Tuesday, 28 February 2012

Why introductory macroeconomics should ditch the LM curve

For teachers and students of economics

                For the next two years I will be taking a break from teaching tutorials in first year undergraduate macro. That will be a relief for just one reason. I have become more and more embarrassed at having to teach the IS-LM model. The IS curve is fine, but the LM curve is not. The reason is obvious enough: central banks do not operate a fixed money supply policy. It would be nice to tell students that the fiction that the monetary authorities fix money is a harmless fairy story, but I do not believe this. Here are just three mistakes or confusions caused by using the LM curve.

1) Is that real or nominal rates on the vertical axis? It does not matter if expected inflation is constant, but we use the apparatus to look at price changes.

2) IS-LM leads to AS/AD. If your goal is controlling inflation, the AS curve suggests that all you need to do is to return to somewhere on the vertical AS curve. If you use a traditional Phillips curve with backward looking expectations then returning to the natural rate will not be enough. Which is the student supposed to believe?

3) IS-LM leads to textbook Mundell Fleming, which in its simplest form has domestic interest rates stuck at world levels. Everyone then learns that fiscal policy is ineffective under floating rates, which is unfortunate, because this is a special consequence of assuming fixed money. The IS curve plus UIP is a much better way to think about this, and gives a more relevant answer: a temporary demand shift is not crowded out by an appreciating real exchange rate if domestic interest rates are unchanged.

                I’m currently lecturing on Oxford’s second year undergraduate macro course. There I largely ignore the LM curve, and (following the textbook by Carlin & Soskice) instead teach a three equation model involving an IS curve, a Phillips curve, and a monetary rule. The monetary rule captures the idea that the monetary authority has preferences over excess inflation and the output gap, and combining this with the IS curve and Phillips curve we can derive a Taylor rule. While the monetary rule curve is far from ideal, representing the result of a static rather than dynamic optimisation exercise, it at least captures the spirit of what central banks try and do today.
The introductory course in macro at Oxford starts with IS-LM, but it also includes the Phillips curve and the Taylor rule, presumably because they are more relevant than the AS curve and the LM curve. So first year students learn about the AS curve and the Phillips curve, the LM curve and a Taylor rule. Perhaps not surprisingly, many get very confused. It would seem to make much more sense to switch things around. Use only the IS curve, Phillips curve and monetary/Taylor rule in a first year introductory course, and only introduce the LM curve in subsequent years. (I would introduce the LM curve in stages. First, think about replacing an inflation target with a price level target for monetary policy. Maybe think about nominal GDP targets, using the current situation as clear motivation. Then go to money supply targets, with of course some discussion of money demand.)
                This sequencing seems clearly preferable, so I find it puzzling that most Principles textbooks do not take this approach, but instead continue to start with IS-LM. Why are they so devoted to the LM curve? Is learning about the LM curve meant to impart some deep wisdom to students? Is there a feeling that because money is important in understanding how business cycles and inflation work, it should be introduced as part of the core model come what may? This seems strange, and is probably counterproductive. Perhaps the explanation is more mundane: that the economics of high volume textbook publishing is such that innovation is difficult to do. 
I sometimes say to master’s students just starting the core macro course that they will spend some time learning about the same stuff as they did when they were undergraduates: inflation, unemployment, business cycles. The key difference is that what they learn will only be 5 or 10 years out of date, compared to material that is 30 years out of date for undergraduates. Strange, but unfortunately true.

20 comments:

  1. Great post!

    BTW, what do you think of the Carlin-Soskice's textbook? Are there better ones out there?

    ReplyDelete
  2. good post. i am an applied microeconomist currently teaching second year macro, trying to keep stuff as close to reality as possible. upon reading your post, i ordered an inspection copy of Carlin-Soskice, maybe will be able to switc to it next time i cover this.

    ReplyDelete
  3. As a recent PPE grad (and a big fan of the Carlin & Soskice book used in the second year course), I was wondering whether you ever considered Tyler Cowen's and Alex Tabarrok's Principles book (http://worthpublishers.com/Catalog/product/modernprinciplesmacroeconomics-secondedition-cowen)?

    ReplyDelete
  4. On the right hand side of your blog, you say it's written for both economists and non-economists.

    Pro-tip: if you're really aiming to reach non-economists, it would be good to define terms like "IS curve" and "LM curve" before you start to say what is wrong with them. Non-economists probably have no idea what you're talking about.

    ReplyDelete
    Replies
    1. Pro/Am-tip: Read your own quote "It's written for economists and non-economists". Clearly this particular blog post was for economists. Non-economists look here: http://en.wikipedia.org/wiki/IS/LM_model

      Delete
  5. I'm in two minds about this really. I agree that IS/LM leaves something to be desired, but I don't really like the alternative approaches either. I especially don't like adding a Taylor rule as if the CB automatically acts in a certain way, I think that really is confusing for students. It's one thing to say, "if output goes up with the money supply fixed, interest rates will have to rise to maintain equilibrium in the money market." It's quite another to say, "if output goes above the natural rate interest rates will rise because the CB follows a Taylor rule". The first is a statement about the impersonal forces that ensure market equilibrium is achieved, the second is a prediction about the future behaviour of a specific group of people. If the CB's decisions were that straightforward we could replace the MPC with a computer (which may not be a bad idea...).

    I'm not sure the fact that LM is unrealistic is much of a criticism: show me a model in economics that is realistic! The idea that people form rational expectations is patently unrealistic, for instance. Sometimes it's useful to make unrealistic simplifying assumptions, such as assuming the money supply is fixed, expectations are fixed etc. I think this is a useful starting point, and having grasped what would happen under those conditions one can move on to consider the implications of more complicated assumptions.

    ReplyDelete
  6. Have a look at the subtitle! I try to remember to add this when I want just to talk to economists.

    ReplyDelete
  7. Blanchard's book has an LM curve yet also addresses all three of your questions. Of course, you have to get past the first chapters.

    ReplyDelete
  8. I agree with "econjon". I don't like the LM curve, but I don't like its alternatives either. In my experience, some undergraduates can have trouble distinguishing the "natural" operation of the economy from policy interventions designed to influence it, endogenous from exogenous variables. So I'm not as keen on monetary policy rules as I might be. As for UIP, this isn't easy to apply because it depends on expectations about the exchange rate and the duration of interest rate deviations, and these aren't at all easy to pin down.

    The LM curve approach could, maybe, have a couple of advantages. First, it's applicable to economies in which the central bank does not pursue an active policy, or does not pursue a systematic policy. Second, it focuses attention on the idea that financial markets can influence the economy, and must also be in equilibrium. A related point is that it provides a natural framework for talking about *how* the central bank influences interest rates, e.g. through open market operations. I'd be curious as to whether you think the LM curve's lessons about financial markets and the implementation of policy are wrong and misleading, or merely of secondary importance and not worth the trouble.

    ReplyDelete
  9. A problem with textbook Taylor rules is that they miss what appears to be the main determinant of policy (in the US) and that is as an asset price stabilization tool.

    ReplyDelete
  10. The best undergraduate macroeconomics book is that by S.Williamson.

    The one by Miller and Upton, although dated, is also excellent.

    ReplyDelete
    Replies
    1. I think the macro textbook by Max Gillman is far superior to that of Williamson.

      Delete
  11. I absolutely totally agree. There is another problem with the LM curve. Back in the 70s when it was current macro research it fit the data very well (if one added an exponential trend in velocity and some micro nonfounded stock adjustment/ error correction dynamics). Then it failed completely in the early 80s. By then research had moved on, but a model which fit well before 1980 and very poorly since then is not ideal. I believe that this is the reason that monetary authorities switched back from targetting monetary aggregates to targetting interests rates after a monetarist experiment which lasted a few years in the USA.

    I note that Brad DeLong has banished the LM curve from his textbook.

    On point 2, I would go with the model which fits the historical data better, that is the one with a backwards looking Phillips curve. Models with vertical AS curves have nothing to do with actual economies, so I think they can and should be ignored.

    ReplyDelete
  12. I agree with Anonymous, the best undergraduate macroeconomics book is that by S.Williamson. I'm a second year student,studying Macro, IS/LM curves help me to understand real economic situation on Us example. I like the way lecturers teach student macro in Georgia.
    http://bestbuycreditcardonline.com/www-hrsaccount-com-bestbuy-pay-bill

    ReplyDelete
  13. For an alternative to IS-LM that is simpler than Carlin-Soskice, look at the papers with "Wicksell" or "Minsky" in the title here: http://www.gettysburg.edu/academics/economics/char_weisehomepage/charles_weise.dot

    ReplyDelete
  14. I'm a wierd political economist who learnt his economics from textbooks after (too) many courses on the philosophy of the social sciences. I still don't understand why economics students aren't forced to do a little bit more of this. Really good economists do seem to know what 'models' are and what they mean....but I get the impression undergrads are rarely encouraged to think in this way....'what assumptions is this model making?' 'which as most likely to be wrong?' 'In what ways might the model therefore bias my findings' 'Is it still a good model anyway?' I mostly teach politics. That's what we do with models all the time (though the models are generally much simpler, so I guess there's more space to do the questioning).

    ReplyDelete
    Replies
    1. I agree fully with Anonymous here.
      I am an undergrad studying economics and have just been going through the Mundell-Fleming model and I have had to constantly remind myself that this model is a way to explain short term determination of output, the interest rate and the exchange rate in an open economy under specific assumptions and conditions. This are JUST an economic models, they are NOT laws.
      Economics education should probably just focus on teaching students how to understand whats going on in the economy and how to build models to describe that and how to appreciate the endogenous and exogenous variables in them before attempting to make policy changes. And most importantly how to troubleshoot when things don't seem to work..

      I wish the whole profession can just come out and admit that the current fashionable model cannot be applied in the next crisis and that most of the time we do not know what the hell is going on.

      Delete
  15. As an ex Oxford PPEist I found the Carlin-Soskice model you taught us quite intuitive, more so than the LM curve. Not easier or simpler, but the idea of a CB having preferences and trying to maximise these chimes in well.

    However, I'm quite glad I was taught ISLM in first year. For better or worse other people still talk about ISLM and it's nice to know what they are on about. In addition being taught a basic model early on and then dumping it keeps student from becoming too attached to a model. It makes you question why we aren't using it so you getter a better grasp of its assumptions and why they may or may not be applicable.

    I think that application of critical thought to models was one of the relative strengths of the way Oxford taught economics at undergraduate, which also partly comes about because we had to write essays rather than do problem sheets. It was nice when you could dismiss a friend at another university's concerns with phrases like "yeah but who targets the money supply" or "you do know that you're assuming adaptive expectations". This neatly covered the deficiencies of our fading maths skills.

    ReplyDelete
  16. Prof. Wren-Lewis,
    I've read a new version of the textbook by Carlin and Soskice is forthcoming. It seems the authors weren't very happy about with the outcome after the bursting of the crisis, which rendered the "old" New-Keynesian (no financial issues) Macro useless.
    The new edition appears to bring those new insights into their models.
    It will for sure be worth taking a look at.

    ReplyDelete
  17. my understanding was that IS-LM is a model made to analyze an economy with a lot of slack, an economy either in or on the verge of deflation and depression, at which point policy rules are irrelevant (the Taylor rule has not been relevant to US policy since 2008 or whenever it was), inflation is non-existent, etc..

    I don't understand the supposed empirical relevance of Romer's model.

    Having said that I don't pay much attention to IS-LM. I find there is far too much that is non-explicit about it, and it is far too confusing for students.

    If you're going to use models that employ a lot of hand-waving, good old Keynesian cross is more than enough. What exactly does the IS-LM say that could not be hand-waved from a Keynesian cross?

    ReplyDelete