John Cochrane has a good post saying something I’ve been thinking about for a while. There are two disjoint orthodoxies in economics, one in policy and one in scholarship. Both are secure on their own territory, but they have little connection with each other. This isn’t obvious from the outside since many of the same institutions and even individuals contribute to the reproduction of both orthodoxies, but as intellectual projects they are entirely distinct.
Cochrane:
There is … a sharp divide between macroeconomics used in the top levels of policy circles, and that used in academia.
Static ISLM / ASAD modeling and thinking really did pretty much disappear from academic research economics around 1980. You won’t find it taught in any PhD programs, you won’t find it at any conferences …, you won’t find it in any academic journals… “New-Keynesian” DSGE (Dynamic Stochastic General Equilibrium) models are much in vogue, but have really nothing to do with static Keynesian ISLM modeling. Many authors would like it to be so, but when you read the equations you will find these are just utterly different models.
Static ISLM thinking pervades the upper reaches of the policy world. … If you read the analysis guiding policy at the IMF, the Fed, the OECD, the CBO; and the larger policy debate in the pages of the Economist, New York Times, and quite often even the Wall Street Journal, policy analysis is pretty much unchanged from the Keynesian ISLM, ASAD, analysis I learned from Dornbush and Fisher’s textbook, taught in Bob Solow’s undergraduate Macro class at MIT about 1978.
Note that Cochrane is agnostic about which of these projects is on the wrong track. This is a habit of mind we should all try to cultivate: The interesting questions are the ones where we can seriously imagine more than one answer.
Those are our choices? Ugh. Why would I study the world using the concept of "equilibrium"? Such a thing has never been observed and there are very good reasons from the real social sciences (ie, not economics) that it never will.
Not our only choices. I should have made this clearer. I reject both orthodoxies. But we have to be clear that they're not the same, in fact they have almost nothing in common. Which I think a lot of left critics of "economics" fail to do.
And didn't Noah Smith make more or less the same point about how financial industry modelers eschew DSGE and using more traditional Keynesian economic models.
Good point. The smarter ones used the models governments are using. The PIMCO guy and Krugman's hedge fundie nemesis went with their gut and were wrong about all sorts of things and lost money.
Cochrane: "New-Keynesian" DSGE (Dynamic Stochastic General Equilibrium) models are much in vogue, but have really nothing to do with static Keynesian ISLM modeling.
Like it or loathe it, the idea of IS/LM is to treat macro as a GE model. We can hardly say DSGE has nothing to do with GE. In this context "Dynamic" simply means that the optimization involves many periods. "Stochastic" simply means we model disturbances in a way that doesn't offend journal editors. I don't agree that these are really such drastic changes as Cochrane pretends. Historically, DSGE grew out of Hicks's GE.
He goes completely off the rails when he says the MPC in the NK model is zero. If that were true it would mean that a $1 increase in the household's disposable income results in $1 added to savings and $0 spent on current consumption. That's obviously wrong.
Cochrane's problem is that his contempt for Keynesian ideas is such that he can't be bothered studying them.
Like it or loathe it, the idea of IS/LM is to treat macro as a GE model.
Sure, but it has that in common with almost all macro models.
In this context "Dynamic" simply means that the optimization involves many periods.
Right. Which is very different from what dynamic means in any other context, which is a process of transition between states. The "D" in DSGE is certainly confusing, if not actively misleading. But in any case, the underlying model of a single representative agent choosing an optimal consumption over infinite (not just "many") periods is radically different from the model underlying ISLM, of distinct actors making saving and investment plans based on expected income for the current period, which plans may not be consistent ex ante.
He goes completely off the rails when he says the MPC in the NK model is zero. If that were true it would mean that a $1 increase in the household's disposable income results in $1 added to savings and $0 spent on current consumption. That's obviously wrong.
I think he is correct. In ISLM, consumption is a function of current income; the derivative dC/dY is what we call the marginal propensity to consume. In DSGE current consumption is not a function of current income, but of expected income over the infinite future. Since C is not a function of Y, there is no marginal propensity to consume in the sense of Keynesian models. It is correct that a $1 increase in current period disposable income will have a negligible effect on current consumption, since the increased consumption will be distributed over the entire lifetime. You may think this is "obviously wrong" as a description of the real world (and I may agree), but as a description of standard DSGE models it is correct. Of course, you can then add financial frictions or whatever to get a larger effect of current-period income, but that is not a feature of DSGE models out of the box.
Cochrane's problem is that his contempt for Keynesian ideas is such that he can't be bothered studying them.
On the contrary. I think Cochrane understands Keynesian models better than many New Keynesians, who ignore the fundamental differences between modern models and the ones used by earlier generations of Keynesians.
In DSGE current consumption is not a function of current income, but of expected income over the infinite future. Since C is not a function of Y, there is no marginal propensity to consume in the sense of Keynesian models.
No, C(0) is a function of Y(0) and expected income over the infinite future. So the MPC, which is the partial derivative of C(0) wrt Y(0), exists and is strictly +ve.
It is correct that a $1 increase in current period disposable income will have a negligible effect on current consumption, since the increased consumption will be distributed over the entire lifetime.
Obviously this depends on the rate of time preference, but I agree: in a NK model the MPC is very low. But Cochrane says it's actually zero. He's quite explicit about this.