Paul Krugman has a nice column on the micro-foundations of macroeconomics. I am hardly one to comment on the methodological issues. I’ve read and admired elegant macro theory papers by Robert Lucas, Peter Diamond, and various New Keynesians. I’ve always enjoyed these contributions as a tourist visiting from far away, as someone who doesn’t really use these analytic frameworks in my own work.
Even so, Krugman notes several points that resonate with me. In practical policy analysis, we rarely insist on using models that percolate up from some tight model of individual optimizing agents, or indeed on grounding policy analysis in any tight individual-level model of human behavior. Krugman notes that market analysts don’t demand micro-models when they analyze supply and demand for cigarettes and automobiles. That’s not how epidemiologists try to analyze HIV prevention interventions, either, how CMS actuaries try to forecast the cost of Medicare policy changes, or how health economists gauge the likely impact of accountable care organizations.
We don’t have the data to do that. We don’t have the tight micro-theories to do that, either. We need tractable, somewhat fuzzily-modeled alternatives, what Krugman calls “intellectual scratch-pads.” Policy analysis is a more rough-and-ready enterprise. Moreover, there is one embarrassing little secret in many areas of applied economics and social science. Most of the time, good intellectual scratch pads provide more accurate forecasts and better policy guidance than do more elegant and tractable micro-grounded models. This isn’t always true, but it often is.
In part, this embarrassing reality reflects the common reality that complex systems such as weather display properties one would not anticipate or reliably predict based on their simple component parts. In part, it reflects another embarrassing reality. Economics, psychology, and sociology will never be as tight or as powerful as the natural sciences. It’s a little silly to pretend otherwise.
I love, for example, Kevin Murphy and Gary Becker’s famous rational addiction paper. It shows that a dynamic optimization model of consuming an addictive good can produce a remarkable range of real behaviors such as binging and cold-turkey quitting. I would never directly apply this model to public policy. That’s not what such models are really good for. Developing these models remains essential work. It draws our attention to important causal mechanisms, to variables we might otherwise ignore, to interventions, subpopulations, and incentives we might not otherwise consider.
Two other points deserve attention.
First, macro-patterns are essential to detect omissions and over-simplifications in our micro-models. Most famously, when we see persistent unemployment, we can try to dogmatically argue this away based on some underlying microeconomic model. Almost everyone after Keynes realized that it’s more fruitful to ask what is missing in underlying models that presume frictionless market-clearing. Here and elsewhere, the Great Depression and our more recent deep recession provided great out-of-sample experiences to learn some important things about credit and labor markets, housing. The lessons may go even broader. How have marriage patterns changed under these economic stresses? How have income shocks altered demand for addictive goods and to crime?
Second, the micro-macro divide isn’t merely an issue for social science. It’s an issue for other areas of inquiry, too. Certainly that’s true of political philosophy and ethics. Many libertarians regard unequal social arrangements and income distributions to be justified if these percolate up (or could percolate up) from a process of fair and free exchange between people. If I want to pay Derrick Rose $100 to watch him sink some threes, so be it. Many other people want to pay him, too. So Derrick Rose will become very rich. The fair process that made created this pattern justifies the resulting wealth disparities.
Yet as Thomas Nagel noted in one of the great book reviews I’ve ever read, there’s something a little crazy about this approach. We can debate tax policy by considering what’s fair and reasonable in the deals people strike, a few at a time, acting in essential isolation from wider social concerns ranging from the intergenerational transmission of inequality to global warming. We’re likely to miss things that become apparent as we watch millions of apparently simple and fair exchanges play out among billions of people, operating through complicated human institutions over generations. Sometimes micro-arguments need macro-foundations, too.