Why I don’t like IS-LM (reply to DeLong)

My normal Sunday post will have to wait, as Brad DeLong has a recent post where he asks why I don’t like IS-LM.  Since I get annoyed when people don’t take my open letters seriously, I can hardly ignore his post.  Plus it’s a very good question.  Because I don’t have complete confidence in any of my answers, I will go for quantity rather than quality.  But I’ll say right up front that I doubt there are any theoretical flaws in the IS-LM model.  As Brad DeLong puts it:

But I have not yet seen a theory of nominal spending or real output determination that does not have an IS-LM representation…

I think he is probably right, and most of my reply will be on pragmatic grounds, not theoretical.  Nevertheless, let me start off by taking a stab at a theoretical argument.

Update:  (4/13/09)  David Glasner just reminded me of my main objection to IS-LM, it doesn’t determine the price level (or NGDP.)  I’m sure that there is a good answer to this objection, so let me just anticipate the only two responses that I can think of right now.  First, some might argue that the classical model of money supply and demand determines the long run price level, but IS-LM is a short run model.  But as I argued earlier, that confuses long run with “future” and short run with “present.”  In fact, “right now” is the long run for policies instituted earlier.  In that case you’d want to model current changes in the price level in terms of past changes in money supply and demand, not IS-LM.  Others might argue that Keynesians have a “bottleneck” view of inflation, which occurs when output pushes against capacity.  Unfortunately the most perfect monetary experiment in American history, the 1933 dollar depreciation program, decisively refuted that mechanism.  Just to review, when the dollar was sharply depreciated after March 1933, the WPI rose 14% in 4 months, (sticky) real wages fell sharply, and industrial output soared 57% in 4 months.  But since output ramained far below capacity, prices clearly weren’t rising due to full employment.  Instead, the rise in prices that began immediately in April supported Friedman’s view of the Phillips Curve, unanticipated price level increases raise output, and not the NAIRU view that when output is below capacity inflation falls.  (End of update.)

A World Without Interest Rates

Michael Woodford and I agree on one thing; changes in the expected future path of monetary policy are much more important than changes in the current stance of some monetary instrument.  But he prefers to define monetary policy in terms of how a central bank sets its policy rate relative to the always changing Wicksellian equilibrium rate (i.e. the rate necessary for macro stability.)  He feels so strongly about this that he developed a model without any money at all.  I am just the opposite; I like to think of policy in terms of the supply of a medium of account (such as base money) relative to the real demand for that asset.  I like to think about economies without interest rates.  Bear with me for a moment while we consider just such a hypothetical case.  (By the way, I believe that there is actually a bit of money hidden in Woodford’s model, so I don’t doubt a sharp-eyed reader will find some hidden interest rate in mine.)

The model I envision is a simple island economy with lots of fruits and vegetables for sale.  Suppose the islanders adopt Federal Reserve notes as a medium of exchange (to avoid the inconvenience of barter.)  It doesn’t matter where the government gets these notes, although it might matter how they are introduced into the economy.  I suppose they could be given away, or used to pay government worker salaries.  In that world, how does a doubling of the money supply cause prices to double?  I have always thought the monetarist excess cash balance story described the process well.  It is counter-intuitive in a way, as it is an application of the fallacy of composition to the demand for money.  This paradox gets resolved by assuming the government determines nominal cash balances and the public determines real balances.  Interest rates play no role.

Obviously our modern economy does have a financial system, and obviously that fact changes the transmission mechanism somewhat.  But my hunch, or intuition, is that it changes it much less than most people imagine, even in a modern economy.  (Later I’ll supply some evidence for this intuition.)  People are fooled into thinking interest rates lie at the center of the monetary policy transmission mechanism for two reasons:

1.  Interest rates delay the response of prices (and NGDP) to monetary injections, and they are very visible.

2.  The excess cash balance mechanism only seems applicable to the long run equilibrium, and thus seems unimportant for current policy issues.  This is a very dangerous misconception, as we will see.

Let’s start by asking why interest rates matter.  In principle, a doubling of the money supply should immediately double the price level.  Most economists think that the reason it does not is that wages and prices are sticky (although other factors also may play a role.)  So if prices don’t double in the short run, then some other variable must equate the now doubled nominal supply of money, with the not yet doubled demand for money.  That variable is nominal interest rates, which are not at all sticky.  They immediately fall to a level where money supply and demand are equated.  Now we are in the world of IS-LM.

But note that in the example I just gave, the short run fall in the interest rate can be viewed as simply a symptom of the stickiness of wages and prices, and also as a factor that slows down the rise in nominal spending (by temporarily raising money demand or lowering velocity.)  So I it doesn’t have to be viewed as essential to the monetary policy transmission mechanism.  It may be central, but as we saw in the island without banks, it need not be central to the determination of nominal aggregates.

At this point an IS-LM supporter might be extremely frustrated, because so far I am only concerned with what determines the path of nominal aggregates, whereas IS-LM is usually viewed as a model of real GDP determination, with inflation a sort of afterthought that depends on SRAS.  I agree that you cannot get real effects from money with a model of perfect wage and price flexibility.  However you can get such real effects with wage and price stickiness, but without interest rates playing any sort of important role in the transmission mechanism.  All you need is a Phillips Curve (or SRAS) model, and then some nominal shocks.

So here is where I end up.  An increase in the money supply that is expected to be permanent will raise the future expected NGDP (or price level) proportionately.  This will increase the current level of nominal spending, but to a lesser extent.  Is it necessary for interest rates to fall in the short run to help this process along?  I don’t see why.  If you keep the current nominal rate stable, real rates would fall as expected inflation rises.  I think that is what a lot of IS-LM types would see as one possible mechanism.  I am told that this example would show up as a shift in the IS curve, if nominal rates are on the vertical axis.  (But I may be wrong.)

Or maybe nominal rates rise one for one with the rise in expected inflation, and real rates are unchanged.  Does that mean the monetary expansion will have no effect?  No, because in that case velocity would rise as the opportunity cost of holding cash rises.  If nominal spending rises and wages and prices are sticky, you will have real effects.  Or maybe nominal interest rates (and hence real interest rates) would fall.  In that case the IS-LM model can easily explain the process.  The point is that I don’t care what happened to interest rates, if the expected future NGDP rises, that will put upward pressure on current AD, and hence current NGDP.  And this will occur for basically the same reason that an increase in the expected future price of commodity X or financial asset Y will increase the current price of that asset.

None of the preceding explains why I have such a negative view of IS-LM model, instead, my real objection is entirely pragmatic—almost every time I see people use IS-LM, they seem to misunderstand what is going on, at least as I see it.  I see the basic problem as follows, most people, and even most economists, seem to think that they can ascertain changes in the stance of monetary policy simply by looking at changes in short term nominal interest rates.  I see short term rates as merely an epiphenomenon, something that accompanies unexpected changes in the supply of money, relative to the demand for money.  But the key driver of current changes in AD is expected future changes in NGDP.  (I believe this may be part of what Keynes meant by “confidence.”)  And this is determined by changes in the expected future supply of money (relative to changes in demand.)

I should probably add, however, that in the current crisis those using the monetarist approach have done no better than Keynesians, as they seem to assume that one can ascertain the stance of monetary policy merely by looking at changes in the stock of money, however defined.  But when the demand for money is unstable, the monetarist approach is also pretty useless.  Add in positive interest payments on reserves, and the monetarist model becomes totally useless.  Before getting into my pragmatic arguments, let’s look at a quotation for Robert King  (JEP, 1994, pp. 77-78):

“If changes in the money stock are persistent, then they lead to persistent changes in aggregate demand. With a rational expectations investment function of the neoclassical form, persistent changes in demand for final output lead to quantitatively major shifts in the investment demand schedule at given real interest rates. These effects are generally sufficiently important that real interest rates actually rise with a monetary expansion rather than fall: the IS curve effect outweighs the direct LM curve effect. Further, if persistent changes in the money stock are only gradually translated into price or wage increases, then there are very large additional expected inflation effects on nominal interest rates.”

When I originally read this passage from King, it confirmed my intuition about the difficulty of applying IS-LM to the real world.  Brad DeLong is right—every policy scenario has an IS-LM representation.  But if they are not intuitively obvious, if most people get them wrong, they will be useless.  Everything that occurs can be expressed in terms of C+I+G+NX, or in terms of MV=PY, but that doesn’t necessarily make those frameworks useful either.

Pragmatic arguments against IS-LM

In an earlier post I talked about the similarities between 1929, 1937 and 2008.  In all three years the expected rate of nominal GDP growth looking many years out in the future seemed to fall precipitously in the autumn.  I believe that all three shocks were monetary, broadly defined.  But I don’t think IS-LM is a useful way to think about these shocks, for a variety of reasons.

1.  In my research on the gold standard I found that almost everyone (even Friedman and Schwartz) seemed to view the initial 1929-30 downturn as a bit of a mystery.  I realized they were looking in the wrong place.  The medium of exchange in those days was gold, and its value or purchasing power was determined in a world market.  Between 1926 and 1929 central bank gold reserve ratios had been rising at 2.5% per year.  In the 12 months after October 1929, that ratio soared 9.6% worldwide, as central banks greatly increased their demand for gold.  This deflationary shock totally changed the expected future path of NGDP, and thus led to sharp declines in stock and commodity markets.  (Essentially the same thing happened to stock and commodity markets during 1937-38 and 2008-09, but for slightly different reasons.)

F&S missed this because they looked at the monetary aggregates, and the real demand for those aggregates will rise under a deflationary scenario (as long as bank panics haven’t hit yet.)  Keynesians saw nominal interest rates fall sharply over that 12 month period (real interest rates were hard to estimate back then) and assumed tight money could not be the problem.   I view that sort of reasoning as a misuse of IS-LM, for the reasons outlined in the King quotation.  Now you can say that a model can’t be blamed simply because its proponents misuse it, but we are talking about most of the leading Keynesians who have studied the Great Depression, unless I am mistaken.  (I am thinking of Keynesian explanations that speculated about “animal spirits” weakening, or consumer sentiment worsening after the stock crash.)

2.  Since 1986 I had been pushing a forward-looking monetary policy aimed at stabilizing NGDP expectations (through the use of NGDP futures contracts.)  When Lars Svensson recently started arguing for a policy of targeting the forecast, I resumed work on this topic.  I came to the conclusion that the only sensible way of describing the stance of monetary policy was relative to the stance expected to hit the policy target.  If the goal is 2% inflation, then a policy producing 1% inflation expectations is too contractionary and one producing 3% inflation expectations is too expansionary.  In addition, I have always preferred a NGDP target, and because of their “dual mandate” I have always assumed that NGDP growth described the Fed’s policy target just as well as inflation.

Last October NGDP expectations obviously fell well below the 5% norm of recent decades, and even below zero.  That’s when I went on this crusade against current Fed policy, and in some ways against much of the profession.  I immediately discovered that Lars Svensson had had much less influence than I imagined (although people like Krugman praise his work.) Many people were incredulous when I described recent Fed policy as the most contractionary of my lifetime.  Most articles about the Fed say something like “well at least no one can accuse Bernanke of not doing enough.”  I can only suppose that this attitude comes from the dominant model in our schools, IS-LM, which if used at only a superficial level (ignoring how monetary policy can shift the IS curve) could make it seem like policy has in fact been very expansionary.

3.  Laidler’s book on monetary history pointed out that Wicksell adopted a very quantity theoretic approach in the early 1920s.  When I read that passage it rang a bell, as I immediately thought of the other most famous proponent of the interest rate approach to monetary policy, Keynes.  Specifically I recalled that in his 1923 book, A Tract on Monetary Reform, he also adopted a quantity theoretic approach.  And for those of you who don’t know much about Keynes, he had already been developing an interest rate approach before WWI.  What was it about the early 1920s?

Then I recalled that there was a resurgence of the quantity theory under the name of ‘monetarism’ in the 1960s and 1970s.  How are these periods similar?  In both periods many countries very experienced high rates of inflation, rates so high that the interest rate approach did not offer the sort of intuitive explanation that the quantity theory offers.  Of course the early 1920s saw hyperinflations lasting only a few years, whereas the 1960s and 1970s saw lower but still very substantial rates of inflation which persisted for several decades (especially in places like Latin America.)  If excessive monetary expansion was the cause, as most believed, then the quantity theory’s money supply indicator seemed easier to visualize than the Keynesian model’s low interest rate mechanism (as most people seem fixated on nominal interest rates, which were then quite high.)

4.  I will admit that the IS-LM model seemed to work reasonably well during the Great Moderation (1983-2007), but I think that was not for the reason many assume.  The Taylor Rule works on several levels.  If inflation is too high you are supposed to sharply raise interest rates.  But the way you raise interest rates is by lowering the money supply (relative to what had been expected.)  And in fact it may have been changes in the money supply, and more importantly changes in the expected future path of the money supply relative to demand, that kept things reasonably stable when the Taylor Rule was still credible.

5.  My basic problem with the IS-LM approach is that it doesn’t seem to be all that useful (compared to monetarism) in high inflation periods when inflation expectations are changing fast and the real interest rate is hard to estimate, and it doesn’t seem very useful in years like 1929, 1937, and 2008 when expectations are falling so fast that policy seems expansionary when it is actually contractionary.    In neo-Wicksellian terms the problem is that the Wicksellian equilibrium rate falls faster than the policy rate.  (And I don’t think monetarism is useful in those deflationary periods either.)  But IS-LM works fine when things are pretty stable.  The analogy I would use is that IS-LM is like a bus that has brakes that work fine when going right down the middle of the road, but which fail if the bus is sliding of the road to the right (deflation) or to the left (high and variable inflation.)

6.  In my post on George Warren I described FDR’s 1933 policy of dollar depreciation.  I argued that the transmission mechanism that was clearly visible in 1933 (higher current price of gold leading to higher expected future money supply and prices) is actually the same mechanism that always drives NGDP growth; it’s just that it is rarely so obvious.  Indeed, in some ways that post is the best way to understand my approach to monetary economics—it’s all about targeting expectations.  Of course you can reach many of the same conclusions using a forward-looking IS-LM approach, and to some extent Gauti Eggertsson did so in his 2008 AER piece.  Whereas I argued the higher gold price was the key, he argued it was lower real interest rates as inflation expectations rose.  In my view the gold price approach is much more fruitful; it allows us to understand specific movements in asset markets during 1933 better than the real interest rate approach.  But either approach seems light years ahead of either a simple Keynesian (nominal interest rate) approach, or a simple monetarist approach.

7.  Those who read this blog know that I have had a lot to say about “liquidity traps.”  I think what most people are missing is that causation tends to run from contractionary monetary policy to low interest rates.  Low rates are not a sign of monetary policy ineffectiveness; rather they are a symptom of excessively tight monetary policy.  So far as I know, virtually every zero interest rate environment in history has occurred when central banks have run deflationary monetary policies.  The IS-LM framework leads to a sort of defeatism about the possible efficacy of monetary policy when rates are near zero.  In one sense that’s not hard to understand, after all it is deflationary policy that got us into the mess, why should we expect more from monetary policy in the future?

In contrast Krugman’s “expectations trap” approach says what really matters is credibility, one needs a policy that is expected to boost inflation (or I would say NGDP growth) expectations.  Krugman clearly understands that monetary injections expected to be permanent can be effective at zero rates, but if he thinks central bank attempts to reflate will often lack credibility, then he might regard IS-LM as a useful heuristic device.  Perhaps he is pessimistic about monetary policy effectiveness because of the way he reads history (1930s America, 1990s-2000s Japan, and right now.)  I find very little evidence in any of those three cases to support monetary policy ineffectiveness, but he probably reads the evidence differently.

8.  Because it’s getting late I’ll commit the sin of quoting myself, from before I had gotten burned out on this topic.  This is from an appreciation of David Laidler:

If this crisis shows anything, it is the importance of having people who can think along a number of different dimensions at once.  I am not impressed by young hotshot theoreticians who tell me that the liquidity trap “all boils down to X.”  Or “monetary policy is simply this.”  No, it doesn’t all boil down to anything, as you may have gathered from my other posts.  You cannot understand liquidity traps without understanding the history of the Great Depression, and 1994-2008 Japan.  (In my other posts I argue that neither “trap” was what it seemed to be.)  You need to understand exactly how the Fed works, what the policy of paying interest on reserves means, and whether that rate could be negative.  You need to understand the issue of policy credibility, and what sort of signals markets are likely to find credible.  You need to understand the pros and cons of unconventional open market operations involving risky assets.  You need to understand the many different channels by which monetary injections can impact AD.  You need to understand the international political friction caused by currency devaluation.  It even helps to understand how word choice subtly shapes our thinking (i.e., use of the term “expectations trap,” for what is not necessarily a trap at all.) It doesn’t “all boil down to” anything.  It’s complicated, not mathematically complicated, but conceptually complicated.

There is no reason why a central bank should not be able to convince markets that it wants to do something that it in fact wants to do. I see no reason why markets wouldn’t believe Bernanke if he said he wanted to do the very thing that as an academic his suggested should be done.  One of my commenters said that Krugman argued that the Fed might not be able to credibly commit to more than 2% inflation.  I think that might be a bit low, but even if it were true, I wouldn’t be concerned at all.   In this severely depressed economy, it would take at least 5% NGDP growth expectations to get 2% inflation expectations.  And that would be far better than what we are likely to get, even with massive fiscal stimulus.  Furthermore, if the Fed can get 2% inflation expectations with the fiscal stimulus, and credibility is the only problem, then the Fed should be able to get 2% inflation expectations without fiscal stimulus.  In other words, we have just ballooned the national debt for no good reason.

Of course there are lots of “foolproof” ways to break the zero bound.  One is the option of currency devaluation, proposed by Svensson, McCallum and others.  Another is the CPI or NGDP futures contract targeting idea.  But even if these ideas aren’t politically acceptable for various reasons, at the very least these options tell us there is something very wrong with the way people interpret IS-LM, as if it can somehow tell us when monetary policy has become ineffective.  It is particularly inappropriate when applied to a country like Japan; that clearly could have stopped its currency from steadily appreciating in recent decades, if it was so inclined.  (And if someone tells me the U.S. would have objected, then we’re back in 1932 when the gold standard was the real problem, not zero rates.)

When I was learning hydraulic Keynesianism at Wisconsin in the 1970s, we were taught that the slopes of the IS and LM curves showed the relative effectiveness of monetary and fiscal policy.  Even as an undergrad that made no sense to me, as I thought (and still think) there is no effective limit on the size of monetary injections.   A few years later people stopped talking that way, and only started again when rates fell close to zero.  I wish people could see that low rates are a sympton of tight money, not a time to revive the IS-LM analysis of the relative strength of fiscal and monetary expansion.

I got very worried in December 2007 when the Fed cut rates less than expected, stocks plunged, and nominal rates in the Treasury market fell from 3 month to 30 year maturities.  I wasn’t concerned so much about the economy, which seemed to revive a bit when the Fed reversed this mistake in January, but rather what worried me was the fact that most economists seemed to assume that rates were falling because of easy money, whereas the bond market was clearly signaling it was tight money that was depressing bond yields.  Markets respond to the unexpected part of the announcement.  I thought the Fed had learned a lesson.  I was wrong.  BTW, I don’t doubt that monetary shocks often do move interest rates in the “right” direction, but the point is that we can’t count on that being true, and it’s least likely to be true when we most need it to be true—when we are in unfamiliar territory.

I could go on and on, my manuscript on the Depression has dozens of examples of where I think a forward-looking approach that focuses on the supply and demand for the medium of account is superior to the IS-LM approach, but I doubt anyone is still reading.  So let me thank Brad DeLong for raising this issue.  Although we have different approaches to the monetary transmission mechanism, I think we have some methodological similarities.  He has also done a lot of research on monetary history, and takes the sort of broad, integrative approach that I think is desirable.

Postscript:  I share Milton Friedman’s preference for looking at lot’s of issues from a partial equilibrium perspective (a model for interest rates, another model for NGDP, another model to partition NGDP into prices and real output, etc.)  I’m sure if he were alive today he could have done a much better anti-IS-LM piece.  BTW, I am now planning an important post on Friedman and Schwartz’s Monetary History, which will complement this post.  It will evaluate Krugman’s critique of F&S.  I have mixed feelings about that critique, and I think that in discussing the issues raised by Krugman you will get a much better idea of where I am coming from.  It should be ready by later in the week, after I finish grading and taxes.

PPS.  The architect Christopher Wren is buried in a simple tomb in the basement of St. Paul’s Cathedral.  I belive the inscription reads “If you seek a monument, look around you.”  I’m not trying to compare myself to that great man, but I suppose I could say something similar about my blog, which is almost entirely an implied critique of the IS-LM way of thinking about monetary economics.



33 Responses to “Why I don’t like IS-LM (reply to DeLong)”

  1. Gravatar of Jon Jon
    12. April 2009 at 18:41

    First–I haven’t entirely comprehended your response–I just thought I’d toss it my objections to IS-LM:

    1) The real-world is quantized–not all configurations are achievable.
    2) The shape of the IS-LM curves is actually unknown (and unknowable).

    Which means… IS-LM is mere bravado. It purports to be a tool by which we can reason about the effects of disequilibria but when it matters: reality and the IS-LM model diverge.

  2. Gravatar of Yehuda Yehuda
    13. April 2009 at 02:13

    I’m confused, which may well mean I missed something.

    You said:”Low rates are not a sign of monetary policy ineffectiveness; rather they are a symptom of excessively tight monetary policy. ”

    I don’t understand this. While its clear that a lowering of interest rates that is smaller then expected means money is tighter then expected, its unclear why a nominal rate of 1% or 0.5% means money is tight, all other things equal.

  3. Gravatar of Bill Woolsey Bill Woolsey
    13. April 2009 at 06:58


    What do you mean by “tight money?”

    If the quantity of money will be smaller than the quantity demanded at the target growth path for nominal income, then nominal income will fall. With sticky prices, real income will fall. If real income is expected to fall, investment demand will fall now, and the level of the real interest rate consistent with real total expenditure equal to productive capacity falls now. Perhaps it turns negative–say -1%.

    Further, the expected drop in nominal income will result in lower expected inflation. Perhaps it could go so far as to turn into expected deflation, say -1%.

    And so, the .5% nominal interest rate could represent a 1.5% real interest rate, while the real interest rate necessary to maintain real expenditure equal to real income has fallen to -1%. Money is “tight.” The real interest rate is 2.5% too high compared to where it should be under current conditions.

    But what caused this to happen? Expectations that the quantity of money will be less than the quantity necessary to keep nominal income equal to expenditures.

    If those expectations change, then investment demand will rise and the expected inflation rate will turn from deflation to say, 2%. Suppose the real interest rate needed to keep total real expenditures equal to capacity rises to 1%. That adds up to a nominal interest rate of 3%. And so, a proper, more expansionary monetary policy results in the nominal interest rate rising from .5% to 3%.

  4. Gravatar of Bill Woolsey Bill Woolsey
    13. April 2009 at 07:00


    “But what caused this to happen? Expectations that the quantity of money will be less than the quantity necessary to keep nominal income equal to expenditures.”

    Expectations that the quantity of money will be less than the quantity demanded when nominal income remains on target.

  5. Gravatar of Noah Yetter Noah Yetter
    13. April 2009 at 08:11

    “…you cannot get real effects from money with a model of perfect wage and price flexibility.”

    Yes you can. Your view is clouded by too much aggregation. It matters HOW money enters the economy. Go back and read the Austrians.

  6. Gravatar of Rafael Rafael
    13. April 2009 at 12:37

    Hi Scott,

    What do you think about following a Taylor Rule-type policy, even if Orphanides says that they are extremely problematic when it comes to using real time data subject to severe bias?

    Furthermore, the zero bound gives rise to multiple equilibria, which make things difficult when it comes to relying blindly on the Taylor principle(see the John Cochrane paper: Inflation Determination with Taylor Rules: A Critical Review).

    Am I too pessimistic about Taylor rules? I like the IS-LM framework because it is intuitive, and, despite the fact that it work reasonably well during the Great Moderation, I don´t think a Taylor rule would work outside the Great Moderation either. If I am wrong, please correct me.


  7. Gravatar of ssumner ssumner
    13. April 2009 at 16:36

    Jon, The shape of the IS-LM curves are hard to estimate, as you say, partly because the curves are affected by expectations in a way that makes it hard to identify the slopes.

    Yehuda, I have a post back in February entitled “What Would Really, Really Tight Money Look Like?” which you might find interesting. I ask what would be the tightest money imaginable? Then I answer that it would be a highly deflationary monetary policy. But whenever governments pursue highly deflationary policies, we end up seeing interest rates fall to very low levels, as Bill explains in his comment.

    Bill, I agree, and the most expansionary monetary policies (Germany 1923) are the most likely to result in high nominal rates. Most people find that extreme case a bit easier to visualize.

    Noah, You are right that the Austrians have a model based on misperceptions, which can generate real effects. I should have said that wage and price stickiness is the most likely factor to generate real effects in the class of rational expectations models that I usually use. I think that aggregation is inevitable in macro, it’s just a question of how you do it.

    I also worry that the Taylor Rule may not be enough. There are two aspects to the Taylor Rule, its policy goal (a mixture of prices and real output–not dissimilar to NGDP) and its backward-looking charateristic. In principle, one could use the same policy goal in a forward-looking way, and i think that would work much better (and would address Orphanides concern.)
    I actually corresponded with Cochrane a bit. I must confess that I don’t understand the indeterminacy issue very well, but don’t think it’s a problem for any policies that I have proposed. He was intrigued by my futures targeting idea, wondered why no one had thought of it before (of course it’s has been around at least since Earl Thompson’s 1982 paper, but most macroeconomists have ignored the idea.)

    I would be careful with “intuitive” models. In my view monetary economics is very counterintuitive. Policymakers in the early 1930s were using an IS-LM approach when they assumed policy was expansionary because rates were low. That error led to the Great Depression. Admittedly a more sophisticated IS-LM analysis might have turned up rising real interest rates, but even that is not necessarily true if expectations for real GDP growth are bearish enough.

  8. Gravatar of Greg Ransom Greg Ransom
    13. April 2009 at 18:38

    Steve writes:

    “the always changing Wicksellian equilibrium rate (i.e. the rate necessary for macro stability.)”

    The crucial question here is — macro stability of what? As Hayek points out, you can attempt to “stabilize” some macro “variable” that are incompatible with overall macro stability, for example, the stabilization of the “price level” in the face of productivity gains will create inevitable macro instability.

    See Hayek on all this — Hayek in particular looks at some of the incompatibilities in Wicksell’s work which were not evident to Wicksell himself.

  9. Gravatar of Greg Ransom Greg Ransom
    13. April 2009 at 18:56

    Steve writes:

    “the always changing Wicksellian equilibrium rate (i.e. the rate necessary for macro stability.)”

    The incompatible accounts of “macro stability” assumed in Wicksell was identified by Hayek in his _Priced and Production_:

    “Wicksell, from the outset, regards the problem as concerning explicitly the average change in the price of goods, which from the theoretical standpoint is quite irrelevant. He starts from the hypothesis that, in the absence of disturbing monetary influences, the average price level must remain unchanged. This assumption is based on another, only incidentally expressed, which is not worked out and which, from the point of view of most of the problems dealt with, is not even permissible; the assumption of a stationary state of the economy. His fundamental thesis is that when the money rate of interest coincides with the natural rate (i.e., that rate which exactly balances the demand for loan capital and the supply of savings) then money bears a completely neutral relationship to the price of goods, and tends neither
    to raise nor to lower it. But, owing to the nature of his
    basic assumptions, this thesis enables him to show deductively
    only that every lag of the money rate behind the natural rate
    must lead to a rise in the general price level, and every increase of the money rate above the natural rate to a fall in general price level. It is only incidentally, in the course of his analysis of the effects on the price level of a money rate of interest differing from the natural rate, that Wicksell touches on the consequences of such a distortion of the natural price formation (made possible by elasticity in the volume of currency) on the development of particular branches of production; and it is this question that is of the most decisive importance to trade cycle theory. If one were to make a systematic attempt to coordinate these ideas into an explanation of the trade cycle (dropping, as is essential, the assumption of the stationary state), a curious contradiction would arise. On the one hand, we are told that the price level remains unaltered when the money rate of interest is the same as the natural rate; and, on the other, that the production
    of capital goods is, at the same time, kept within the limits imposed by the supply of real savings. One need say no more in order to show that there are cases “” certainly all cases of an expanding economy, which are those most relevant to trade cycle theory “” in which the rate of interest that equilibrates the supply of real savings and the demand for capital cannot be the rate of interest that also prevents changes in the price level.59 In this case, stability of the price level presupposes changes in the volume of money; but these changes must always lead to a discrepancy between the amount of real savings and the volume of investment. The rate of interest at which, in an expanding economy, the amount of new money entering circulation is just sufficient to keep the price level stable, is always lower than the rate that would keep
    the amount of available loan capital equal to the amount simultaneously saved by the public; and thus, despite the stability of the price level, it makes possible a development leading away from the equilibrium position. But Wicksell does not recognize here a monetary influence tending, independently of changes in the price level, to break down the equilibrium system of barter economics; so long as the stability of the price level is undisturbed, everything appears to him to be in order.60 Obsessed by the notion that the only aim of monetary theory is to explain those phenomena that cause the value of money to alter, he thinks himself justified in neglecting all deviations of the processes of money economy from those of barter economy, so long as they throw no direct light on the determination of the value of money; and thus he shuts the door on the possibility of a general theory covering all the consequences of the phenomena he indicates. But although his thesis of a direct relationship between movements in the price level and deviations of the money rate of interest from its natural level, holds good only in a stationary state, and is therefore inadequate for an explanation of cyclical fluctuations, his account of the effects of this deviation on the price structure and the development of the various branches of production constitutes the most important basis for any future monetary trade cycle theory. But this future theory, unlike that of Wicksell, will have to examine not movements in the general price level but rather those deviations of particular prices from their equilibrium position that were caused by the monetary factor.”

  10. Gravatar of Greg Ransom Greg Ransom
    13. April 2009 at 19:05

    If you’ve read Hume or Cantillon, you know that this sentence is obviously false:

    “In principle, a doubling of the money supply should immediately double the price level.”

    If you’ve read Hayek, you know that increases in the money supply in fact have to work their way through through the whole structure of production.

    The 19th century Marshallian idea of “stickiness” really has nothing to do with it.

  11. Gravatar of Greg Ransom Greg Ransom
    13. April 2009 at 19:12

    Apologies for the name mix up.

  12. Gravatar of Joe Calhoun Joe Calhoun
    14. April 2009 at 05:41


    This is a little off topic, but I am curious about how you view current monetary policy. I understand you felt it was way too restrictive over the last year, but unless I missed it, you haven’t said what your view is of current policy. Has the Fed at least gotten to an expansionary point now even if not enough to get to a future 5% NGDP? Or are they still way behind the curve? Absent the futures contract you advocate, how will we know when policy has gotten to its proper level? If an IS-LM model won’t tell us, what will? Did I miss something in this post?

  13. Gravatar of Joe Calhoun Joe Calhoun
    14. April 2009 at 05:57


    By the way, I watch the TIPS market as my market indicator of inflation expectations. The spread right now is roughly 1.3% on the ten year. What should be the Fed be targeting? 2%?

  14. Gravatar of ssumner ssumner
    14. April 2009 at 17:47

    Greg, I agree that Wicksell thought in terms of price stability. Both Hayek and myself prefer NGDP rules, but for different reasons. I am not an Austrian economist, although I plan to look at the subject a bit more this summer. Hayek wouldn’t like my 5% NGDP growth rule, unless I am mistaken. But the neutrality of money suggests that differences in the average rate of inflation among policy rules shouldn’t matter (except as a tax on money) what matters is the aggregate being targeted.

    Regarding Hume and Cantillon, they didn’t have rational expectations models. With Ratex, I think you do need stickiness, although I am a pragmatist and have an open mind on this issue. I am not ruling out misperceptions or other factors as possibly contributing to short run disequilibrium.

    Joe, We are still way too contractionary, with inflation expectations at below 1% over five years. It needs to be closer to 2%. And NGDP growth expectations are also very low, although a bit harder to estimate. I usually add the consensus growth estimate to the TIPS spread to get a rough NGDP estimate. It needs to be 5%.

  15. Gravatar of Greg Ransom Greg Ransom
    14. April 2009 at 19:25

    Scott — It’s a straight shot from Hayek’s Hume/Cantillon model of money caused market-wide disequilibrium through the structure of production to the whole literature on the limits and division of knowledge — but not only that, this stuff gave Hayek the idea for the key building blocks of the modern intertemporal equilibrium construction itself.

    As Hayek says, you have to abstract away from the key problems and facts of economic science to model around this stuff via “rational expectations” and macroeconomic aggregates. To provide a valid explanatory framework that takes into account these core facts and problems of economics is impossible using modeling “tools” and maths that exclude the most central economics facts and economic problems by formal necessity.

    Scott writes:

    “Regarding Hume and Cantillon, they didn’t have rational expectations models. With Ratex, I think you do need stickiness, although I am a pragmatist and have an open mind on this issue. I am not ruling out misperceptions or other factors as possibly contributing to short run disequilibrium.”

  16. Gravatar of Greg Ransom Greg Ransom
    14. April 2009 at 19:29

    As a follow-up on Hume, Cantillon & Hayek, the point is that the “misperceptions” issue is a microeconomic issue and a production coordination issue as discussed in Hayek’s famous essays “Economics & Knowledge”, “The Use of Knowledge in Society” and his essays on socialism and capital production, and elsewhere. “Ratex” is a modeling choice that utterly fails to address these core economic facts and core economic problems of economic science.

    Scott writes:

    “Regarding Hume and Cantillon, they didn’t have rational expectations models. With Ratex, I think you do need stickiness, although I am a pragmatist and have an open mind on this issue. I am not ruling out misperceptions or other factors as possibly contributing to short run disequilibrium.”

  17. Gravatar of ssumner ssumner
    15. April 2009 at 18:03

    Greg, Obviously I may have overlooked important perspectives, but all I can do is adopt the modeling strategy that seems most effective to me (again using pragmatic criteria.) I have focused on a few key principles:

    1. Wage stickiness
    2. NGDP instability as a source of cycles
    3. Monetary rules aimed at NGDP stability
    4 Forward-looking monetary rules
    5. Ratex and efficient markets

    As I said in my defense of efficient markets a few weeks back, I have yet to find a single anti-EMH or anti-Ratex theory that has any useful implications. If someone finds one, I’ll gladly look at it. I’m not an ideologue as should be obvious from the fact that my list above does not adhere to any single doctrine. I accept certain ideas from each of the following schools:

    1. New Keynesian (wage stickiness)
    2. Monetarism (excess cash balance mechanism, policy rules)
    3. New classical (Ratex, EMH)
    4. New Monetary Economics of the early 1980s (futures targeting, price of money approach)
    5. Austrian/Hayekian (NGDP rules, distrust of structural models of the economy, preference for market-based policies.)

    If you can convince me that Austrian/Hayekian economics explains recent events better than my hodgepodge theory fine. But right now I see NGDP fluctuations causing huge problems for the economy, and I think we have the tools to prevent NGDP fluctuations, but aren’t using them. Maybe even with a stable NGDP, there would be all sorts of problems related to overinvestment, misperceptions, etc. But we won’t know that until we get stable NGDP.

  18. Gravatar of Greg Ransom Greg Ransom
    15. April 2009 at 22:03

    Scott, the crux of the problem is the issue of explanation — what can provide an explanation, and what can’t, what we can hope for in an explanation, and what we cannot hope for.

    These are hard issues — no doubt too hard for a blog.

    You want a math/logical “model” to do the explaining.

    Hayek offers an alternative _causal_ picture of economic explanation, involving real world entrepreneurial learning as a cause that can’t be “modeled” in a “God’s eye view” math or logical construct. Think Knightian uncertainty to image _why_ this cause can’t be modeled in a “God’s eye view” math or logical construct.

    These are radically different world views about what economics can do and how economics explains.

    Again. Too much for a blog.

  19. Gravatar of Greg Ransom Greg Ransom
    15. April 2009 at 22:11

    2. “NGDP instability as a source of cycles”

    Here we have some common ground. Why is NGDP unstable?

    Well, because money and credit and perhaps other factors systematically distort the time dimension of production, involving also individual mistakes in investment, savings, and consumption across time.

    Can we imagine this distortion and illustrate it in marginal valuation terms? To a degree.

    Can we build a mathematically tractable equilibrium “model” of this disequilibrium phenomena? There is no reason at all to think so. Nor is there a clear reason why we should think so, or what it must be necessary to do “science” or to produce a powerful causal explanation.

  20. Gravatar of Greg Ransom Greg Ransom
    15. April 2009 at 22:12

    Make that a “WHY” and not a “what” in the following:

    “Can we build a mathematically tractable equilibrium “model” of this disequilibrium phenomena? There is no reason at all to think so. Nor is there a clear reason why we should think so, or WHY it must be necessary to do “science” or to produce a powerful causal explanation.”

  21. Gravatar of ssumner ssumner
    16. April 2009 at 17:48

    Greg, We are not as far apart as you might imagine. Let me separate out two issues:

    1. Why does NGDP fluctuate?
    2. Why do NGDP fluctuations have real effects?

    On the second question I mentioned wage stickiness, but I imagine there are many reasons, including some the Austrians have identified that I am not even aware of. But if we both think NGDP instability is a problem, we don’t necessarily need to worry if we don’t think it is a problem for exactly the same reason.

    On the first question, I have a very non-structural modeling answer. I don’t think economists should set up mathematical models of the economy and use them to set monetary policy instruments. Indeed, one of my papers (with Aaron Jackson) had “Is structural modeling necessary?” as its subtitle, and answered “no.” (Economic Inquiry , 2006.)

    Instead, I favor establishing the monetary policy goal (stable NGDP growth) and let the market decide how much money is necessary to hit that goal (by essentially using NGDP futures contracts as a sort of gold standard–pegging the price of those contracts.) I don’t see why Austrians would find that approach highly mathematical.
    My views may not seem that way in this blog, as I often give the Fed advice on interest rate targets, penalties, etc. But it’s their choice to use those flawed structural models, and that flawed discretionary approach. I can’t stop them. If they insist on doing that I will give them advice on how to avoid another depression, with whatever tool they choose to use. The last depression set back free market economics by decades.

  22. Gravatar of Ivan Ivan
    28. February 2011 at 06:16

    May I ask you why economists authors of textbooks on intermediate macroeconomics like you keep using the IS-LM model even though we already know that the Central Bank does not set the monetary supply. Instead, it does set the interest rate. Shouldn´t you do like Wendy Carlin and David Soskice in their recent and fantastic book “Macroeconomics: Imperfections, Institutions and Policies” where they replace the LM curve by a monetary rule (for example, a Taylor rule). Wouldn´t that be more representative of what occurs in reality rather than supposing that the institution gets the control of the quantity of money?

  23. Gravatar of ssumner ssumner
    1. March 2011 at 06:30

    Ivan, I never use IS-LM, and indeed did a whole post trashing the model. I’ve never taught IS-LM, and have never written a textbook.

  24. Gravatar of TheMoneyIllusion » Nick Rowe presents the first plausible IS-LM model. TheMoneyIllusion » Nick Rowe presents the first plausible IS-LM model.
    9. September 2011 at 18:57

    […] ago Brad DeLong did a post asking me to explain why I didn’t like IS-LM.  I wrote one of my typical ultra-long early posts, throwing everything against it I could think of.  But I so dislike IS-LM that I never learned […]

  25. Gravatar of MTD MTD
    10. September 2011 at 09:47

    Scott – If BB came out and said instead of Operation Twisted, the Fed would adjust current and future base money by any volume necessary to support sustained inflation break-even TIPS spreads between 300-400 bps, I think we’d be one step closer to your NGDP level targeting idea. We need to get V going. This may do the trick, no?

  26. Gravatar of ssumner ssumner
    12. September 2011 at 14:21

    That would be even more than we need, especially if you are talking about 5 year TIPS or less.

  27. Gravatar of TheMoneyIllusion » Links to my views on money/macro TheMoneyIllusion » Links to my views on money/macro
    23. September 2011 at 04:50

    […] And why I don’t like the IS-LM approach […]

  28. Gravatar of Why I do not like the IS-LM model — Marginal Revolution Why I do not like the IS-LM model — Marginal Revolution
    4. October 2011 at 00:27

    […] is why Scott Sumner does not like IS-LM.  After writing this post, I remember I had an earlier 2005 post on why I do not like IS-LM; I […]

  29. Gravatar of On IS-LM «  Modeled Behavior On IS-LM «  Modeled Behavior
    4. October 2011 at 11:52

    […] and Scott are not fans. Brad […]

  30. Gravatar of IS-LMが嫌いな理由 by Scott Sumner – 道草 IS-LMが嫌いな理由 by Scott Sumner – 道草
    9. October 2011 at 00:48

    […] Scott Sumner // またサムナーの古いエントリから、まだ熱かった”Why I don’t like IS-LM (reply to Delong) “(12. April 2009) […]

  31. Gravatar of More On the Sumnerian Holy War Against Fiscal Stimulus | Last Men and OverMen More On the Sumnerian Holy War Against Fiscal Stimulus | Last Men and OverMen
    21. February 2017 at 04:40

    […] this is the real reason Sumner doesn’t like the IS-LM model.        http://www.themoneyillusion.com/?p=915       Sumner often construes his problem with fiscal stimulus in terms of that […]

  32. Gravatar of On Friedman's Keynesianism Sumner Just Can't Let it Go | Last Men and OverMen On Friedman's Keynesianism Sumner Just Can't Let it Go | Last Men and OverMen
    18. March 2017 at 06:45

    […]    http://www.themoneyillusion.com/?p=915 […]

  33. Gravatar of Is There Something Wrong With Market Monetarism | Last Men and OverMen Is There Something Wrong With Market Monetarism | Last Men and OverMen
    18. April 2017 at 01:10

    […] German data, so I’m not sure exactly what happened to productivity.”      http://www.themoneyillusion.com/?p=915      Those labor reforms were wonderful Sumner is at pains to tell us. Can you […]

Leave a Reply