Link to the original.
How Much Hoc to Add? (Wonkish and Methodological)
Mark Thoma objects to Tyler Cowen’s attempt to pigeonhole some of us as “Old New Keynesians”, and makes the case for flexibility in use of models. Indeed. But there’s something more systematic to the macro story – something that very much informs my work. Namely, “old Keynesian” and “new Keynesian” models represent two intelligent but imperfect responses to the problem of thinking about an economy in which people must make future-oriented decisions. Neither is fully convincing, but there are no better alternatives at this point. So what I do, and I hope others do as well, is to think about policy issues both ways, and try to “bracket” things in a way that leads to reasonably secure conclusions.
Like a lot of macro, this goes back to John Hicks. In his classic Value and Capital (1939, but he had much of the work done when he invented IS-LM), he laid out the basics of general equilibrium analysis – that is, thinking about how all market interrelate, instead of thinking about each one in isolation – but realized that he had a problem when key decisions in markets depended on expectations about the future. One way out is to assume Arrow-Debreu markets in which contracts can be signed now governing all possible future states of the world; never mind.
What Hicks proposed was an ad hoc solution (“temporary equilibrium”): let’s just assume that people expect future prices to be the same as current prices (“static expectations”). This lets you collapse the dynamic problem into a sort of quasi-static analysis of the short run.
And in particular, Hicks suggested thinking in terms of a three-good economy, in which the three goods were money, bonds, and physical output. When you do that and add temporary price stickiness, you get IS-LM.
Now, you can add a bit more flexibility by making assumptions about expectations more complicated in ways that seem more realistic – adaptive expectations, regressive expectations, etc.. In international macro, for examples, the Mundell-Fleming model – which is just IS-LM with borders – gets more convincing if you assume that investors expect exchange rates to revert to some kind of normal level. But it’s basically about squeezing dynamic issues into something that looks like a static framework.
And that’s not a criticism! IS-LM is in fact a deeply sophisticated framework, as evidenced by the number of economists who dismiss this crude stuff, then stumble ludicrously over even very simple issues.
But can’t we do better? The only workable alternative to temporary equilibrium at this point is rational expectations – assuming that what people expect about the future is what your model says they should expect. When you do that, and again add some realistic temporary price stickiness, you get New Keynesian macro, which is in a way more intellectually satisfying that old Keynesianism.
The trouble with New Keynesian macro is that it goes too far the other way – people aren’t that rational, or if you prefer, in the real world nobody knows what the right model is, so rational expectations breaks down as a concept. Also, NK models are much harder to work with; compare Obstfeld-Rogoff with Mundell-Fleming, and you’re going from simple diagrams to 60-plus-equation derivations.
So what’s an economist to do? Well, what I do is use both. Typically, I work things out first in terms of IS-LM, but then try to write down a stripped-down NK model with similar results; if I can’t make that work, it’s time for some hard thinking about where the difference lies. And sometimes this leads to insights that I don’t think would have come from either approach alone, like the crucial role of monetary credibility in making a liquidity trap possible.
The point is always to realize that the map is not the territory, but also to realize that IS-LM is a pretty good map, and that you can often fill in the gaps with a fancier—but not better, just different – NK model.