Tyler Cowen linked to a post by Frances Coppola:

But one thing seems clear. How negative rates would work in practice is no clearer than how QE works in practice. They are experimental, and their effects are complex. Hydraulic monetarist arguments (“if you can get rates low enough the economy will rebound”) are simplistic.

Monetarist?  Don’t they focus on the money supply and/or NGDP expectations?  Most people would consider Milton Friedman a monetarist, and here’s what he had to say about low rates:

Low interest rates are generally a sign that money has been tight, as in Japan; high interest rates, that money has been easy.   .   .   .

After the U.S. experience during the Great Depression, and after inflation and rising interest rates in the 1970s and disinflation and falling interest rates in the 1980s, I thought the fallacy of identifying tight money with high interest rates and easy money with low interest rates was dead. Apparently, old fallacies never die.

In a monetarist model, lower market interest rates are contractionary for any given monetary base, because they reduce velocity.  It’s the Keynesians who are likely to claim that lower rates are expansionary.  Now of course Friedman was talking about market interest rates, not IOR. Lower IOR is theoretically expansionary, and so far markets have reacted to negative IOR announcements as if they are expansionary.

Will that be true in the future?  Nothing is certain, as monetary policy is very complex, and concrete action A can always be interpreted as a signal of future concrete action B.  Anything is possible.  A $1 billion increase in the base can have a contractionary impact if a $2 billion increase was expected.  But any monetary analysis should start from the presumption that reducing the demand for an asset will reduce its value.  A reduction in the value of base money is expansionary.  That means lower IOR has a direct expansionary impact on NGDP.  (Secondary effects are complex, as Coppola suggests.)

Indeed, individual banks can avoid paying negative rates on excess reserves. They can discourage customers from making deposits; they can choose to hold reserves in the form of physical cash; or they can increase new lending (not refinancing), since the balance sheet result of new lending is replacement of reserves with loan assets.*

But of course the reserves do not disappear from the system. They simply move to another bank, which then incurs the tax. The banking system AS A WHOLE cannot avoid negative rates on reserves. The reserves are in the system, so someone has to pay the negative rate. If banks increase lending to avoid the negative rate, the velocity of reserves increases. It’s rather like a game of pass-the-parcel.

This is a common misconception, which I see all the time.  If individual banks don’t want to hold a lot of excess reserves, they can simply buy other assets with them.  If the banking system as a whole wants to reduce its holding of excess reserves, it can reduce the attractiveness of deposits until the excess reserves flow out into currency held by the public.  The central bank controls the monetary base, not bank reserves. The composition of the base is determined by banks and the public.

It seems reasonable to suppose that negative interest rates might increase loan demand. Negative interest rates on reserves put downwards pressure on benchmark rates and thus on bank lending rates, attracting those who would otherwise be priced out of borrowing. But typically those are riskier borrowers. We have just spent eight years forcing banks to reduce their balance sheet risk. Do we really want to force banks to lend to riskier borrowers? Of course, tight underwriting standards could be used to deny those people or businesses loans: but doesn’t that rather defeat the purpose of negative rates? It’s something of a double bind.

We need an easier money policy and, if banks are taking excessive risks, a tighter credit policy.  It’s best not to mix up monetary and credit issues.  Negative IOR is about reducing the demand for base money, which is inflationary; it’s not about increasing bank loans.  Central banks should not be trying to encourage more debt creation—unless you want to end up like China (where the policies are unfortunately linked together.)

In the end I agree with those who are skeptical of negative interest rates.  These ultra-low interest rates are a sign that monetary policy is too contractionary. The developed world needs much higher interest rates, but only if we get there with an expansionary monetary policy.  In December the Fed tried a short cut, raising rates without boosting NGDP growth. This is putting the cart before the horse.  You need to generate higher NGDP growth expectations first, and then you can achieve a permanently higher level of interest rates.

Multipliers are ratios. That’s really all they are. There is the money multiplier (M2/MB), the fiscal multiplier (1/MPS) and the velocity of circulation (NGDP/MB, or NGDP/M2). If you assume these ratios are stable, you can derive some very interesting policy results. Of course the ratios are not completely stable, but may be stable enough to be of some value. Sometimes. My own view is that multipliers aren’t particularly useful, but today I’d like to assume the opposite, and show that the implications are not necessarily what you might assume.  (And please, no comments from MMT zombies “explaining” to me that multipliers don’t exist.)

Milton Friedman faced a quandary when trying to explain how bad government policies led to the Great Depression. If he defined the money supply as “the monetary base” (as I prefer), people would have pointed out that the base increased sharply during the Great Depression. Alternatively, he could have adopted the market monetarist practice of defining the stance of monetary policy in terms of changes in NGDP. Thus falling NGDP during 1929-33 was, ipso facto, tight money. His critics would have objected that this begged the question of how could the Fed have prevented NGDP from falling.

So he split the difference, and settled on M2 as both the definition of money, and the indicator of the stance of monetary policy. He suggested that, “What is money?” was essentially an empirical question, not to be determined on theoretical first principles. His statistical analysis led him to conclude that M2 (which unlike the base did fall during the early 1930s) was the preferred definition of money. And also that growth in M2 should be kept stable at roughly 4%/year.

In my view M2 no longer represents a good definition of money, using Friedman’s pragmatic criterion. Look at M2 growth in recent years:

I don’t know about you, but I see almost no correlation with the business cycle. Indeed M2 growth soared in the first half of 2009, making money look “easy”, which is obviously crazy. So if Friedman were alive today, how would he define money? The base still doesn’t work, as reserves also soared in 2008-09. Nor does M2. I don’t have a good answer, but I suspect that coins might be the best definition. Unlike the base and M1, periods of illiquidity probably don’t lead to massive hoarding of coins.  They are primarily useful for making transactions (although a sizable stock is held in piggy banks.)

Unfortunately, I could not find any data for the stock of coins in circulation. (Which is a disgrace, when you think about the 100,000s of data series the St Louis Fred does carry. As I recall, back in the 1990s coins were almost as important a part of the base as bank reserves.) But I did find data on annual coin output. For simplicity, I chose unit output, but value of output (which counts quarters 5 times more than nickels) would almost certainly lead to broadly similar results. In the list below I will show the change in annual coin output, compared to the year before, and also the change in the unemployment rate at mid-year (June) compared to the year before. The unemployment rate change is in absolute terms:

Year  * Coin Output  * delta Un

2000:   +28.1%           -0.3%
2001:    -30.9%          +0.5%
2002:    -25.7%          +1.3%
2003:    -16.5%          +0.5%
2004:    +9.5%          -0.7%
2005:   +16.1%          -0.6%
2006:    +1.4%           -0.3%
2007:    -6.9%             0.0%
2008:    -29.8%        +1.0%
2009:   -65.0%          +3.9%
2010:   +79.6%          -0.1%
2011:    +28.7%         -0.3%
2012:   +13.9%          -0.9%

Unfortunately my data ends at 2012, but that’s a really interesting pattern. Especially given that I don’t have the data I’d actually prefer.  I’d like the change in the size of the coin stock; instead I have the change in the flow of new coins (but not data on old coins withdrawn.)  It’s more like a second derivative.

In any case, it’s an amazing correlation. The signs are opposite in every case except the one where unemployment doesn’t change at all.  Coin output falls during years when unemployment is rising, even years like 2003 when unemployment is rising during a non-recession year.  And even better, the biggest change by far in coin output (proportionally) is in 2009, which also saw the biggest change by far in unemployment.

If you are not good at math then you’ll have to take my word for 2010 being a smaller change in proportional terms.  Indeed if you look at actual coin output in levels, 2010 was the second smallest in the sample, 2011 the third smallest, and 2012 the 4th smallest.  The decline in 2009 was so great that we never really climbed out of the hole.

Now let me emphasize that there’s an element of luck here.  If we had coin data for 2013 and 2014 I doubt the relationship would hold up.  Coin output seems to be in a steep secular decline.  So it’s partly coincidence that the signs are reversed in virtually every case.  But not entirely coincidence.  Perhaps someone could do a regression (using first differences of logs of coin output—so that the 2009 change will be larger than 2010) and confirm my suspicion that this relationship does show something real.  Falling coin output is associated with recessions.

But does it cause recessions?  If only you knew how tricky the term ’cause’ really is!  Krugman basically called Friedman a liar (soon after Friedman died) for claiming that tight money caused the Great Depression, whereas in Krugman’s view Friedman’s data pointed to the real problem being a non-activist Fed—they didn’t do enough to prevent M2 from falling. But they didn’t cause it to fall with concrete steppes.  The base didn’t fall.

I’ve always believed we should think of “causation” in terms of policy counterfactuals.  Suppose the Fed had acted in such a way that M2 didn’t fall.  And suppose that in that case there would have been no Great Depression.  Then if the Fed was capable of preventing M2 from falling (which is itself a highly debatable claim) then there is a sense in which Friedman was right, the Fed did cause the Great Depression.  Again, that’s if they could have prevented M2 from falling, and if stable M2 would have prevented a depression–both debatable (but plausible) claims.

My claim is that if we use Friedman’s pragmatic criterion for defining money, then coins might possibly be the best definition of money for the 21st century.  If the Fed had acted in such a way that coin output was stable in 2007-09, or at worst declined along its long run downward trend, then there would have been no Great Recession.  So in that sense the fall in coin output “caused” the Great Recession. But I could also find a 1000 other “causes,” such as plunging auto sales.

Can the Fed control the coin stock?  I’d say they could in exactly the same way they can control M2 (or nominal auto sales), via a multiplier.  The baseline assumption is that both the coin stock and M2 move in proportion to the base.  That would be the case if the M2 and coin multipliers were stable.  If the multipliers change, then the Fed simply adjusts the base to offset the effect of any change in the coin multiplier.

No let me quickly emphasize that I view the preceding as an extremely unhelpful way of thinking about monetary policy and the Great Recession.  I still prefer to define money as the base, as the base is directly controlled by the Fed.  And I prefer to define the stance of monetary policy as NGDP growth expectations.  And I prefer to think of tight money as setting the monetary base at a level where NGDP growth expectations fall below target, as in 2008-09.  I’d just as soon leave coins to children with piggy banks and nerdy collectors.  But if you insist on defining money using Friedman’s pragmatic criterion, then coins are my definition of the money stock.

A penny for your thoughts?

PS.  I have a new post on the Phillips Curve at Econlog.

It’s possible that I’m the one that’s confused.  But since it’s my blog, I’ll write the post as if others are confused.

Picture the AS/AD diagram.  Now shift AS to the right, due to population growth, capital accumulation, resource discovery, or technological developments.  What happens to AD?  I guess it depends what you mean by “AD.”  I’d say nothing happens, although the quantity demanded rises at a lower price level.  When I read others I often get the impression they have in mind some sort of “real AD” concept, which would drain AD of all meaning.  After all, if it’s quantity demanded, then any and all changes in output are changes in aggregate demand.   This would allow no debate as to whether recessions were caused by AD shocks, as a recession is defined as a significant fall in output.  It becomes a tautology!   Yet I get the feeling reading people like Stiglitz that he views AD is a real concept, not a nominal concept.

Or consider an increase in AD when the economy is at capacity.  The textbooks say you just get inflation in that case.  But if you used a “real AD” concept, then there would have been no increase in AD in the first place.  That’s right, even in Zimbabwe AD did not rise, because output didn’t rise.

Why does this confusion exist?  Perhaps because we have two radically different ways of thinking about AD; the monetary approach (M*V) which is obviously a nominal concept, and the Keynesian approach (C+I+G+NX) , which could be visualized in either nominal or real terms.  Most people are Keynesians, and think in terms of actual purchases of goods and services.  To take a micro analogy, most people views the terms ‘consumer purchases’ and the term ‘demand’ as being synonymous.   Even though purchases are also sales, and could just as well be termed “quantity supplied.”    (Remember those graphs of “oil demand” over the next 50 years?)  When I read popular writers on macroeconomics I see them break the economy down into sectors, and talk about things like “December demand for US made cars,” what they really mean is “quantity demanded.”

The deeper problem is that the Keynesian and monetarist worldviews are nearly incommensurable.  It’s very hard to mentally toggle back and forth between the two approaches, because they are so radically different.

When I read the following quotation from Tyler Cowen, I initially wondered whether he was confusing AD with real quantity demanded:

Weak job creation remains at the heart of America’s unemployment problem.  Accepting this hypothesis does not require the rejection of Keynesian economics; for instance you can think of weak job and start-up creation as one reason why AD is not recovering so well on its own, with causation running both ways of course.

(BTW, commenters should not complain that his first sentence is tautological, he’s talking about gross job creation, not net job creation.)

I see economic dynamism, creative destruction, as something that affects AS, not AD.  Yet in the very next line he shows that he’s not confused.  Like me, he views AD as a nominal concept:

Remember “” monetary velocity is endogenous to perceived gains from trade.

But I’m still not happy, because I don’t agree with his implicit assumption that velocity shocks affect AD.  They do under Friedman’s 4% money growth rule, and they do under a gold standard.  But velocity shocks have no impact on AD under the following monetary regimes:

1.  Inflation targeting.

2.  NGDP targeting.

3.  A Taylor Rule.

4.  A hybrid policy where the central bank does just enough QE to prevent inflation from falling below 1%, but no more.

I’m not quite sure what sort of regime we have today, but my hunch is that it’s closer to the 4 items on that list, then it is to either a 4% money rule or a gold standard.  If I had to guess I’d assume Tyler might have made the following error:

1.  He developed a real theory of unemployment (no problem there.)

2.  Saw market monetarists looking over his shoulder, or perhaps felt uncomfortable with empirical evidence that AD matters too, and decided that the theory was in some way compatible with AD theories of the recession.  But I don’t think you can do that.  An AD theory must be 100% nominal.  That means it must move the monetary policy process front and center into any explanation.

This doesn’t mean that real shocks can’t matter.  For instance, I speculated that in 2008 and 2011 the oil price shocks made monetary policy more contractionary, which reduced AD.  In both cases the Fed saw high headline inflation rates, and became squeamish about monetary stimulus.  In both cases they tightened enough to reduce NGDP growth, even as inflation was still above target.  The slower NGDP growth slowed the economy.  It’s possible that a similar explanation could be developed with Tyler’s job creation story.  But I don’t think it’s enough to tack on a “velocity might fall” explanation.  To me, that seems too much like someone working out a real theory of AD, and then assuming nominal AD must move in the right way to make it work.  As when Keynesians convince themselves that fiscal policy must affect AD, and then offhandedly suggest that velocity will move in the right direction to make it happen.  Maybe so, but the theory needs to be developed in terms of actual central bank practice, i.e. changes in M*V, not just a add on assumption about velocity.

PS.  A whole different issue is the question of how nominal AD shocks get translated into real changes in quantity demanded (those December car sales.)  For that you need wage/price stickiness, and Tyler Cowen has a new post that discusses fascinating evidence on wage stickiness in rural India (where you might expect wages to be flexible.)

PPS.  Here’s how I’d make the argument if a gun was pointed at my head.  A more dynamic job creation process would somehow raise the Wicksellian equilibrium real interest rate.  This would allow the Fed to do less of the “unconventional monetary stimulus” that is it squeamish about doing, and more conventional stimulus, for any given inflation rate.  Do I believe that?  I’m not sure.

Here’s Ed Dolan:

I see NGDP targeting as the natural heir to monetarist policy prescriptions of the 1960s and 70s.

If we look at the textbook version of monetarism, the point is almost trivial. Textbook monetarism begins from the equation of exchange, MV=PQ, where M is money (M1, back in the day), V is velocity, P is the price level, Q is real GDP, and PQ is NGDP. Next it adds the simplifying assumption that velocity is constant. It follows that targeting a steady rate of money growth is identical to targeting a steady rate of NGDP growth.

There is a rapidly developing consensus among economists of both the left and the right that NGDP targeting is the way to go.  How did this consensus develop, given their wildly different views on other important issues?

The original push for NGDP targeting seemed to come mostly from economists on the right.  It was seen as a good post-monetarist policy rule.  Recall that the monetarists had favored a stable rate of increase in a monetary aggregate, usually M1 or M2.  The hope was that velocity would be stable.  Unfortunately, fluctuations in velocity during the early 1980s seemed to discredit monetarism (although in fairness Friedman seems to have erred in focusing on M1 at the time.) So conservatives began to look for alternatives.  Bennett McCallum advocated a NGDP targeting rule, where the monetary base would be adjusted to offset changes in velocity.  And I think it’s fair to say that most market monetarists have vaguely libertarian leanings.

This basic approach got a lot of attention from economists like Greg Mankiw, Robert Hall, and John Taylor, although they didn’t end up in exactly the same place.  Matt Yglesias linked to a Tim Lee tweet that reminds us that the late William Niskanen had NGDP targeting enshrined in the Cato Handbook for Policymakers:

The intent of Congress would be better served and monetary policy would be more effective if Congress instructed the Federal Reserve to establish a monetary policy that reflects both their concerns in a single target. The best such target, I suggest, would be the nominal final sales to domestic purchasers””the sum of nominal gross domestic product plus imports minus exports minus the change in private inventories.

[Slightly different from my version, I believe Beckworth and Woolsey prefer nominal final sales, but I am not certain.]

So how about the left?  Why has Matt Yglesias, Brad DeLong, Paul Krugman, Christina Romer, et al, become interested in NGDP targeting?  In my view they have recognized the advantage of NGDP targeting over inflation targeting, which in its simplest version pays no attention to unemployment.  In fairness, most central banks, including the Fed, do pay some attention to unemployment.  This is called flexible inflation targeting.  But in the US that flexibility allows a lot of wiggle room.  The Fed seems to be reluctant to focus aggressively on the growth side of their mandate, especially when all they have to work with are unconventional policy tools.  As a result they have ended up not even hitting their inflation target over the past three years (on average), despite very high unemployment.  No wonder the left is frustrated.  NGDP targeting would make it very clear to the Fed that inflation and growth are equally important in the short run (while assuring low inflation in the long run.)  So the left is on board.  Indeed I’d even put Goldman Sachs into that category, if “left” means “Keynesian approach,” rather than “wild-eyed radical.”

So we have finally achieved agreement between the left and right on a critical issue to policymakers.  Now we just need to convince that moderate who’s in charge of the Fed.  Glad to hear that he found yesterday’s discussion “interesting.”  When the left and right agree, what’s there not to like for moderates?  Time for a David Brooks column?