The ISLM takes quite a beating from Post-Keynesians — and, I would argue, rightly so. There are any number of reasons for this but let me just here highlight one that is not very regularly talked about.
As is well-known and can be seen in the below diagram the ISLM considers output to be a function of the interest rate. At a higher level of interest rates output is thought to be lower and at a lower level of interest rates output is thought to be higher.
The problem with this presentation? It is not true. You see, even if we allow that interest rates have a substantial effect on output, it is not so much the absolute levels of interest rates that matters so much as it is the relative rate. Relative to what? To itself of course. What I mean is that if interest rates have effects on output it is the change in the interest rate rather than the absolute level that leads to expansions or contractions in output.
This can clearly be seen by simply looking at data for a wide variety of countries. If the ISLM were correct we would assume that countries with high interest rates would have low output, but this is simply not the case. Take Brazil as an example. Their interest rate has been between 18% and 8% since 2006,
From the perspective of the ISLM these extremely high interest rates should translate into low output growth, but this is simply not the case. Throughout the period — barring an interruption by the worldwide recession in 2008 — output growth in Brazil has been fairly high, running between about 2% and 9% annually,
This is not simply a case of the nominal interest rate being high while the real interest rate is low either. If we look at the inflation rate over this period it is not particularly high at all — at least, for a developing country — and has proved relatively stable,
I would say that over this period we are seeing an average interest rate of about 12% or so and an average rate of inflation of about 5% or so. Real interest rates, on average, then are about 7% — not to be sniffed at — and yet GDP growth averaged maybe 4-5% if we control for the recession.
The lesson here is obvious: it is not the absolute level of the interest rate that has an effect on output but rather the change in the interest rate. Large changes can cause expansions and contractions in output, but the effects of the absolute level is far less clear.
In this regard, the ISLM — even if we take it on its own terms — is somewhat misleading. The relationship between interest rates and output is not a mechanical one that can be represented in two-dimensions, rather it is one that rests on changes taking place in the variables.
Some will now say: “Oh, but we’re not stupid… everyone already knows that…”. Actually, I think that this is far from clear. When economists — like Krugman, for example — talk about a natural rate of interest that would result in full employment they tend to talk in absolute terms; “The natural rate is –x%…”.
Now, I obviously do not believe in this natural rate but even if I did I would point out that it is the rate at which the interest rate is changed from the present rate to the natural rate that is important, not the absolute level of the natural rate per se. Even if the so-called natural rate were assumed to function it would likely not do so if the interest rate was lowered very gradually to this rate, over the course of, say, 5 years. Rather it would have to be done rather quickly; maybe over the course of a few weeks. Central banks are fully aware of this, of course, which is why they always try to time their interest rate changes precisely so as to ensure the intended effects.
So, the ISLM framework, with its static ideas about interest rates and output, does indeed lead to confusion among those who use it. One more reason, among many, to throw it out.
Fixing the Economists 
I think you overstate the case here. Just because there are positive correlations between growth and interest rates, this does not contradict the IS-LM model. Why are you forgetting about the IS curve? High output and high interest rates can arise from outward shifts in IS. If you did some analysis that dealt with the Granger causation (is it high growth causing high interest rates or vice versa) then maybe you’d have a point (but even then forward looking monetary policy or financial markets could rationalize it). Moreover the static vs. dynamic argument seems mixed up unless I am misreading you. The IS LM model says that expansionary monetary policy all else equal would lead in a “static” sense to higher output per period, odd to discuss a “static” flow measure I know, (this is the low rates and high growth you are trying to refute), but then you talk about growth, which in the IS LM model still implies that it is the change in monetary policy (a shift in LM) that causes growth (a change in output), in this way the IS-LM model agrees with one of the points you are trying to make to refute it. I just don’t think the evidence you have offered either empirically or logically undermine the IS-LM model in any way.
The evidence is not perfect, but it never is in economics. I think it is quite clear that high interest rates are often accompanied by growth while low interest rates are often accompanied by stagnation (the Eurozone, the US, Japan, the UK and so on…). This intuitively refutes the ISLM framework which would lead to the opposite conclusion. Yes, you can add in ‘ifs’ and ‘buts’, but these are largely ad hoc.
As to my other point, no you have not understood it. I’m saying that the rate at which the interest rate is shifted is more important than the level at which it is set. So, the rate of change of the interest rate is more important than the level of the interest rate per se. This is not captured in the framework and thus the framework misses how monetary policy really works. (I.e. it mainly works via expectations). In the model it is implied that we are dealing with absolute levels of interest rates, not rates of change of the interest rate. This can clearly be seen in the statements of those that use it — e.g. Krugman saying that the full employment rate is at x%, which he has taken from a central bank estimate. You can reinterpret the meaning of the model in the here and now, but again what you’re doing is ad hoc.
You can interpret and reinterpret the ISLM in as many ways as you’d like to get around these obvious shortcomings. But the other people using it will be using it in the way that I’m saying because that is how it is presented in textbooks etc. Also, such reinterpretations have the smell of ad hoc pseudoscience about them; i.e. the ISLM can be tortured to the point that it can be made to say ANYTHING. I am fairly confident that what I have presented above and in my comment is the consensus view and it is that which I am interested in criticising.
I’ll just try to restate my two main points one last time in response to your reply. A positive correlation between interest rates and output DOES NOT refute the IS LM model in any way. The IS LM model says this correlation depends on the source of the shock, if on the LM side then its negative, if on the IS side then its positive. These aren’t ifs and buts these are the basic workings of the model. You can say the interpretation is ad hoc, but there are testable implications if you can identify the “exogenous” monetary or IS shocks. Does expansionary monetary policy lower interest rates and raise growth compared to what they would otherwise be? This is a fundamental research question that can be falsified.
To the second point, you say that it is the change in the interest rate which matters rather than the level, but for what? If for growth (as you are discussing in your article) then it is the change in interest rates, that is also what the IS LM model says. This is not inconsistent with Krugman’s statement then that what matters for the level of output and employment is the level of the interest rate. Further, to go back to the first point, you say that decreasing the interest rate faster should be associated with faster growth, if you subject your hypothesis to the same test you did for the level, you will, it seems, similarly find that quickly growing countries often have quickly growing interest rates (and countries falling into recession have falling interest rates). You wouldn’t say this correlation refutes your causal hypothesis per se (and I agree fully that expectations are a major aspect of monetary policies effect), but you aren’t controlling for anything else and are running into reverse causation problems.
I enjoy your postings though, keep up the good work.
Okay, I’ll go step-by-step.
(1) What do you think causes the positive correlation between output and interest rates in countries like Brazil? You say it might be a shift in the IS-curve, but be more specific. Is it government spending shifting? Give me something concrete here because I think you’re mystifying this somewhat.
(2) I didn’t say that the rate of change of the interest rate determined growth. I said that monetary policy is used in this way. I am totally agnostic as to whether this works or not. But it is a more accurate description of how monetary policy actually works than simply assuming a level.
(3) You write:
This is just wrong. Krugman and other ISLMists say, for example, that the full employment interest rate in the economy is currently -5%. That is a figure that he gets from Federal Reserve economists.
Now, I think that figure is bogus for a number of different reasons. But let’s put those aside for now and take the argument at face value. If the interest rate was lowered from 0.25% to -5% in a week would it have different effects than if it was lowered to the same level over the course of, say, 8 years? Of course it would. But the ISLMists never mention this.
This is further reflected in the current arguments of trying to generate inflation to create a negative rate. Again, I think that this cannot be done by the central bank (unless they crash the currency) but let’s leave that aside. It is never specified how quickly that inflation would be generated. If it was generated overnight it would have vastly different effects than if it was generated over the course of 4 years. But these considerations are just thrown by the wayside because the ISLM stifles thinking on such matters.
This is because the ISLM as is commonly conceived (yes, you can create an idiosyncratic interpretation, but this is not consensus) suggests that it is the level that matters. Not the rate of change. This is because the ISLM is static. It does not integrate time. It is based on static equilibrium. It cannot depict time at all. That is the key criticism of both the above and of the post I wrote today.