Thursday, September 6, 2012

Scott Sumner, meet the New Keynesian IS curve. New Keynesian IS curve, meet Scott Sumner

This is really getting bad. From Scott:

"First the Keynesians say QE “won’t work” because in their flawed interest rate-oriented models it doesn’t work at the zero bound.  Then the Fed does QE1 and the dollar plunges 4% on the news.  Mere rumors of QE2 are enough to push stocks and commodities up sharply, as well as TIPS spreads.  So now they are forced to argue that it works, but it works for the wrong reason, so it still doesn’t work.  Or something like that.  I.e. it works because it’s a signal for more expansionary future monetary policies, not because of more money today.  Fine, but isn’t that arguing about how many QE transmission mechanisms can dance on the head of a pin?  Does it work or not, that’s the question.  (If you haven’t read my preceding post, do so first.)

But the bigger flaw in the Keynesian argument is that they don’t seem to realize that monetary policy always works through signaling, not just at the zero bound."

Ummm... right. Isn't that what we've always thought?

The zero lower bound isn't the point where expectations start to matter. It's the point where current interest rates lose traction because debt starts to look an awful lot like cash (what Glasner has called "some unpleasant Fisherian arithmetic"*). As Sumner says, "monetary policy always works through signaling", but that's the entire point of the New Keynesian IS curve, which was around long before the market monetarist blogosphere emerged to save the day. So why is Sumner suggesting that Keynesians "don't seem to realize" a point they've been making for decades?

Nick Rowe likes to talk about this as an upward sloping IS curve. I think it's more useful to think about it as a downward sloping IS curve that shifts right when monetary policymakers do what Krugman and Sumner both want. Analytically it amounts to the same thing, it just seems wrong to me to embed a monetary policy reaction function that says the monetary policymaker will act how they're clearly not acting right now! [UPDATE: thinking more, I think this might be the wrong way to make the case. I feel like I need to go back and reread that post more closely. I feel like I had an OK grasp of it when he first wrote it and it was pretty clear that it was almost a semantic difference between an upward sloping IS curve or a downward sloping one that moves in predictable ways... I need to revisit that before saying more, though].

I know Brad DeLong's probable answer but I hope there are others because as you all know that answer troubles me.

The other point I'd make is that just because this is something that was made explicit and formal by New Keynesianism does not mean it was something the Old Keynesians missed. The point was made explicitly by Keynes himself (although it was not made formally), and it's not a particularly complicated point. The reason the Old Keynesian models didn't have it was because they were simpler and the Rational Expectations revolution hadn't happened yet. The fact that expectations were not formally modeled does not mean that expectational issues weren't understood by Old Keynesians.

Here's my advice: If you want to know what a caricature Old Keynesian would say, take Scott Sumner at his word on what the disagreements are on monetary policy today. If you are interested in what they actually say you're going to need to look elsewhere.

And what endlessly frustrates me about this is that all this artificial division among economists makes good policy less likely not more likely.


  1. I've been reading a MM blogs for a long time, being a post-grad in econ I am also of course aware of New Keynesian macro models. I've come to the conclusion that MMists have a more radical and convoluted concept of expectations, more complex than rational expectations, but I don't really understand it. I think they just have a different philosophical approach to inter-temporal decisions regarding nominal spending/demand for money, based on some old esoteric monetary epistemology; I really can't make heads or tales of it.

    The main thing is that it's NOT expectations of interest rates, this is where New Keynesian economics departs from MMists: New Keynesian models generally ONLY have the short term nominal interest rate as the monetary policy tool. But to MMists it's about expectations of NGDP, and expectations that the central bank will do whatever it takes, whatever policy (unconventional or oherwise) is needed, or not. I.e. the signals are not about the future path of interest rates, the signal is about the underlying philosophy or attitude of the central bank.

  2. Daniel: I think the bit you crossed out was more or less OK.

    The main reason I prefer to think of it as an upward-sloping IS curve rather than a shifting IS curve? Well, the whole point of having 2 curves is that you can separate out the things that shift the first from the things that shift the second. So we define the IS curves as "you know, that other curve thingy, the one that does NOT shift when monetary policy changes, so you can figure out where you end up when you change monetary policy."


All anonymous comments will be deleted. Consistent pseudonyms are fine.