An upward-sloping IS curve

Brad DeLong's post on John Cochrane's upward-sloping IS curve triggered this post. But this is not about John Cochrane. It's about why tight monetary policy may cause real interest rates to fall, if monetary policy is expected to stay tight for long enough.

The story of an upward-sloping IS curve I'm putting forward here isn't really new. I read something roughly similar a few months back, but I have forgotten who wrote it. (It was a paper linked by a commenter on a previous post, had "Monetarist" in the title, and was written about 10 years back.) It's also not that different from the Old Keynesian "accelerator" story.

Understanding why an IS curve might slope up is empirically important, because it could help us understand why real interest rates might be low in a recession even if that recession had a monetary cause. (Normally, with a standard downward-sloping IS curve, an upward/leftward shift in the LM curve would cause higher real interest rates as we move along the IS curve.)

Stability may be a bit tricky with an upward-sloping IS curve and flattish LM curve, so you have to be careful how you tell the story about how you get to the new ISLM intersection.

The IS curve is a semi-equilibrium condition. Each point on the IS curve is a point in {r,Y} space such that the level of output demanded, at that real interest rate r, and at that level of real output Y, equals that same level of real output Y. The IS curve says absolutely nothing about output supplied -- about the level of output people and firms want to sell. That's why it's a semi-equilibrium condition. In full equilibrium, output demanded equals actual output and also equals output supplied.

Ignore government and foreigners for simplicity. C+I=Y.

Start in full equilibrium. Now let there be a monetary shock -- something that causes the supply of money to decrease and/or the demand for money to increase, so the LM curve shifts up/left. In a standard ISLM model, with sticky prices, that will cause a movement along the downward-sloping IS curve to a higher real interest rate and lower real output. The standard way of telling the story of the transmission mechanism is that monetary tightening caused the real interest rate to increase, which in turn caused consumption and investment demand to fall, which in turn caused output to fall.

Now lets just add one twist to the standard story. Suppose the monetary shock is not about today's money supply and/or demand, but is news about future money supply and/or demand. People suddenly learn news that causes them to expect the future LM curve will shift up/left. So they expect a future recession.

That drop in expected future demand, output and income will cause a drop in desired consumption and investment today, for any given real interest rate. Desired current consumption is an increasing function of expected future income, for anybody who does not live hand-to-mouth. That's standard. And the drop in expected future demand for goods means firms expect to be unable to sell as much output, and so the effective marginal return on investment falls. The extra capital can still produce as much extra output as before, but if firms cannot sell that extra output, the marginal return on investment falls. A recession causes unemployment of non-human capital just as it causes unemployment of human capital.

An expected future monetary tightening causes consumption and investment demand to fall.

Now, if you insist you could say that the IS curve slopes down, as normal, but the expected future monetary tightening, and consequent drop in expected future income, causes the IS curve to shift down/left. Each IS curve is drawn holding expected future income constant.

You could say that, but is it useful to say that? The whole point of drawing two curves -- a supply and demand curve, an indifference curve and budget constraint, an IS curve and LM curve -- is that most of the time the things that shift one curve are different from the things that shift the other curve. If an expected future shift of the LM curve automatically shifts the current IS curve, the two curves are not independent.

Rather than draw a very short run IS curve, which shifts every time expected future income changes, even if that future is the immediate future, might it not be better to draw a medium run IS curve, which takes a coarser-grained measure of time, so that changes in the near-future expected income cause movements along a given IS curve? Old Keynesians did this, when they built the multiplier right into the IS curve. The older Keynesians weren't so stupid as to believe that current consumption demand depends only on income received this very instant. Instead, they coarsened the measurement of time so they could roll the immediate past and immediate future into the current period, so that "current" demand depends on "current" income, and the two are equal at semi-equilibrium points on the IS curve.

If the recession caused by expected future monetary tightening is expected to last long enough, the effect on investment demand could be large. If the whole recession is rolled into the "current" period, the marginal propensity to consume plus the marginal propensity to invest out of "current" income could easily exceed one. And as every old/Old Keynesian macroeconomist knows, that means the AE curve in the Keynesian Cross diagram is steeper than the 45 degree AE=Y line, and that also means the IS curve slopes the wrong way. The IS curve is upward-sloping, because an increase in r shifts the AE curve down, which causes the intersection between the AE curve and 45 degree line to shift to the right, with a higher semi-equilibrium level of Y. (It's also unstable of course in the Old Keynesian analysis, but that analysis held the rate of interest fixed, and assumed all dynamics were in the response of output to output demanded.)

This sort of fuzziness due to coarsening the measurement of time will be anathema to rigorous New Keynesians. But we do it all the time. Firms don't adjust output and employment instantly in response to changes in demand. People's beliefs about the current period don't adjust instantly to what is currently happening. Stuff takes time, so we coarsen the measurement of time to make some of it happen instantly.

And current monetary policy is not something that can be separated from expected future monetary policy. By far the most important part of current monetary policy is expected future monetary policy. An expected future monetary tightening is a current monetary tightening. Monetary policy consists largely of managing people's expectations about the Aggregate Demand curve.

So when monetary policy tightens, and people expect this will cause a recession of a significant duration, we should expect to see both real output and real interest rates fall in response. If we want to draw a curve in {r,Y} space that stays roughly fixed in response to that monetary tightening, and if we want to call that curve an IS curve, that IS curve will slope up.

Comments

[Scott Sumner]

Yes, I've been making the same argument, albeit not using the IS-LM model (which I don't understand.) Perhaps the reason I never understood the model was it seemed to imply tight money raised real interest rates, and it clearly didn't in the 1930s, or today.

[Scott Sumner]

Nick, You keep stealing my comments. You said:

"Now lets just add one twist to the standard story. Suppose the monetary shock is not about today's money supply and/or demand, but is news about future money supply and/or demand."

I was about to pounce, to say "that would be the only type of monetary policy that really matters."

And then at the end you said:

"And current monetary policy is not something that can be separated from expected future monetary policy. By far the most important part of current monetary policy is expected future monetary policy. An expected future monetary tightening is a current monetary tightening. Monetary policy consists largely of managing people's expectations about the Aggregate Demand curve."

Your perfection is making it hard to find comments.

King showed in 1993 that in a dynamic forward-looking IS-LM model with ratex, tight money can reduce real rates. I'm horrified that most new Keynesians still don't seem to understand this.

[Nick Rowe]

Scott: yes, you were standing at my shoulder when I wrote this. Translating you into ISLM. (I have a certain fondness for the ISLM model. It is very flexible, and able to portray a number of different views).

"I'm horrified that most new Keynesians still don't seem to understand this."

I'm not sure if they understand it or not.

[Tom Geraghty]

How does all of this relate to the point Krugman made in "Japan's Trap"?

Monetary policy: It may seem strange to return to monetary policy as an option. After all, haven't we just seen that it is ineffective? But it is important to realize that the monetary thought experiments we have performed have a special characteristic: they all involve only temporary changes in the money supply.

. . . A permanent as opposed to temporary monetary expansion would . . . be effective - because it would cause expectations of inflation.

[Nick Rowe]

Tom: it's related, but a bit different.

In Paul Krugman's model, the original problem (the "shock") was an increased supply of saving that caused the IS curve to shift left and forced the natural real rate of interest down to zero. In my post, the (very short run) IS curve shifts left because expected future monetary policy is tightened.

Plus, I would add to what Paul said as follows: ". . . A permanent as opposed to temporary monetary expansion would . . . be effective - because it would cause expectations of inflation. *and it would increase expected future real income also*"

[Scott Sumner]

I'm not sure how many don't understand either, but it seems like a lot. I frequently confronted Keynesian arguments that 1929-30 couldn't have been tight money, because interest rates fell. And of course that seemed to be the standard Keynesian view of 2008. I seem to recall a certain NYT columnist frequently saying "This wasn't like 1982, a recession caused by tight money by the Fed." But why not?

BTW, what's wrong with you other commenters--this is a important post! Where are you?

[Nick Rowe]

Scott: they are all over at your place, fully employed commenting on your massive quantity of good quality posts this last few days!

[Scott Sumner]

Nick: It just shows that people prefer sensationalism over substance. Anyway, I'm going back to one a day.

[Your blog is cool]

Future monetary tightening causes desired investment and consumption to decline in anticipation of a decline in future income, sales, etc. Gotcha.

Wouldn't the expectation of a future upward LM shift cause the demand for liquidity to increase in the current period, shifting interest rates today?

This is fun. The central bank of IS/LMylvania must have outstanding credibility - the citizens do its work for it. It's barely even necessary. All the government needs to control inflation is a spokesman who will occasionally say, "If you don't do this, we're going to make a central bank."