A Disturbing Sentence in a Disturbing Journal Issue

Chari did not respond to Solow at all. He dodged Solow's comment. So what if DSGE models are coherent? They still don't have anything to contribute. So Chari leaves us with two choices: 1) inchoherence with utility or 2) coherence that is useless. Great!

I consider a sharp fall in NGDP to be a monetary contraction. Of course most economists don't agree with me, but then most also don't agree with each other (as to what constitutes monetary contraction.) Some consider high interest rates to be a contractionary monetary policy--others think that view is absurd.

But there certainly are conventional indicators that showed contraction before the Depression became "Great." For example, the monetary base declined 7.2% between October 1929 and October 1930. That's not my preferred indicator, but it certainly suggests contraction.

I like to use the world gold reserve ratio, which is the ratio of world monetary gold reserves to world currency stocks. That ratio rose 9.62% between October 1929 and October 1930. (An increase is contractionary.) That's a far more contractionary policy than the previous three years.

Because I'm not a monetarist, I'm not going to try to defend a model that links changes in NGDP and M, however defined. A fall in V has exactly the same effect as a fall in M. Of course a central bank can offset a fall in V with a rise in M. The question is not whether the central bank can prevent NGDP from falling, there's no doubt that it can under a fiat regime. The question is whether a policy of maintaining stable NGDP growth would have prevented the Great Depression. Ohanian's observation has no bearing on that question. But that's the relevant question for Ohanian's model of the Depression. And it's also the relevant question for re-allocation.

Oh, and rational expectations explain everything.

Sumner:

there's no doubt that it can under a fiat regime.

I don't understand how you can say that. If the Fed could forecast a coming decline in ngdp accurately and if they could see it coming at least 3 years in advance then they might could stop it, but the long lag times between policy and effect prevent them from stopping declines in ngdp without very good long range forecasting.

Sumner:

there's no doubt that it can under a fiat regime.

I don't understand how you can say that. If the Fed could forecast a coming decline in ngdp accurately and if they could see it coming at least 3 years in advance then they might could stop it, but the long lag times between policy and effect prevent them from stopping declines in ngdp without very good long range forecasting.

PS, Austrian theory says that even if you could foresee a coming ngdp decline say 3 years early, you still might not be able to stop it, because what caused the decline in ngdp is a shortage of capital, not a sudden, irrational lust for cash on the part of the people.

Fundamentalist, I don't think there are long lags in monetary policy, and I don't agree with what you call the Austrian view. I think drops in NGDP are caused by a shortfall in money.

"Level targeting" can help make NGDP growth more stable.

Scott Sumner 1 fundamentalist 0

Careful introspection (or alternatively read up on the new keynesian philips curve literature) will show that inflation today is proportional to the expected path of future prices.

It is therefore not necessary to anticipate a drop in NGDP or average prices so long as there is some proxy for market expectations of future inflation. All that is needed to stabilize inflation is to stabilize expectations of future inflation by targeting the spread on inflation indexed bonds.

You need to have a remarkably low opinion of markets to think this wouldnt work. If it didnt then all the Austrians could get rich shorting the over priced bonds...it is a win win for the Austrians, I dont see why you all dont endorse futures targeting...

"If you need proof that DSGE models are to understanding the economy what the lamp post is to the drunk with the lost watch"

Except that a DSGE model can be sufficiently generalised that you can really tell any story with it - it just forces you to recognise the implicit assumptions that lead you to that conclusion. This sounds like a reasonable thing.

In terms of your analogy it is the lamp post that lights up the entire area - and so of course it makes sense to continue searching under its light.

Matt Nolan wrote:

Except that a DSGE model can be sufficiently generalised that you can really tell any story with it - it just forces you to recognise the implicit assumptions that lead you to that conclusion. This sounds like a reasonable thing.

Not true. It's taken 30 years of hard work just to get DSGE models to where they are now, and they still leave a lot of important stuff out. They rarely have more than one or two kinds of agents, they usually have only one or a few kinds of assets, and they are still so difficult to work with that they remain the domain of a high priesthood. There are good reasons to wonder whether models with representative agents and no reason for money to exist are actually all that informative about the world we live in. We do know that, in general, the aggregation problem is insoluble. So the burden of proof is on those who advocate these monstrosities: show us that they actually predict better out-of-sample than simpler models.

To keep the math tractable in an economic model, you always have to leave some stuff out. No matter how complicated your model is, there is always something you can add to complicate it further and render it unsolvable. Since you can't solve the harder version, modeling alone can never tell you whether the part you left out changes the story. To judge that, you have to use the model to make out-of-sample forecasts. If your high-powered DSGE doesn't perform better than a simple time series model (and they generally don't), than how do you know that it's telling you anything useful at all?