Money multipliers, real and imagined
Arnold Kling has a new post that contains several misconceptions about “money multipliers.”
He starts by citing a paper with the following phrase contained in the title:
Does the Money Multiplier Exist?
This is a misleading title, equivalent to asking: “Does the velocity of circulation exist?”
Clearly what the authors meant to say was; “Is the money multiplier stable?” Here is how Kling reacts to their paper:
I think that the popular saying among monetary economists these days is that attention has shifted from the Fed’s liabilities to the Fed’s assets. The old story was that the Fed’s liabilities were currency and bank reserves, and the banks lent out a predictable multiple of their reserves. The new story is that banks hold a ton of excess reserves. Also, if you include retail money market mutual funds in M2 (when did that happen? I’m so out of it, I thought that M2 was still, you know M2), then Carpenter and Demilrap are right that the money multiplier was never so reliable, anyway.
It would be more accurate to say that economists argued that under most circumstances the ratio of deposits to reserves was fairly stable. But obviously not under all circumstances, Friedman and Schwartz showed that the money multiplier was highly unstable during the 1930s, for instance. In the late 1930s this was mostly due to the near-zero interest rates, much like today.
Of course the money multiplier depends on more than the banking system, the behavior of the public also matters:
multiplier = (1 + C/D)/(R/D + C/D), where money = C + D and the base equals R + C.
The fact that reserves are not required for time deposits may or may not make the multiplier highly unstable, just as the C/D ratio might or might not make the multiplier highly unstable. So there’s really nothing new here, the same old questions as before.
Old monetarists tended to assume that the money multiplier was typically fairly stable, and that with sound monetary policy V would also be fairly stable. Thus the monetary base and the monetary aggregates are useful indicators of the stance of monetary policy. Obviously I don’t accept either argument; I think both the money multiplier (however defined) and V are too unstable for the base and M2 to be useful indicators for policymakers.
So far these are minor quibbles, but then Arnold goes dangerously off course:
Anyway, back to the Fed’s assets. When the Fed buys long-term Treasuries, this takes them out of the hands of private investors, who then have to find something else to buy. They bid up the prices of other bonds and drive down interest rates, or so the theory goes.
My own view is that in an enormous world capital market, the Fed is not driving long-term interest rates. I am willing to be wrong. But my null hypothesis is that the Fed is always in an asset substitutability trap. Financial markets work to create substitutability. As a result, you have Goodhart’s Law: if the Fed can control the supply of an asset class (or definition of money), then that asset class will not have much effect on the economy; if an asset class correlates strongly with economic activity, the Fed will not be able to control it.
Another way to put this is that monetary and financial arrangements are endogenous with respect to Fed procedures. The financial markets will evolve ways to insulate the economy from what the Fed does.
Actually “the theory” is very different. According to the liquidity preference theory it doesn’t much matter what the Fed buys. A purchase of gold would drive up the price of Treasuries almost as effectively as a purchase of Treasuries.
Second, Arnold has not described Goodhart’s Law correctly. It does not say that control of an asset class will prevent that asset class from having much effect on the economy. Rather that the “multiplier” or “velocity” (the terms means essentially the same thing) between asset X and NGDP will change unpredictably when the Fed begins to target X.
And this means that there is utterly no reason to presume that “the financial markets will evolve ways to insulate the economy from what the Fed does.” Indeed so far as I know no economist has ever even proposed a model where this is true. And that’s because it would have to be a very strange model. Banks and the public choose the R/D and C/D ratios in order to maximize utility. Ditto for velocity. It would be exceedingly bizarre if the utility-maximizing value of these multipliers moved precisely inversely to the monetary base, thus neutralizing the impact on NGDP. (Except obviously in the case of temporary base injections.) The mistake is to jump from the very reasonable claim that the multiplier and V are unstable and hence not useful, to the totally unwarranted claim that this means monetary policy has no impact on NGDP.
And since wages and prices are sticky in the short run, when monetary policy impacts NGDP it also impacts the real economy. That’s why markets keep freaking out over rumors of tapering, despite dozens of famous bloggers and elite economists telling the markets that QE doesn’t matter. The markets know better.
Tags:
6. October 2013 at 10:08
“A purchase of gold would drive up the price of Treasuries almost as effectively as a purchase of Treasuries.”
This is not correct. There is a very good reason why the Fed buys Treasuries and not gold. They buy Treasuries and not gold because what they buy will be more greatly affected in terms of price, than if they bought something else.
During the crisis starting in 2007/2007, the Fed bought the loans of many distressed banks and non-financial corporations such as Caterpillar, instead of more Treasuries, precisely because they knew that the prices of that debt would be more greatly affected through direct purchase, than through devaluing the dollar through Treasury debt purchase only, and hoping that “spending” would rise sufficiently for the output or debt of the specific corporations in question to be sold at profit generating prices.
6. October 2013 at 10:17
NGDPLT would be subject to Goodhart’s Law.
Goodhart’s Law is conventionally summarized as
“When a measure becomes a target, it ceases to be a good measure.”
Thus, if a central bank targets price levels, then price levels cease to be a good measure. Market monetarists agree with this.
If a central bank targets aggregate spending, then aggregate spending will cease to be a good measure. Market monetarists ought to agree with this.
If NGDPLT were ever adopted, and we realize that NGDP is not a good measure either, then you can bet that yet another change in monetarism will take place, another crop of monetarists with the new key, the new religion, will arise. Maybe aggregate wage targeting will be next. And then Goodhart’s Law will apply to that as well.
And on and on, the whole time true believers refusing to accept that central planning does not work because it corrupts the pricing system.
6. October 2013 at 10:27
Scott, sometimes you say something I really don’t understand, this is one of them:
“According to the liquidity preference theory it doesn’t much matter what the Fed buys. A purchase of gold would drive up the price of Treasuries almost as effectively as a purchase of Treasuries.”
Assume the Fed never ever bought Treasuries or and public sector debt.
Assume Fed only bought / sold random single shares of stock (weight by price).
I think you mean that if they creating money on their digital ledger and were buying a bunch of stock, that the nominal value of Treasuries would rise as well.
Is this right?
But my gut says this also means that relative to the status quo the Govt. would be paying higher levels of interest.
My gut says this change over the long run would fundamentally improve the growth rate of the economy, mainly bc govt would shrink and private sector would grow relative to today, but also bc the transmission mechanism would be “faster” the stock market.
Am I wrong on this? why?
Or does that fit with what you meant?
6. October 2013 at 11:23
Morgan, During normal times when rates are positive they don’t buy enough to materially affect the asset price directly (from the purchase). The effects are mostly indirect–how asset prices of all sorts respond to the effects of monetary stimulus.
Maybe at the zero bound it would make a difference, as they’d buy large amounts. In that case I’d rather they buy Treasuries, I’m not sure why you want the Fed owning companies.
6. October 2013 at 12:30
“Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.”
– Goodhart 1975.
6. October 2013 at 12:37
Dear Scott,
I have three amateur undergrad questions which I think are all the same.
What is the relationship between Velocity and the money multiplier? I mean when I say this, is it causal in either direction?
Is there anything that can influence the money multiplier besides how much deposits the public wants to hold and how much currency they want to hold? Are you describing Bernanke’s view?
When you say… “Of course the money multiplier depends on more than the banking system, the behavior of the public also matters.”
The equation you gave, I think, explains how the public’s actions influence the money multiplier. Could you be more precise in what WAY the banking system influence the money multiplier, as separate from the WAY the public influences the money multiplier.
Thank you.
6. October 2013 at 13:24
“When the Fed buys long-term Treasuries, this takes them out of the hands of private investors, who then have to find something else to buy.”
This is a common view amongst non-economists, who think that the Fed drives rates by bidding up bond-prices. People who have taken college macro however know this effect to be tiny, compared to the ISLM effect of people substituting from money to bonds when they have more than they want, until the opportunity cost of holding money is low enough that they are satisfied. I am really surprised to see Arnold Kling making this mistake however.
6. October 2013 at 13:25
Excellent blogging.
I do wonder why the private sector fails to create “safe assets”. It seems to me first tranche (the first 30 percent LTV) property loans on class A real estate, simple underwriting, would be a safe asset. Such debt might “insulate” bond buyers in credit markets from QE purchases…but would probably be stimulative. Curiously, the private sector seems unable to create safe assets…but we have another anti-QE argument here.
QE is hyperinflationary but the market “insulates” itself against Fed actions.
I guess the Fed is impotent—so raising rates will be ineffective also…
6. October 2013 at 13:53
“Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.”
This is one of those ‘cutisms’ that can be true surprisingly often (like Murphy’s Law and Parkinson’s Law) and which so can be used to get people to think a bit — but which in no way even approach passing the test to be a real “Law”.
If I observe a statistical regularity between the flow of fuel to an engine and the speed at which it runs, and then apply pressure to the gas peddle for control purposes as to that speed, the relationship does not collapse.
Also, note the weasel words “tend to”. Tend to a lot? Or tend to only a tiny little bit?
E.g., if Arnold’s objection really is actually kinda, partly true, then it is not true at all — because given the Fed’s unlimited scope of action it can accomplish whatever it wants through the part of the objection that is false.
So Arnold is making a really absolutist claim by assumption, for which I see no apparent justification, fallacy of the excluded middle. ISTM.
6. October 2013 at 14:42
I don’t think this fits exactly, but if banks are holding T-bills, and the interest rate on reserves is approximately equal to that on T-bills, then if the Fed buys T-bills, will tend to just to increase reserve demand and reduce the money multiplier. If the Fed buys from a bank, it is a pure substitution. If the Fed buys from someone else and it lowers the T-bill rate, then banks will tend to sell off T-bills and accumulate reserves.
What the Fed would need to is buy enough T-bills so that the banks no long hold any–one way or another.
Of course, why pay interest on reserves at the same rate as T-bills and buy T-bills?
Of course, the Fed isn’t buy interest
6. October 2013 at 14:45
One other thought. I assigned a Roger Garrison paper to my students last summer. He argued that the stable money multiplier/stable velocity version of monetarism requires financial repression.
6. October 2013 at 15:21
Jim Glass:
“If I observe a statistical regularity between the flow of fuel to an engine and the speed at which it runs, and then apply pressure to the gas peddle for control purposes as to that speed, the relationship does not collapse.”
Your example begs the question of precisely what pressure to the gas peddle you used in order to even make a statement as the “statistical regularity of the flow of fuel.”
A better analogy would be if you had no control over the vehicle to begin with (default position of you vis a vis the market economy), and then you observed that over time, without your control, the average relationship of fuel to gas peddle pressure was, say, 40 gallons of fuel per week to 20 Newtons of pressure.
Then you attempted to “fix” the pressure to 20 Newtons of pressure every minute of every day, and then you wonder why 20 Newtons of pressure causes all sorts of accidents, such that the gallons actually used is different from the 40 gallons per week you expected.
6. October 2013 at 15:34
It’s rather disturbing that so many people, on this board included, are oblivious to the breakdown of the monetary system taking place.
An important reason why the focus of financial markets is almost entirely on the Fed, is that the Fed has so distorted the markets that markets can no longer remain self-sufficient through voluntary savings.
6. October 2013 at 16:29
Scott: your argument in this post, like your argument in the previous “banks maximise profits” post, hinges on the classical dichotomy between real and nominal variables. Anyone who understands that dichotomy will see what you are getting at. But many people have no knowledge of the classical dichotomy, and will be scratching their heads. For someone like you (a MOA man) that dichotomy is central. But it’s not easy to explain. They are looking for the transmission mechanism between hacksawing into two pieces the metre rod in Paris and the sun doubling in size.
6. October 2013 at 17:05
A puzzling post. Of course I am aware that the Fed offers three definitions of money supply. There are many more. Most puzzling is that Federal Debt is not counted as money supply.
In two post, I make the case that Federal Debt should be counted as money supply.
The first link is “Government Debt is NOT Money Supply?” found at http://mechanicalmoney.blogspot.com/2013/10/government-debt-is-not-money-supply.html .
The second link is “Government Provided Money Supply” found at http://mechanicalmoney.blogspot.com/2013/09/government-provided-money-supply.html.
There is no reason to think that one definition of money supply will support all arguments. Instead, the choice of money supply definition should complement an inclusive description of economic effect. Economist should remember that the public does not react using the theories held by economist, but instead reacts in accordance with perceptions learned from the lessons of daily living.
6. October 2013 at 17:14
Geoff—
You have it backwards. The globe is glutted with savings.
Check long-term trends in sovereign debt yields, or oversubscribed Treasury auctions…money managers of every stripe stuffed to the gills…huge sovereign wealth funds (scary idea, that)…capital is everywhere and cheap…
6. October 2013 at 17:24
Bill Woolsey: Okay, then can the Fed monetize debt without inflationary impact? If so, should not annual debt monetization of, say, $1 trillion a year be done for next several years? In other words, taxpayers give bondholders cash and extinguish federal IOUs. I know this is taboo, but now critics of QE are implying this is possible…
6. October 2013 at 17:25
JoeMac:
Let’s use M2 as the money supply:
M2*V = P*Y
MB*(multiplier)*V = P*Y
So both the multiplier and V serve roughly the same role. Both are variables that change over time. Both tend to decrease sharply as nominal interest rates fall to zero.
The banking system can influence the multiplier via its influence on R/D. The government sets a minimum R/D, but only for demand deposits. In addition, banks can hold more than the legal minimum, and recently have done so.
Saturos, Yes, that’s the popular view. Kling is more a finance expert than a monetary expert, so he takes a more finance-oriented view, even where a money view is needed.
Jim, Yes, Arnold’s problem wasn’t the assumption that the linkages can vary, that part’s true, but rather assuming they’d vary in such a way as to make monetary policy ineffective, which is very unlikely to be true.
Bill, I have some sympathy for Garrison’s view, although it’s a matter of degree. Even financial repression doesn’t get you a constant multiplier, and no repression doesn’t mean the multiplier is completely unstable. It approaches (1 + C/D)/C/D as the reserve ratio approaches zero. And the C/D ratio might well be fairly stable.
Nick, Not sure what they’ll be scratching their heads over, this post doesn’t contain anything at all that’s even slightly controversial. Kling’s claim that the financial system insulates the economy from monetary shocks is the odd claim, I’ve never seen anyone make that claim before. Why would the financial system behave in such an odd way? Keynesians, monetarists, Austrians, RBC-types, Mundellians, etc, all agree that monetary policy affects NGDP when not at the zero bound.
Roger, you said;
“Economist should remember that the public does not react using the theories held by economist, but instead reacts in accordance with perceptions learned from the lessons of daily living.”
Often true, but it depends which “public” you mean. When rumors of tapering drive down bond prices, it’s not everyday experience, but economic theory that explains the move. Ditto for when the yen plunged on the announcement in Japan of the 2% inflation target.
6. October 2013 at 17:32
Nick, Rereading your comment I think you misread the post. I did not in any way rely on the classical dichotomy. Indeed I don’t think I believe in the classical dichotomy anymore than Bernanke, Krugman, Woodford, or any other mainstream economist. As far as I know we all think it applies in the long run, but not the short run. (For level changes, you can debate rates of change changes.)
6. October 2013 at 19:17
“Kling’s claim that the financial system insulates the economy from monetary shocks is the odd claim, I’ve never seen anyone make that claim before”
I’m pretty sure he gets it from Fischer Black, “Business Cycles and Equilibrium.” I don’t know the work well, and most of it was never published. Perry Mehrling describes it in Fischer Black’s bio, which I know Kling has read.
6. October 2013 at 19:31
Bill, is there a link to the Garrison paper?
7. October 2013 at 04:59
“Another way to put this is that monetary and financial arrangements are endogenous with respect to Fed procedures. The financial markets will evolve ways to insulate the economy from what the Fed does.
the Fed can control the supply of an asset class (or definition of money), then that asset class will not have much effect on the economy; if an asset class correlates strongly with economic activity, the Fed will not be able to control it.”
It just depends on how relevant the asset class controled by the fed is IMO. It doesn’t matter how much base the fed creates because most of the economy doesn’t use it. Of course there is a demand for currency but it is only a small proportion of overall money supply. The fed can create as much base as it likes and the banks will just park it in reserve accounts. In order for the fed to have an effect on the economy it needs to affect assets which are relevant.
If the fed directly affected the broader money supply the endogenous and exogenous aspects of financial markets will be more connected . This can be done by allowing people to directly transact in base or crediting deposit accounts of people at commercial banks when conducting policy.
7. October 2013 at 07:50
What about MZM? Money market funds likely should be considered, M2 does not fully capture all money market funds, and the Fed cares a lot about breaking the buck.
The ratio of NGDP/MZM also explains the 10 year Treasury yield.
http://research.stlouisfed.org/fred2/graph/?g=n5i
For the most part, MZM “multipliers” were incredibly stable to reserves. So it was a classical world from 1959-1994. The other side of base money, currency, was a bit more unstable, but not in the long run.
http://research.stlouisfed.org/fred2/graph/?g=n5j
The Fed disrupted multipliers by drastically slowing reserve formation: shorthand, the ratio of MZM to reserves *was* the “shadow banking” problem — money that is clearly vulnerable to a “bank run”. Insufficient reserves against zero-maturity liquid money!
http://research.stlouisfed.org/fred2/graph/?g=n5l
Can’t scream it enough. The Fed actively failed to provide reserves from 1994-2007. The unbacked “shadow banking” was the clear result. A lack of reserves inevitably and ultimately destabilized the financial system and markets.
I’d love to understand why the Fed chose this path: to starve banks of reserves.
The Fed has since provided funds to the reserve-starved banks, but guess what? It’s at least one-half a return-to-trend than a new and shocking balance sheet expansion: the banks and Treasury suck up their off-balance-sheet obligations, the Fed provides reserves, and it’s been less of a base money explosion than it seems.
7. October 2013 at 09:52
jknarr,
This reserve draw-down is shown in a different way by comparing total bank deposits to total loans.
http://research.stlouisfed.org/fred2/graph/?g=n8o
Bank liquidity problems corresponded with bank loans exceeding bank deposits in about 2001 and 2007.
The second line on the graph is a total of bank deposits and federal debt held by the public. To the extent that this sum might be considered “money supply”, the increasing sum did not prevent the 2007 crash.
7. October 2013 at 10:20
jknarr,
I should have pointed out that the DPSACBW027SBOG plus FDHBPIN line is annual change in billions. DPSACBW027SBOG less TOTLL is the difference in billions.
I believe that the correlation between the lines until about 2001 is due to Federal Debt growth basically matching bank excess deposits (you might say “excess reserves”).
This would be a demonstration of how deposits remain the same while Federal Debt grows. Treasury borrows deposits and then, after the Federal Government pays it’s bills, the deposits are restored to pre-loan levels.
The liquidity draw-down remains very obvious in this graph.
7. October 2013 at 10:42
“It does not say that control of an asset class will prevent that asset class from having much effect on the economy. Rather that the “multiplier” or “velocity” (the terms means essentially the same thing) between asset X and NGDP will change unpredictably when the Fed begins to target X.”
This makes sense to a layman like me. It sounds me similar to the complaint I’ve heard that targeting M2 (or was it M1?) in the early 1980’s didn’t “work” because it led to very volatile interest rates. My thought was always “of course if you target something other than interest rates then interest rates will fluctuate.”
Likewise, if we target NGDP, I would expect more volatile CPI numbers than if we target CPI. That’s not a bad thing. And I would expect the ratio of NGDP/Monetary Base to be volatile. I just don’t see why we should care about that ratio (or money multiplier * velocity) in and of itself – we should only care about it in so far as it effects NGDP. But then we’re back to targeting NGDP by using the base to offset those changes.
7. October 2013 at 15:44
Mike, You need to read more carefully. In this post we are not discussing the zero bound case. Kling says that’s always true. Banks don’t park funds in ERs when not at the zero bound.
Negation, Good point.
7. October 2013 at 18:42
I didn’t refer to the zlb I was referring to the following paragraph:
“Another way to put this is that monetary and financial arrangements are endogenous with respect to Fed procedures. The financial markets will evolve ways to insulate the economy from what the Fed does.”
Most money is endogenous in the form of deposits as observed in the C/D ratio. If endogenously not enough deposits are created then because C/D is relatively stable the fed just creates excess reserves.
If people could directly hold and transact in base efficiently through electronic depository and payments systems instead of only physically carrying notes and coins it would be a different story. Base would dominate over deposits and then the fed could more directly influence broader monetary aggregates making money more exogenous than endogenous.
8. October 2013 at 05:27
Mike, The frequency that base money is used has absolutely no bearing on its importance in determining NGDP. What matters is how base money velocity reacts to changes in the base.
If your statement was not referring to the zlb, then it was totally wrong. Banks don’t sit on ERs when not at the zlb (or positive IOR.)
Recall that you said;
“The fed can create as much base as it likes and the banks will just park it in reserve accounts.”
That’s flat out wrong.
8. October 2013 at 06:09
“The frequency that base money is used has absolutely no bearing on its importance in determining NGDP. What matters is how base money velocity reacts to changes in the base.”
The demand for money (frequency) has a massive importance in determining ngdp IMO. If people could directly deposit and transact in base its demand would greatly increase. If it was increased into the hands of the broader public through the wealth effect, balance sheet effect and through portfolio re balancing NGDP would be very correlated to an increase in base because its velocity would also increase.
If we refer to the current system I agree with you but I am referring to what base can do without being under the unnecessary limitations of the current system.
“”The fed can create as much base as it likes and the banks will just park it in reserve accounts.”
That’s flat out wrong.”
The reason no excess reserves happened in the past before zlb is because the fed always increases base at the same speed and is not really being too proactive with the amount of base to target interest rates on base becuase money is mostly endogenous and whatever interest rate target it announces just enter the expectations of banks and it trades to that level. The main determinant of the rate on reserves is expectations and demand. If the majority of the money supply is endogenous then it doesnt really mean too much what the rate on reserves is IMO. We may not agree on this and I guess it doesnt matter so much anyway.
8. October 2013 at 06:49
[…] Scott Sumner writes, among other things, […]
8. October 2013 at 08:42
ssumner and Mike.
I may have something wrong in my understanding of banks and reserves. I think every one of the statements below is correct.
The Fed controls bank lending by targeting bank liquidity.
Reserves increase when deposits increase.
Reserves are unchanged when the Fed sells Federal Debt into bank ownership.
Reserves and deposits decrease when Fed sells Federal Debt to private ownership.
Interest rates increase when Fed sells Federal Debt to private ownership.
The Fed counters interest rate increase by buying Federal Debt directly from Federal Government (or indirectly through agents).
I think that every statement is correct. Does my understanding agree with yours?
8. October 2013 at 18:20
lkdr, I don’t understand your first point.
Regarding ERs, the reason they were low before 2008 is that banks lost money by holding ERs when rates were positive. The Fed was not targeting the level of ERs.
Roger, The first statement is false, the second is true sometimes, not true other times. The third is false. The fourth is true for reserves, often true for deposits. The 5th is sometimes true and sometime false. The sixth is false, at least regarding buying directly from the Federal government.