What happened to the Keynesian SRAS?

I’m not a huge fan of the Keynesian SRAS.   I think many Keynesians overestimate how flat it is at less than full employment.  But in recent weeks I have started to miss the SRAS, as the economics profession seems to have driven the IS-LM framework into an intellectual cul de sac.  A couple posts back I cited a Financial Times story that mentioned a Fed study suggesting negative 5% real interest rates were needed to provide the desired level of stimulus.

[Oops; Kevin just pointed out that the article did not say real rates.  I just sort of assumed that, because nominal rates can’t go below zero.  I also think this is probably how others interpret the finding, as the follow quotation from Krugman hints:

Today, however, with expected inflation roughly zero and a recession that is the fruit of past irrational exuberance, conventional monetary policy has run out of room. That’s the point of the Goldman Sachs exercise, shown below, which asks what the familiar Taylor rule would prescribe for monetary policy over the next few years; the answer is a Fed funds rate of -6, which isn’t possible.

I withdraw the implication in the following paragraph that Mankiw or Krugman erred in citing this study.  However, I still think the study’s findings are nonsense, as I believe expected inflation is non-negative zero over the next 12 months, and thus I interpret the Fed study as calling for a roughly 5% reduction in real rates, which, given nominal rates are near zero, can only be achieved by at least a 5% expected inflation rate.]  Back to the original post:

Recently both Krugman and Mankiw cited this estimate.  True, neither explicitly endorsed the -5% figure, but then neither seemed to regard it as patently absurd, which was my initial reaction.  Krugman and Mankiw are both brilliant theoreticians at elite Ivy League universities.  I’m at Bentley.  Even worse, I barely understand IS-LM.  But hey, that’s never stopped me before—so here’s my reaction.

Let’s start with the idea that we need a negative 5% real interest rate.    Doesn’t that mean that monetary policy would have to be expansionary enough to create expectations of at least 5% inflation?  And if expectations are rational, doesn’t that mean monetary policy would have to be expansionary enough to actually create 5% inflation?  And if the SRAS is very flat when the economy is deeply depressed (like right now) doesn’t that mean we’d need an extraordinarily large amount of NGDP growth to get 5% inflation.  In other words, are these Fed estimates (reportedly derived from a Taylor Rule model) at all consistent with the Keynesian SRAS?  I don’t see how.  But more likely I have misinterpreted something fundamental here, and some commenter will set me straight.

I suppose the reason I am having so much trouble here is that I have never understood why interest rates (or any price) has an important role to play in business cycle theory.  Let me use this analogy to illustrate my confusion.  We’ve all read articles saying something like “the price of gasoline has shot up to $4.00 a gallon, and thus we can expect consumers to start cutting back on consumption.”  If you are an economist, or perhaps I’d better say if you understand the laws of supply and demand (a much smaller group) you will start pulling out your hair and screaming “no, high prices don’t mean lower consumption.  Less supply of oil means lower consumption.  More demand for oil means higher consumption.  Higher prices mean . . . absolutely nothing.”  And a rise or fall in the dollar has no implications for the size of our trade deficit, and a rise or fall in interest rates has absolutely no implications for investment.  These are prices.  And the law of supply and demand says price changes have no implications for quantities.  None at all.

Now if someone like Brad DeLong or Paul Krugman is reading this they are probably thinking “Aha, another conservative who doesn’t understand the basics.  The thought experiment involved the Fed changing the interest rate, and the Fed does this by shifting the LM curve.  So there is no identification problem.”  But I do understand IS-LM well enough to know that monetary stimulus can increase both the expected rate of inflation, and the real interest rate.  And I know enough economic history to know that this “perverse case” of money affecting the IS curve becomes more and more probable the more dramatic (and clearly identified) the monetary shock.

Business cycle theory is about changes in the supply and demand for base money creating unanticipated NGDP shocks, which cause temporary employment fluctuations.  And except for a supply shock footnote, that is all.  The key assumption is sticky wages and prices.  The role of interest rates?  Almost entirely passive, and when they do play a role it is almost exactly the opposite of the prediction of the IS-LM model.  For example, during normal times the nominal interest rate has little short run impact on the demand for base money.  But when deflationary monetary policy pushes nominal rates close to zero, the demand for base money rises (although not as much as you’d think from recent events.)  This increase in base demand is deflationary.  Thus in one of the few cases where we can actually see interest rates impacting AD, the effect is perverse, lower rates actually reduce AD.

Now that I’ve reiterated my contempt for interest rate-oriented macro, let’s return to the SRAS problem and ask what went wrong with the Fed study.  After all, the Taylor Rule is a respectable model, and it seemed to work pretty well for about 25 years.  Why is it now giving bizarre policy advice, such as the need for the Fed to generate 5% inflation expectations?  I have several theories:

Let’s imagine the Fed did succeed in getting 5% expected inflation, and also assume that Keynesians are right about SRAS being relatively flat in a deep slump.  Then what sort of real GDP growth would we be looking at?  I can’t even imagine a plausible number.  I find it easier to start with some hypothetical NGDP growth rates.  Here are two hypothetical NGDP growth rates for the next 12 months, and in each case I’ll give you my view of the likely price/output breakdown, given what we know about SRAS:

a.  NGDP growth = 5%,  inflation = 1 to 2%, RGDP growth = 3 to 4%

b.  NGDP growth = 8%,  inflation = 2 to 3%, RGDP growth = 5 to 6%

Now you may differ from me as to the current slope of the SRAS, but is there any conceivable way one could get 5% inflation out of any plausible choice of optimal NGDP growth?  What would happen to nominal wages?  We are at 8.9% unemployment and going much higher, does anyone think nominal wages are about to accelerate?  Certainly Krugman doesn’t, he’s worried about wage deflation.  But then from where are we to get this rapid rise in the GDP deflator?

Let me be clear about one thing, I’m not denying the Fed could generate 5% inflation if they showed enough recklessness; you’ve seen me point to Zimbabwe numerous times.  Rather what I’m arguing is that there is no plausible NGDP growth path that the Fed would have as a policy goal, that would be expected to generate anything like 5% inflation.  But that means the Fed believes it could generate all the NGDP growth it thinks prudent at inflation rates far below 5%.  And that means the Fed could generate all the NGDP growth it thinks prudent at real interest rates much higher than negative 5%.  Somehow the “appropriate real interest rate” coming out of these Taylor Rule-type models is nonsense.  Why is that?

Perhaps the model is too backward-looking.  The authors probably plugged in numbers for recent inflation and real growth, and assumed that that macroeconomic environment was relevant when considering policy counterfactuals.  In the relatively slowly evolving business cycle of 1982-2007 that might have worked, but in the past year it has proven woefully inadequate.

Here is the basic problem as I see it.  If we assume that the policy works, in the sense of generating 5% or 8% expected NGDP growth, or whatever number you think the Fed is shooting for, then the IS curve will shift up vertically by a large amount, relative to where it would be if we continued with the real growth and inflation numbers of the past 12 months.  But if the IS curve shifted much higher, then we wouldn’t need anywhere near as low a real interest rate to generate adequate growth.  In other words, I wonder whether the Fed study has come up with an estimate of the appropriate real interest rate, assuming the policy is expected to be a complete failure.  You don’t have to be an extreme Ratex guy like me to see problems with that scenario.  A highly effective Fed policy capable of producing 5% actual inflation may not produce exactly 5% expected inflation, but it would certainly have some impact on the expected rate of inflation.  We’ve already got lots of people worried about inflation without any effective monetary stimulus.  Even liberals like Brad DeLong are talking about a potential inflationary time bomb in monetary stimulus.  Imagine what would happen to inflation expectations if the Fed actually did some effective stimulus (i.e. a penalty rate on ERs, an explicit NGDP growth path, etc.)

If you step away from the morass of IS-LM into the bright daylight of NGDP futures targeting, then everything becomes much clearer.  Now the market sets the supply and demand for base money to accommodate 5% expected NGDP growth.  On plausible estimates of SRAS, we’d be likely to get 1 to 2% inflation.  Interest rates—who cares?  But if you have to ask I’d guess the robust growth from a 5% NGDP target would raise real rates slightly above current levels.  An 8% NGDP growth target would raise real rates even further.  However given the low level of current output, I’d also guess that real rates would stay below their historical norms for a few years.  The closest analogy for what I am proposing is the dollar depreciation program of 1933, which dramatically reduced risk spreads between Aaa and Baa bonds.  That should be no surprise, as a strong increase in NGDP growth will sharply reduce debt defaults.

When you step away from a forward-looking NGDP futures targeting world, and back into the morass of IS-LM, everything becomes confusing again.  The Fed study says we need 5% inflation?  Why not?  If you see macro through the lens of interest rates, then I suppose negative 5% real rates might make a lot of sense.  Although low interest rates are the fruits of our deflationary monetary policies, people somehow believe that still lower rates will help.  No matter how deep the hole we’ve dug ourselves into, some will insist that we only need to dig a bit deeper.  Instead of seeking ever lower rates, how about trying an effective monetary policy that is expected to produce on target nominal growth (such as futures targeting) and hence expected to restore interest rates back up to the level of a healthy economy.

The Taylor Rule tells us that the deeper the slump, the more inflation we need (at zero nominal rates.)  The Keynesian SRAS tells us that the deeper the slump the less inflation we need.  I’ll take the model in which interest rates play no role.



14 Responses to “What happened to the Keynesian SRAS?”

  1. Gravatar of Jon Jon
    13. May 2009 at 08:03

    The taylor rule can be understood in the context of a linear control dynamics–it is itself a linear controller in a very basic engineering sense.

    Couple of observations then:
    1) The output of the taylor rule lacks intrinsic meaning. If you attenuate its prescription you may get different “performance” but converge nonetheless. This is why the “gain” or “weighting” parameters of the Taylor rule are ill-specified. The common “0.5” weighting is ad hoc.

    2) As you describe scott, the “economic system” is nonlinear. Therefore its fair to say that a historically tuned taylor rule is likely to be correct only around a narrow range of values consistent with ‘linearizing’ the economy. This isn’t just a question of the model being “backward” looking, but a consequence of essentially using linear approximations.

  2. Gravatar of Kevin Donoghue Kevin Donoghue
    13. May 2009 at 09:14

    I don’t see the FT, Mankiw or Krugman referrring to a real minus 5% rate. As I read it, all of them are referring to nominal rates and Krugman explicitly refers to the desired Fed funds rate.

    “But I do understand IS-LM well enough to know that monetary stimulus can increase both the expected rate of inflation, and the real interest rate.”

    Arguably IS-LM would be a better model if that were so, but it’s not so; at least in the Mark I version. Of course IS-LM has had a lot of bells and whistles attached to it over the years. But in the simple version, increased M means a rightward shift in the LM curve which raises Y and lowers r. Prices and wages adjust slowly (it’s basically a model for a slump after all) so you don’t need to think about the inflationary impact at that stage. As I read the Krugman post you link to, he has nothing fancier in mind than this short PDF file, which is presumably designed for an introductory macro course. (It’s not Krugman’s just one I googled; pages 5-8 deal with an increase in the money supply.) After all, the point he’s making is that Ferguson still hasn’t reached the point Hicks had got to in 1937, so in that context it wouldn’t make sense to use a modern model.

    I’m pretty sure Krugman wouldn’t use ye olde IS-LM for a situation where he thought inflationary expectations were playing a big role. Certainly it can be done – but AFAICR Sargent and Wallace showed back in the 1970s, in the early days of the RE craze, that if you modify an IS-LM model so that expectations of inflation are rational you get a model which is as neoclassical as they come. If Krugman thinks that way he should be in Chicago. Most likely he thinks the old-fashioned, static-expectations IS-LM is a good enough first approximation for journalistic work in the depths of a recession. In better times he can use more sophisticated (but not necessarily better) models. There’s horses for courses.

  3. Gravatar of Nick Rowe Nick Rowe
    13. May 2009 at 09:20

    Just to try to forestall one possible misunderstanding, there are 3 ways to draw the ISLM:

    1. With the real interest rate on the axis. Changes in expected inflation shift the LM.

    2. With the nominal interest rate on the axis. Changes in expected inflation shift the IS. (This is Scott’s way of drawing it).

    3. With both real and nominal rates on the axis. Changes in expected inflation change the wedge between IS and LM curves.

    All 3 ways are the same.

    And changes in expected future real output (possibly caused by current monetary policy) shift the IS also.

  4. Gravatar of David Pearson David Pearson
    13. May 2009 at 09:25


    Forgive me, but don’t your real growth “options” above betray a Keynesian bias?

    Each option seems to imply that real growth must accompany a 5% inflation rate. What is the justification for this assumption? Why can’t you have a 5% inflation rate and -2% real growth? How about 10% inflation and minus 5% real growth?

    Impossible? Of course, if you believe we must reach “full capacity” before the economy generates meaningful inflation.

    There is an alternate model. Each downgrade of real gdp estimates is met with an increase in bond yields (this is known everywhere but the U.S. as the “sovereign risk premium). The increase in government borrowing costs and needs causes upgrades in government deficit forecasts, which in turn cause interest rates to rise further. The expectation sets in the Central Bank would rather permanently monetize the structural deficits then face higher unemployment. Inflation expectations rise even while real gdp shrinks.

    Everyone tells me the above “emerging markets” model does not apply to the U.S.. Partially its their Keynesian outlook (no inflation with excess capacity), and partly its their going-in assumption that the Treasury and Fed will always retain credibility. I have no problem with these assumptions as long as proponents bother to defend them, which they usually don’t. The defense typically boils down to “foreigners and private savers will keep buying our debt because we’re the only game in town.” A nice statement, but not exactly an elegant piece of analysis that takes account of the very real alternatives (most notably short-duration T-bills).

  5. Gravatar of ssumner ssumner
    13. May 2009 at 10:45

    Jon, I agree about the nonlinear part, but that’s not my primary argument. Not only do we not need negative 5% interest rates, I’m not sure we need any reduction in interest rates, real or nominal. BTW, Robert King made this same point back in 1993, so it is a very well established proposition that monetary stimulus can raise both real and nominal interest ratres. I think the main problem is the backward-looking nature of the Taylor Rule.

    Kevin, Thank you very much. That was a very embarrassing error on my part. Because I believe the expected inflation rate is now roughly zero percent, or slightly higher, I am going to leave the post up, as I still disagree with the implied message of the Fed study (if I understand that message correctly.) I still believe that, given the near zero nominal rates, the Fed is hinting that we need a monetary policy capable of generating at least 5% inflation expectations.

    Nick, I knew about the first two versions, but not the third. Whichever version one uses, there is always some way to make it “correct”, the question is whether it is useful in any of its incarnations. My argument is that it leads one astray. We shouldn’t be thinking about business cycles and AD from a interest rate perspective. The fact that so few economists understand the point you made in your post today, speaks volumes about the effects of educating our students with IS-LM.

    David, The estimates do show Keynesian bias. But the entire Taylor Rule analysis is based on Keynesian theory, as is the SRAS curve, and I was trying to show that they lead to very different policy implications. BTW, I do think the SRAS is fairly flat right now, but I also believe we would get some inflation long before full employment–as is implicit in the numerical example I provided.

    The hypothetical case you discuss is a possibility, although I doubt it will play a major factor in the near term U.S. business cycle. On the other hand I think a bit of stagflation may come from statist economic policies out of Washington. But I expect that problem to look more like 3% inflation, 2% real growth. And only after recovery, during recovery the output/inflation tradeoff will be more favorable.

  6. Gravatar of Bill Woolsey Bill Woolsey
    13. May 2009 at 11:22


    It is simple. The only way out of the recession is “overshooting” Monetary policy cannot raise spending at all if will cause only slight inflation and large increases in real income. So, the monetary policy must be so expansionary that it would go beyond capacity and create excess demand pressures raising prices enough to raise inflation. Expecations of this will cause real interest rates to fall enough, to motivate the large “overshooting” increase in expending.

    So, a moderate increase in money that would just cause outpuat to recover won’t do anything. A large one will not only cause output to recover, it will cause inflation.

    This is the _same_ issue that I always bring up. Can monetary policy keep nominal income on target if interest rates hit the zero bound. I am especially interest in a 3% growth path with zero expected inflation, so that the zero nominal bound is a zero real bound–if nominal income stays on target.

    All of this talk about creating inflation expectations and lowering real interest rates so that there can be no liquidity trap is “academic” to me.


    I can see how expected inflation could rise based upon your arguments. Real interest rates then fall given nominal interest rates (including at the zero bound.) How does actual inflation rise?

    It seems to me you are making the argument that people will raise prices now in the face of surpluses. Finding that implausible isn’t uniquely “Keyesian.”

  7. Gravatar of travis travis
    13. May 2009 at 13:24

    Kevin: If nominal rates are supposed to be -5%, and t-bills are at 0%, aren’t real rates at -5%? I don’t see how focusing on nominal rates instead of real rates changes what Scott was saying: it’s not plausible that we’ll get nominal rates of -5% unless the Fed went Zimbabwe crazy.

    Keep posting Scott. I’ve learned more economics from this blog than anywhere else.

  8. Gravatar of ssumner ssumner
    13. May 2009 at 15:02

    Bill, I am a bit confused by your response. Are you describing the Keynesian view (that monetary policy can only work through interest rates?) That is certainly not my view (zero rates or not) and I had assumed it wasn’t your view either. Just create enough excess cash balances to get 3% expected NGDP growth, or 5%, or whatever number you are targeting. NGDP futures will do this, regardless of any zero bound. Indeed under an optimal policy it is very unlikely that nominal rates would be at 0%. Nick has a post today showing that expansionary monetary policy raises nominal rates. So I don’t see the problem.

    Travis, Thanks. I find this whole topic more and more confusing, the more I think about it. I think you are right that the study was probably implying we need 5% expected inflation. But I can’t be sure. I suppose they might have meant something like “we need to increase the money supply by an amount that would normally reduce rates by 5%.” Or maybe just that “we are 5% away from where the Taylor model says we should be.” In any case, I just wanted to show that if people inferred from this study that we need 5% expected inflation, then I have a big problem with that inference.

  9. Gravatar of Bill Stepp Bill Stepp
    13. May 2009 at 16:33


    Why would any “rational” lender lend money at a negative real interest rate? Seriously.
    [I use quotes around “rational” because, as a follower of Mises on this point, I don’t think there’s such a thing as irrational action, and that the rational/irrational distinction is much ado about nothing or at least very little.]

  10. Gravatar of Nick Rowe Nick Rowe
    13. May 2009 at 18:17

    Bill Stepp: I think you are right (or very nearly right).

    We argue that negative nominal rates are impossible because people would store cash under the mattress.

    We could equally argue that negative real rates are impossible, because people would store cans of baked beans under the mattress.

    Maybe I should do a post on this!

  11. Gravatar of Bill Stepp Bill Stepp
    14. May 2009 at 01:40


    Go for it!

  12. Gravatar of Nick Rowe Nick Rowe
    14. May 2009 at 04:17

    Bill: Thanks! I did. http://worthwhile.typepad.com/worthwhile_canadian_initi/2009/05/could-the-natural-rate-of-interest-really-be-negative.html

  13. Gravatar of ssumner ssumner
    14. May 2009 at 16:43

    Bill, If your only choices are cash at a negative 5% real rate (because of 5% inflation), or lending the money to a bank at 4% (negative 1%) I could easily see a rational person lending to the bank. But I also agree with Nick that significantly negative real rates are very unlikely, except perhaps for cash and a few risk-free assets like T-bills.

    Nick, Interest post. I mostly agree.

  14. Gravatar of TheMoneyIllusion » That’s “former central bankers” TheMoneyIllusion » That’s “former central bankers”
    19. May 2009 at 10:42

    […] few days ago I posted a discussion of the apparent conflict between a well-publicized Fed study of the Taylor Rule’s policy […]

Leave a Reply