'Trapped in the ''Dark Corners'''?

A small part of Brad DeLong's response to Olivier Blanchard. I posted a shortened version of Blanchard's argument a week or two ago:

Where Danger Lurks: Until the 2008 global financial crisis, mainstream U.S. macroeconomics had taken an increasingly benign view of economic fluctuations in output and employment. The crisis has made it clear that this view was wrong and that there is a need for a deep reassessment. ...

That small shocks could sometimes have large effects and, as a result, that things could turn really bad, was not completely ignored by economists. But such an outcome was thought to be a thing of the past that would not happen again, or at least not in advanced economies thanks to their sound economic policies. ... We all knew that there were “dark corners”—situations in which the economy could badly malfunction. But we thought we were far away from those corners, and could for the most part ignore them. ...

The main lesson of the crisis is that we were much closer to those dark corners than we thought—and the corners were even darker than we had thought too. ...

How should we modify our benchmark models—the so-called dynamic stochastic general equilibrium (DSGE) models...? The easy and uncontroversial part of the answer is that the DSGE models should be expanded to better recognize the role of the financial system—and this is happening. But should these models be able to describe how the economy behaves in the dark corners?

Let me offer a pragmatic answer. If macroeconomic policy and financial regulation are set in such a way as to maintain a healthy distance from dark corners, then our models that portray normal times may still be largely appropriate. Another class of economic models, aimed at measuring systemic risk, can be used to give warning signals that we are getting too close to dark corners, and that steps must be taken to reduce risk and increase distance. Trying to create a model that integrates normal times and systemic risks may be beyond the profession’s conceptual and technical reach at this stage.

The crisis has been immensely painful. But one of its silver linings has been to jolt macroeconomics and macroeconomic policy. The main policy lesson is a simple one: Stay away from dark corners.

And I responded:

That may be the best we can do for now (have separate models for normal times and "dark corners"), but an integrated model would be preferable. An integrated model would, for example, be better for conducting "policy and financial regulation ... to maintain a healthy distance from dark corners," and our aspirations ought to include models that can explain both normal and abnormal times. That may mean moving beyond the DSGE class of models, or perhaps the technical reach of DSGE models can be extended to incorporate the kinds of problems that can lead to Great Recessions, but we shouldn't be satisfied with models of normal times that cannot explain and anticipate major economic problems.

Here's part of Brad's response:

But… but… but… Macroeconomic policy and financial regulation are not set in such a way as to maintain a healthy distance from dark corners. We are still in a dark corner now. There is no sign of the 4% per year inflation target, the commitments to do what it takes via quantitative easing and rate guidance to attain it, or a fiscal policy that recognizes how the rules of the game are different for reserve currency printing sovereigns when r < n+g. Thus not only are we still in a dark corner, but there is every reason to believe that should we get out the sub-2% per year effective inflation targets of North Atlantic central banks and the inappropriate rhetoric and groupthink surrounding fiscal policy makes it highly likely that we will soon get back into yet another dark corner. Blanchard’s pragmatic answer is thus the most unpragmatic thing imaginable: the “if” test fails, and so the “then” part of the argument seems to me to be simply inoperative. Perhaps on another planet in which North Atlantic central banks and governments aggressively pursued 6% per year nominal GDP growth targets Blanchard’s answer would be “pragmatic”. But we are not on that planet, are we?

Moreover, even were we on Planet Pragmatic, it still seems to be wrong. Using current or any visible future DSGE models for forecasting and mainstream scenario planning makes no sense: the DSGE framework imposes restrictions on the allowable emergent properties of the aggregate time series that are routinely rejected at whatever level of frequentist statistical confidence that one cares to specify. The right road is that of Christopher Sims: that of forecasting and scenario planning using relatively instructured time-series methods that use rather than ignore the correlations in the recent historical data. And for policy evaluation? One should take the historical correlations and argue why reverse-causation and errors-in-variables lead them to underestimate or overestimate policy effects, and possibly get it right. One should not impose a structural DSGE model that identifies the effects of policies but certainly gets it wrong. Sims won that argument. Why do so few people recognize his victory?

Blanchard continues:

Another class of economic models, aimed at measuring systemic risk, can be used to give warning signals that we are getting too close to dark corners, and that steps must be taken to reduce risk and increase distance. Trying to create a model that integrates normal times and systemic risks may be beyond the profession’s conceptual and technical reach at this stage…

For the second task, the question is: whose models of tail risk based on what traditions get to count in the tail risks discussion?

And missing is the third task: understanding what Paul Krugman calls the “Dark Age of macroeconomics”, that jahiliyyah that descended on so much of the economic research, economic policy analysis, and economic policymaking communities starting in the fall of 2007, and in which the center of gravity of our economic policymakers still dwell.