Why I do not like the IS-LM model

1. It fudges the distinction between real and nominal interest rates, so it can put the two curves on the same graph.  Every time you write down an IS-LM model you should hear a clock start ticking in your head.  The longer the clock ticks, you more you need to worry about this problem because the more that a) the price level may change, or b) expectations about future price level changes will start to matter.

2. It fudges the distinction between short-term interest rates (for the money market curve) and long-term interest rates (a determinant of investment).  They’re not the same!  Don’t assume they are the same, just to squash the two curves onto the same graph.

3. It leads you to think that the distinction between non-interest bearing currency and short-term interest-bearing securities is a critical wedge for the economy.  It also implies that if all currency paid interest (a minor change, most likely, macroeconomically speaking), the economy would behave in a totally screwy way.  It probably wouldn’t.

3b. The model leads you to believe that interest rates are more important than they probably are.

3c. For a while it treats “money” as the non-interest-bearing security, and then for a while it treats money as the transactions media behind AD, something closer to M2.

4. It overemphasizes flows and under-emphasizes stocks of wealth.  The quantity theory approach, as wielded by Fisher and Friedman, does not induce individuals to make this same judgment.  For one thing, this distinction really matters when you’re trying to predict the macro effects of “window breaking.”  The flows perspective will usually be more optimistic than a perspective which recognizes both stocks and flows.

5. Those aggregate curves are not invariant with respect to expectations, including expectations of government policy.  You don’t have to believe in an extreme version of the Lucas critique to worry about this one.  Those curves are conditional and the ceteris paribus assumption is not to be taken lightly here.

6. In the LM curve, what is the embedded reaction function of the Fed?  Good luck with that one.  Pondering this issue leads you to conclude that the whole model was written for an economy fundamentally different than ours.

7. The most important points, for instance about the significance of AD, one can derive from a quantity theory or nominal gdp perspective (for the latter, see my Principles text with Alex).

Why take on all that extra baggage?

Here 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 didn’t like it back then either.

Addendum: For a few sources, here is Roubini on IS-LM.  Here is Wikipedia.


and Charles Goodhart on same : The Continuing Muddles of Monetary Theory: A Steadfast Refusal to Face Facts (PDF here : http://www.boeckler.de/pdf/v_2008_10_31_goodhart.pdf)

Oh...but it's so "clean" and "neat"...even undergrads can understand it.

Thanks for this, Tyler. It's always good to see prof's that are willing to tear apart one of the most widely taught models.

Here is Krugman's explanation of why IS-LM is used despite its known faults. He's way ahead of Tyler.


I notice that Tyler says what he doesn't like, but mentions no substitute. A typical dog-whistle to his "base".

I was going to make a similar comment, but sure wasn't going to say Krugman is in the ballpark. In fact, people who expect too much from models are the reason that models get too much blame for not being reality.

And the alternative to using models is....? Libertarian slogans? Marketz Rulz! Or a great air of disdain? Yes, that gives us lots of guidance -- not!

No, you just don't say something like "Well, suppose you wanted a first-pass framework for thinking coherently about macro-type issues..." and then skip right from that humble description to it being good enough to crank ginormous dials attempting to steer the entire economy- breaking a crap ton of windows.

And anytime someone murmurs 'I'm not entirely comfortable with that' respond with some version of "noone knows any macro!!!"

Don't do THAT.

I notice that Krugman regularly is able to explain his predictions with models, but that Tyler Cowen seldom makes such predictions. Consequently, Cowan has very little record, and Krugman has many correct predictions.

It seems to me that scoffing at models does not produce as many results as using the models intelligently.

Almost all economics starts with flawed assumptions that enable enough simplification for the problems to be tractable. The question is when the resultant models tend to give good results. A perfect but intractable model doesn't help: it doesn't give any results.

Your response here to the limitations of models reminds me of how a medieval medic might have reacted to the suggestion that we stop bleeding patients. Could you even consider the possibility that our economic "science" is on the same level as medieval medical "science", particularly regarding our ability to influence outcomes in a beneficial way and the certainty thereof?

I also hate it when bloggers don't spell out their entire worldview in every single blog post.

Seriously, if you want Tyler's views in macroeconomics, he has about 8 years worth of posts, many of which deal with his views on macroeconomics. If you need an alternative way of looking at the macroeconomy, feel free to read some of them.

"Pondering this issue leads you to conclude that the whole model was written for an economy fundamentally different than ours."

An economy under a gold standard perhaps.

Good post Tyler. Some points:

- It completely ignores irreducible uncertainty, which was central to the analysis it supposedly represented.

- It doesn't include a banking system.

- It treats the rate of interest as a dependent variable determined by S, I, liquidity preference etc., rather than an independent variable that determines these things (which is what it is).

+1 Cahal.

Agree with Tyler and you on all point, but would expand to say that the banking system is so important to macro, yet macro-economic models fail to properly consider the dynamics of banking.

The interest rate is a market price, which is determined simultaneously by S and I and all the rest. You're not really qualified to speak about economics, are you?

For a while it treats “money” as the non-interest-bearing security, and then for a while it treats money as the transactions media behind AD, something closer to M2.

A banknote is non-interest-bearing when it circulates, when it's not deposited in a bank, when it's money.

A banknote deposted in a bank earns interest, because the note then is not money. It is a share of the collateral backing the bank's loans. The bank collects rent on this collateral, and the bank pays dividends to depositors from this rent, after subtracting the costs of operating the bank and the dividends it pays to the bank's shareholders. The bank's shareholders own assets other than the collateral backing loans, like the bank's buildings and other property.

In other words, money never bears interest, because banknotes are money only when they're circulating, and circulating notes don't bear interest. The interest rate on money is always zero, because money is an accounting device with no intrinsic value (no marginal value). Any other impression is an illusion, an accounting fiction.

While many modern macro sorts, including "austrians," are contemptuous of large-scale macro models, such models have their uses.

The large models do make a distinction between real and nominal rates, have multiple rates short and long, and have credit spreads. They have multiple investment sectors with different elasticities. They include stock prices and real estate values. They have capital stocks. They have household wealth. They have tax effects. Some models have liquity constraints.

They have expectations, although typically they are adaptive expectations. There are fancier models that use rational expectations.

If your issue is that IS-LM is too simple, the solution is a bigger model.

If the issue is that the foundations are flawed, that is different. IS-LM is still Keynesian. The model does not handle intersectoral structural shifts and cannot explain trend technology advance.

Most of these criticisms apply with more force to the Quantity Theory which Tyler seems to endorse. He has no other model to offer. Every model makes assumptions. The question is can you learn something from using it.

I learned a version of IS-LM from Thomas Sargent, say, about 1971, which easily incorporated a simple form of expected inflation, and hence the real rate of interest, and a flexible price level. Wages were stuck, not prices. I'm guessing financial effects could be incorporated by recognizing different risk premia.

But can't we learn something from a critique of the model too? When IS-LM material was presented to me at PSU, I don't remember any of these points even being mentioned. Surely the distinction about the interest rates would have been useful to know at the very least.

the nominal/real distinction is easily addressed (see Frank) by explicitly noting that the interest rate is nominal, and then shifting the IS curve up (in the theory, a function of the real interest rate) by the level of expected inflation.
price movements push the LM curve around; you can move it as slowly or as quickly as you see fit.
you can incorporate a financial friction (and financial market disturbances) into the model, see Michael Woodford's JEP piece
the point about investment and the real interest rate is well-taken (a weak link in practice). You could make the IS shift around via some other investment-determining factor, such as optimism or a noisy signal about good times ahead (which may turn out to be false).
you must make consumption and investment forward looking, which means that - while you can take a disturbance and then solve what happens to the economy and then (after that) display IS and LM shifts - you cannot actually solve it via simple curve shifts in the diagram; Bob King has a nice discussion in the Richmond Fed Quarterly (or whatever it is called).
you can re-do the IS-LM by making the LM the Fed reaction curve. the Fed will react to inflation/expected inflation ...
I do all of this in my undergrad course in international finance.
Basically: the typical simplified textbook IS-LM is quite misleading. It takes a lot of work to get the IS-LM in shape for proper analysis of economic events. Not sure there is a good alternative at present.

Interesting comments, thanks for sharing.

It also implies that if all currency paid interest (a minor change, most likely, macroeconomically speaking), the economy would behave in a totally screwy way. It probably wouldn’t.

Really, now? I've long had this idea to issue zero-coupon inflation-indexed bearer bonds[1]. Basically, cash that appreciates with inflation. You'd have to achieve a certain critical level of acceptable before it would be a true "money" but it would certainly have big impacts:

- The inability to rob cash hoarders through inflation and taxes on inflation beating investments.
- The corresponding shift in relationship between consumers and banks.
- The automatic inflation-indexing of all loans denominated therein.
- Businesses only need to adjust relative prices, not price levels.

[1] Yes, I know "bearer bonds" are illegal, but that's simply a matter of choosing a name for them that appropriately throws off the government dogs.

Do these "models" actually have some explanatory power, or are they just toys to give the modelers the illusion of real understanding? Are there some well-known successes that someone can point to where the model predicted or explained some period of economic history?

Economists are lousey when it comes to concepts and definitions. Try to find one reasonable, measurable definition of 'utility' consistent with what we know about neurology and psychology! Economists don't even try. In this case, it's surely not as bad but let's start with some better definitions. With 'I', investments, interest sensitive expenditures are meant. Which expenditures are interest sensitive, in reality or in theory?

- Households: consumer durables and more especially cars. Also: houses. Cars and houses are by the way about the largest markets of the entire economy.

However, stating that some expenditures are interest-sensitive is about equal to stating that they are dependend on available credit/mortgages. Houses (a very large chunk of total 'I') are largely financed with mortgages.

Now take your typical Austrian deregulation induced too easy credit malinvestment boom which goes bust:

* asset prices go down
* expenditures on 'I' go down and GDP either declines or grows less fast
* unemployment goes up
* inflation goes down or even turns into deflation
* people want to rebuild their balances
* there is a large stock of unsold houses

And there is your IS/LM liquidity crisis. No matter what the interest rate is, people won't start to spend as they want to save. It's really not that complicated.

Mr. Cowen,

inspired by yuur puzzlement I've just published a little graph on the whole IS/LM thing, or in fact on the I-thing which, as I see this, is central to the whole issue. We have to ask the question why economists developed the concept in the thirties of the twentieth century, what led them to dismiss classical theory and to turn to other ways to understand the economy in exactly that period? I think that Keynes rightly guessed that a prolongued low level of investment, caused by among other things a rapid increase of productivity DESPITE THIS LOW LEVEL was the main culprit:


Modern (economic) historic reaearch of course shows that, indeed, productivity increases due to investments in real new technology (in those days: aviation, electrification, those kind of things) often lag investments by decades... this is the answer to the marginal productivity of capital problem of Keynes (p.s. - the graph is for the Netherlands, most other countries show comparable developments, the rise of the investment rate prior to 1929 started earlier in the USA. A historian called Michelangelo van Meerten has published a nice book on historical capital stocks).

Ah, Delong and Krugman, pointing out the obvious obliviousness of the author of this post. Ingrained prejudice is just so hard to let go of, despite the intelligence of the author.

A lot of lame points do not add up to a good point. Bad post.

Hicks didn't like IS-LM but so what? It's here to stay. Too much sunk cost in this endeavor by now. It's a windmill of our minds, ever turning, ever churning. It's not a Keynesian model at all. Keynes drew his economics from Marshall and Wicksell, not from Walras. His concern was dynamic adjustments of many actors with disparate beliefs and information, not the general equilibrium of a model with two non-state actors and sticky wages. But that's water under the bridge by now. IS-LM will live in textbooks, and ever more irrelevant, nasty DSGE models will keep multplying, like locusts. That's the only real multiplier effect in macro, come to think of it.

There's a lot of confused ranting in these sections.

First, ISLM is junk and everyone knows it. It is not an economic model, it is a statistical model fit to economic time series. It can fit anything, but explains nothing because there is NO ECONOMICS IN THERE.

Second, Tyler Cowen knows about as much about macroeconomics as I do about the mating behavior of tree sloths -- basically, that it happens. Taking him to task for not offering a model is pointless, he doesn't understand modern macroeconomic models well enough to use them. He's a political scientist, and a rightwing extremist one to boot.

Merjin Knibbe seems so completely lost, it is a wonder he hasn't inured himself.

There's no physics in the Ideal Gas Law. The detailed behavior of atoms gets boiled down to the ideal gas constant. The ideal gas law explains nothing -- people figured it out before we knew atoms were real. Up until then it was a statistical model fit to a bunch of empirical data.

No one says the Ideal Gas Law is junk.

These seem like theorist complaints. Physicists don't like the Standard Model, yet it makes predictions for the most part within the accuracy of being able to measure stuff.

The question is: does the IS-LM model make predictions within the accuracy of being able to measure stuff? As far as I can see, the answer is yes unless e.g. Paul Krugman is only pretending to use the model.

Therefore, the theoretical complaints cannot actually be important. Reality must fudge the distinction between real and nominal interest rates. Reality must fudge the distinction between short and long term interest rates. Reality is invariant with respect to expectations. Reality does not have a well defined Fed reaction function.

To do an extreme version of this argument, post-war nominal GDP is fit rather well by a line (in log space). If I say GDP will follow that curve with Gaussian deviations, then I only need 0.087 more bits per sample based on KL divergence to explain future GDP. And a line has a slope, and intercept and the Gaussian has a mean and a standard deviation. I know nothing about interest rates.

As a deeper (but related) critique from information theory, is that in coarse graining (e.g. macroeconomics), you necessarily lose more and more of your degrees of freedom. Think of the ideal gas law. The detailed properties of atoms get boiled down to the ideal gas constant.

Maybe your distinctions between real and nominal interest rates do not even exist at the macroeconomic level? If one can make predictions ignoring that fact then maybe the information distinguishing them is destroyed.

Comments for this post are closed