economy.

The reason is clear: emergent properties at the macro level have a life of their own. And its is these emergent properties like recessions and unemployment that directly affect the behavior of the economic person. I take here the strong version of emergent: a property or behavior at the aggregate level which cannot be observed or predicted at the individual level.

We do have methods that attempt to model this: the Keynesian aggregate expenditure model, but that is static. Dynamic macro econometric models, but those are difficult, and maybe it was because of these difficulties that the Lucas critique was so successful in attacking them.

I propose two things to restore the dominating importance of emergent macro properties on economic behavior. One is a recommitment to econometric modelling. Ever increasing data and increasingly better tools will continually improve modelling and forecasting results.

The other is a methodology that is vastly underused in economics, but widely used in various other sciences: network system analysis based on the mathematical theory of graphs. These methods lets us directly measure and model emergent dynamic behaviors from groups, like the individuals in an economy. No added up methodological individualism required; no agent-based model needed. Observe, model, and predict directly at the macro level.

While I believe empirical models, properly done, are fundamental to understanding and policy, network models provide us with a dynamic theory, emergent macro behaviors, that  support our correct Keynesian beliefs that it is the macro foundations of micro behavior that matter, not the other Lucasian way around.