Tag Archives: islm

The dark, dark era of money demand

Did you know that Greg Mankiw and Larry Summers once wrote a paper showing that tax cuts are probably contractionary?

Neither did I.

Of course, in an effort to be sensationalist, I’m being unfair to Mankiw and Summers. I doubt either of them has ever actually believed that tax cuts depress the economy. Mankiw, after all, has some innovative ideas about how paycheck-to-paycheck consumers might make tax cuts effective as a tool to boost aggregate demand. As high-ranking economic advisers, both Mankiw and Summers presided over large tax cuts intended as stimulus.

Yet for some reason, they wrote a paper in 1986 whose working title was “Are Tax Cuts Really Expansionary?” They concluded the answer was quite possibly no:

In this paper, we re-examine the standard analysis of the short-run effect of a personal tax cut. If consumer spending generates more money demand than other components of GNP, then tax cuts may, by increasing the demand for money, depress aggregate demand. We examine a variety of evidence and conclude that the necessary condition for contractionary tax cuts is probably satisfied for the U.S. economy. (emphasis mine)

I can’t think of a better demonstration of how “money demand” once warped economists’ minds.

Mankiw and Summers, you see, were following up on a long literature that used the IS-LM model to analyze the effects of fiscal policy. In that literature, debt-financed transfers stimulated the economy: with more money in their pockets, consumers spent more. (Needless to say, this literature didn’t bother with Ricardian equivalence—but that’s another story.)

Now, there was some feedback from the “LM curve”. As consumption rose, there was greater demand for the (fixed) quantity of money, leading to higher interest rates and a partially offsetting drop in consumption and output. But this could only be a partial offset. After all, interest rates only rose because output rose: if output stagnated or fell, there’d be no dampening effect from interest rates, and nothing to offset the positive effect from the transfer, meaning that output would have to rise after all. (Contradiction!)

In IS-LM lingo, an upward movement in the “IS” curve would inevitably boost the value of “Y”:

Mankiw and Summers challenged this result with a simple observation. Maybe money demand doesn’t depend on output in the aggregate—instead, it depends separately on different components of output, like consumption and investment. In particular, Mankiw and Summers argued that money demand was influenced mainly by consumption—households hold a lot of money for consumption purposes, but they don’t need nearly as much for durable goods purchases, and businesses don’t use much cash for investment.

This slight modification of the IS-LM model makes it possible for tax cuts to be contractionary. Here’s the logic: tax cuts boost consumption, which dramatically increases money demand and forces up interest rates. Higher interest rates put such a damper on investment that the overall movement in output is negative.

Interesting story. But it misses the obvious question: why does the Fed stand by, complacently, and keep the money supply exactly the same as such clearly unintended consequences play themselves out? And assuming the Fed will never anything so crazy, why do we care about a thought experiment where it does? If it’s even just following an interest rate rule, the paper’s entire chain of reasoning is meaningless.

To be fair, Mankiw and Summers recognized these issues:

Second, our analysis considers the effect of tax cuts assuming a constant path of some monetary aggregate. Depending on the Fed’s reaction function, a wide range of alternate outcomes is possible.

But, for some reason, they still found it to be a worthwhile exercise:

Our assumption that the money stock is held constant in the face of tax changes, however, is a natural and conventional benchmark

A natural benchmark? Really? A constant money stock is natural? If this is your framework for policy analysis, you might as well conclude that the War on Drugs is the most contractionary economic program in the United States: after all, it creates a tremendous demand for cash in illicit transactions, one that quite plausibly swamps any short-term variation due to fiscal policy. If Mexican drug lords could safely wire their money around, there wouldn’t be any need for so many $100 bills. Holding the money stock constant, interest rates would plummet, investment would soar, and we’d experience a massive investment boom.

But I shouldn’t be too hard on circa-1986 Mankiw and Summers: after all, they were prisoners of their time. A constant money stock in the face of policy changes has never been a “natural” benchmark, but it certainly was conventional. Everybody used it. The Fed even toyed for a few years with its own form of monetarism—the only policy rule under which the Mankiw-Summers result (and all its IS-LM precursors) might have had a grain of truth.

This is one of those historical episodes that makes you realize how far economics has come. Somehow, in 1986, it seemed perfectly natural to write a paper on the obscure properties of the money demand function—even to two economists as sharp as Mankiw and Summers. (They don’t come much sharper than that.) Today, thanks to Michael Woodford and fellow travelers, we realize that the money part of monetary policy isn’t really so important, and that the perverse feedbacks of the old model are little more than intellectual curiosities—unless, of course, you have a central bank crazy enough to implement a money supply rule, which we fortunately do not.

And this is liberating! Macroeconomics is hard enough without having to worry about how every single policy might interact with money demand. (For instance, in extreme cases we need to discuss changes in liquidity demand; a related concept, but one with very different policy implications.) Let’s be glad that the era of money demand is over—hopefully for good.



Filed under macro

What’s needed in a macro model?

Paul Krugman defends IS-LM as a pedagogical device on the grounds that it’s part of “the minimal model that has goods, bonds, and money”. Greg Mankiw circa 2006 does much the same, favoring the IS-LM model “because it keeps the student focused on the important connections between the money supply, interest rates, and economic activity, whereas the IS-MP model leaves some of that in the background”.

But do the “important connections” in the model bear any correspondence to reality? Not really—and understanding why not is a great deal more interesting than any attempt to muddle through outdated diagrams.

As I pointed out last week, the “LM curve” represents a version of monetary policy that disappeared decades ago: a target for the money supply. Given a particular value for the money supply, higher output must be accompanied by higher nominal interest rates, which offset the increase in money demand that tends to accompany a larger economy. We’re left with an upward-sloping curve in (i,Y) space—that’s LM.

Now that Fed uses interest rates to implement monetary policy, does this make any sense? Excerpting his textbook, Greg Mankiw claims that the LM mechanism is still a useful way to understand how central banks do business. After all, with a few exceptions they still implement interest rates by using open-market operations to adjust the supply of reserves. In this light, we can say that the Fed is moving the LM curve to achieve its desired interest rate. Right?

Not quite. For LM to be useful in understanding the implementation of monetary policy, it can’t be just a long term relationship—that’s merely the near-obvious statement that all else equal, a larger economy will eventually need more money. It needs to be valid in the short term as well, the horizon over which the nitty-gritty of monetary implementation takes place. And there’s no reason why that should be true.

Indeed plenty of reason to think exactly the opposite—that at high frequencies, declines in output lead to increases in money demand. Anyone who’s read a newspaper over the last few years has surely come across the notion of a flight to liquidity. When the economy dips, there’s an increase in demand for liquid assets that vastly outweighs whatever tiny drop you’d expect in transactions demand for money. In practice, LM probably slopes the wrong way. (This is also the difficulty with Brad DeLong’s argument that LM applies to quantitative easing—QE tries to change the spread between long rates and the expected path of short rates, but there’s no reason to assume that spread has any particular relationship with output, much less a positive one.)

This isn’t to say, of course, that we should force undergraduates to scribble downward-sloping LM curves. Of course not. Rather, the exact relationship between “M”, “Y”, and “i” is so complicated and time-contingent that we shouldn’t waste time trying to model it at all. As far as I know, the guys at the New York Fed who actually implement interest rate targets don’t rely on some hyper-complicated model of the relationship between reserve demand and 132 macro variables. Instead, they inject reserves into the system when rates are above target, and take them out when rates are below target. It’s a pretty mechanical process, but it works, and you don’t need any more than supply and demand to understand why.

There’s a broader question here: what mechanisms do you really need in a macro model? For decades, monetary economists painstakingly hashed out functions for “money demand”, and spent untold amounts of econometric energy trying to estimate them. You’d see horrendously tedious papers exploring how the effects of government policy X depended on the exact specification of the money demand function. Even as late as 1999, one prominent monetary economist worried that innovations in financial markets would turn the central bank into an “army with only a signal corps”, as they brought down the demand for government-issued money.

As Mike Woodford pointed out a decade ago, none of this actually matters. Central banks today (at least the ones in developed countries) only care about money demand to the extent that it affects their ability to control interest rates, and this remains perfectly feasible when money demand is small or even zero. The messy regulatory and technical issues that determine banks’ demand for reserves on the fed funds market have virtually nothing to do with the effects of monetary policy on the economy. Obsessing over money demand is a waste of time.

When you think about it, this is fairly obvious. The “LM” curve embeds two claims about the demand for money: that it increases with output and decreases with interest rates. But there are countless other influences on the demand for money (at least base money created by the Fed), many of which are just as important in the short term. How many $100 bills do drug dealers need to evade notice? Are paper dollars still popular in countries with underdeveloped banking systems? How many ATMs has Bank of America built? Do gas stations demand payment in cash?

If you seriously believed that modeling money demand was important, you’d be working overtime to build a model with all these features. Sure, you’d probably have a “transactions demand” block like everybody else, but you’d also be surveying coke dealers to keep abreast of changes in their cash management. The fact that no one actually deems a survey of coke dealers necessary to understand monetary policy—even when their effect on money demand is quite plausibly larger than the effect from most other economic activity put together—is powerful evidence that no one really thinks the details of money demand matter.

And that’s the great thing about economic modeling: you don’t have to include every conceivable, small-bore mechanism. You can’t! Instead, you need to focus on what matters—and as the economics profession has finally come to realize, the precise characteristics of money demand just don’t make much difference. LM is irrelevant.

Fortunately, monetary economics offers plenty of other material to keep us busy.


Filed under macro

Why is LM still there?

With all the recent discussion of IS-LM, I can’t help but repeat a longstanding question of mine: what on earth is “LM” still doing in the diagram?

The “IS” curve is logical enough: by encouraging investment (not to mention spending more generally), lower interest rates lead to higher output. Sure, there are some flaws. We should really be looking at the full future path of interest rates (which is what matters for spending and investment decisions), not today’s interest rate in isolation. We should also emphasize the difference between nominal and real interest rates—so that if the vertical axis denotes nominal interest rates, inflation will shift the IS curve (which depends on real rates) upward. But as a starting point, this isn’t really so bad: the fact that higher real interest rates, all else equal, push down output is probably the most fundamental observation in macroeconomics.

But the “LM” curve? It’s implicitly describing a monetary policy rule that disappeared decades ago. Here’s the story: the central bank has a target for the nominal money supply. Due to sticky prices, in the short run this corresponds to a target for real money balances as well. Generally, higher real output (“Y”) will increase the demand for real money balances, while higher nominal interest rates (“i”) decrease it. The set of possible equilibrium pairs (Y,i), therefore, has positive slope: when high Y is elevating demand for real money, i has to rise as well to bring demand back into line with the fixed supply.

Fair enough. But central banks today don’t target the nominal money supply: in the short run, they target nominal interest rates directly. In this light, a more sensible “LM” curve would be horizontal. Now, admittedly central banks try to operate according to policy rules, under which the response of interest rates to output (or, more accurately, deviations of output from its potential level) is generally positive. If we reinterpret LM as a monetary policy rule, the upward slope makes a little more sense. In the past few decades, however, the most important feature in central banks’ policy rules has been the response to inflation, not output; the runaway inflation of the 1970s was blamed on an overeager response to the (misperceived) output gap, and for better or worse no one wanted to repeat the same mistake twice. Meanwhile, although the US has never officially joined the bandwagon, many countries now operate under an inflation targeting framework, in which responding to inflation is the key feature of the policy rule. In this environment, depicting policy as a relationship between “Y” and “i” misses what’s really going on—better to abandon the upward-sloping LM curve altogether and use a simple horizontal line to depict the current policy rate.

I’m not alone in this sentiment. David Romer wrote an entire piece for the JEP in 2000 called Keynesian Macroeconomics without the LM Curve. (As the title suggests, he shares my feelings on the matter.) Tyler Cowen puts this at #6 on his list of grievances. It’s a pretty obvious point—yet, for reasons I don’t entirely understand, we still print thousands of undergraduate textbooks a year with LM front and center.

This is nothing, of course, compared to the abomination that is the AD/AS model, also included in undergraduate textbooks. AD slopes down for the same outdated reason that LM slopes up: given a constant money supply rule, lower prices imply higher real money balances and therefore lower real interest rates, which lead to higher demand. (It can also be justified using real balance effects, which are quantitatively irrelevant, or fixed exchange rates, which only exist in a few cases.) This has absolutely nothing to do with monetary policy as it’s currently implemented. Yet the simple AS and AD curves, made appealing by the apparent (but false) analogy to ordinary supply and demand, lurk somewhere in the minds of countless former economics students. This leads to all kinds of bad intuition—like the notion that sticky prices are problematic because they prevent the adjustment to equilibrium on the AD/AS diagram. (Wrong. Under current Fed policy, the price decrease -> lower interest rate -> improvement in demand mechanism is no longer operative, unless deflation combines with the Taylor rule to force a policy that the Fed should have chosen anyway. Certainly this is no use at the zero lower bound, where price flexibility is actually harmful, because it leads to more deflation and higher real interest rates.)

Somehow, the economics profession hasn’t quite completed the transition to a world where money supply is no longer the target of choice. In research papers, of course, the change happened years ago—but intuition and the hallowed undergraduate canon are much slower to change. Meanwhile, we’re left with LM, the strange vestige of an earlier era.


Filed under macro