The Non-Relationship Between Interest Rates and the Money Supply
by Mike Kimel
This piece has been cross-posted at The Presimetrics Blog.
The Non-Relationship Between Interest Rates and the Money Supply
Figure 1
The graph shows that all but one recession since 1948 was preceded by a big drop in the real money supply per person over the length of a year. The exception – July of 1953 – wasn’t much of an exception; the 12 month change in the real money stock per capita went negative in August of 1953 and remained negative throughout the length of that recession.
Now, I use real M1 per capita a lot in my posts; I think it’s a largely unused but very good (in part because it is largely unused) measure of monetary policy, and its one Michael Kanell and I use in
But because it isn’t used much, every time I do use it, I find there is a bit of confusion. And because it is so useful, I think it is worth spending a bit of time covering it.
I’d like to start with a question I get asked a lot, which will be the topic of this post: what sense does it make to use real M1 per capita (or any other measure of the money supply) as a way to monitor the Fed’s monetary policy? After all, what the Fed does is set interest rates, or rather, one interest rate in particular: the Federal Funds rate. And when it does that, the money supply gets set by default. A fair number of college seniors majoring in econ, when pressed, could probably tell you a story like this:
The Fed determines what the Federal Funds interest rate should be. If the Fed wants to reduce interest rates, it will create money out of thin air and use it to buy bonds. Because there is more money competing to buy bonds, the interest rate bonds have to pay falls. At the same time, the added money sloshing around becomes cheaper for anyone to borrow, whether they’re issuing bonds or not. On the flip side, if the Fed wants to raise interest rates, it sells bonds. That forces anyone else trying to sell bonds to raise interest rates to compete. Additionally, because the Fed retires the money it gets paid for the bonds it sells, that process reduces the amount of money available in the economy, making it harder to come by and hence more costly.
It’s a nice story, and like many others that is taught in economics classes across the country, it works fine in theory. It may even apply in a lot of real world situations. As I will show below, however, it doesn’t have much if anything to do with the way the U.S. economy operates.
Now, I’ve got a lot of ground to cover today, and being charge of watching the newborn while the wife is out and about, I don’t have a whole heck of a lot of time, so whereas I’d normally be throwing up a lot of graphs to show what I mean, instead I’m going to tell a little story and then show you that the folks responsible for setting the money supply agree with the story.
Here’s the story: Americans like to shop, and they seem to really like to shop in December. (Apologies for the lack of a graph, but you can find one example of the data here.) Shopping is easier when there’s more liquidity in the system – that is to say, when money is looser and easier to come by – so the Fed should be expected to loosen the real money supply somewhat to accommodate the end –of-year shopping season. On the other hand, as anyone who watches the news can tell you, the Fed doesn’t exactly move interest rates around in a way that would suggest any loosening around December each year with a subsequent tightening up early in the next year. Which means either: a) the Fed does nothing to accommodate the end –of-year shopping season or b) the money supply and the interest rates, at least within some not-so-narrow range, have very little to do with each other.
Now, if the Fed is trying to loosen the money spigot to help the shopping season along every year, it isn’t doing it with interest rates. Interest rates don’t seem to display much of a pattern when it comes to the calendar. You don’t hear anyone say, “I’m waiting until to December to refinance the house since rates are always lower then.”
What about real M1 per capita? Well, there we see a pattern. Using data from January of 1948 to December of 2008, we can construct 721 blocks made up of 12 consecutive months. The first of these blocks runs from January of 1948 to December of 1948, the second from February of 1948 to January of 1948, and so forth, until January of 2008 through December of 2008.
The graph below shows the percentage of these blocks in which each month had the highest real M1 per capita. For example, the percent of blocks for which the peak real M1 per capita for the block occurred in January is shown in the first bar, the percent of blocks for which the peak real M1 per capita for the block occurred in February is shown in the second bar, etc. So look what happens:
Figure 2.
Clearly, by far, the real money supply per capita is looser in December than any other month, and by a degree well outside the range that anything resembling chance would support. The fact that January comes in second is probably a function of spillover from December – the Fed doesn’t always drain out the money supply as quickly as it might otherwise like.
So… to sum up what we have… the Fed has a reason to loosen the money supply in December. Interest rates don’t show any pattern that works with the Fed’s goals, but real M1 per capita does. This indicates that unless there’s some really big coincidence going on, interest rates and real M1 per capita do not move together and the Fed is/seems to be using real M1 per capita (or something very similar) to accomplish many of its liquidity goals.
While I tend to let data speak for itself, I’d like share a couple of quotes. Here’s the first
Because of these difficulties in achieving a subtle response of the Federal funds rate to changes in the amount of borrowing, achieving the degree of reserve pressures specified in the directive has been interpreted since the late 1980s to mean creating conditions consistent with the FOMC’s desired Federal funds rate. That rate has generally been apparent to the banks; since 1994 it has been announced formally and in prior years it was clearly indicated through an open market operation.
The rate has tended to move to the new, preferred level as soon as the banks knew the intended rate, with little or no change in the amount of borrowing allowed for when constructing the path for nonborrowed reserves (described below).
I bolded a key sentence in the above paragraph; it indicates that the Federal funds rate is set not by the buying and selling of bonds, as textbooks will tell us, but rather by the Fed’s wish. Since 1994, the Fed has announced its target rate, and the rate starts moving in that direction immediately upon the announcement, not waiting for open market operations. Prior to 1994, the target rate wasn’t announced, but banks tried to get figure out that target through statements made by Fed officials and head to that rate anyhow.
A second quote from the same source:
A prominent seasonal factor affecting deposits is the buildup in balances to accommodate the extra transactions during the holiday period, stretching from late November to early January (and the sharp reversal during January). A shorter term seasonal pattern arises from the payment of social security benefits on the third of each month; most recipients allow their cash balances to rise initially, then gradually work down the deposits as they pay their bills. (The Treasury’s total cash position might show offsetting movements, but most Treasury cash is not subject to reserve requirements.3)
This quote indicates that the Fed does, indeed, try to accommodate additional seasonal liquidity (including those in December I noted in Figure 2).
So, these quotes match the data I described and support my conclusion. Which is to say, I’ve found another heathen who refuses to accept the standard textbook monetary policy story about how the money supply and interest rates behave. So where do these quotes come from, you ask?
You’ll find the quotes I cited on pp. 141 – 142 and p. 142, respectively, of the chapter on Open Market Operations in a textbook on monetary policy produced by the New York Fed. That is to say – the people directly responsible for the buying and selling of bonds on behalf of the Federal Reserve aren’t don’t seem to feel that their activities have anything to do with the setting of interest rates. But what does it say about the economics profession that the rest of us are under the misimpression that it does?
—
I’ll have at least one more post soon on real M1 per capita and its effect on the economy.
—
Data Sources
FRED, the Federal Reserve Database, was the source for most of the data used to compute real M1 per capita: population from 1952 to the present , M1 from 1958 on and M1 from 1958 on.
Quarterly data on real GDP per capita and population. Note – the quarterly population figures were used to extrapolate monthly population for 1947 to 1952.
Finally, money stock figures were substituted in for M1 from 1947 to 1957. Those were copied by hand from this document at the
Mike,
Interesting and timely post in view of my current focus on the theory of money. I’ll be poring over it carefully. Thanks.
Here’s a question for you: does anything change when you use real M0 per capita instead of M1 per capita?
I still feel burned by Milton Friedman’s sleight of hand of hand in using M1, which the Fed only influences not controls, to falsely conclude that the Fed caused the Great Depression. If you look at M0 for the period in question, you’ll see an M0 expanded even as M1 contracted. So, you could reasonably conclude that the Fed could have done more to prevent M1 from contracting, but the fact is that M1 contracted due to decisions by banks not to lend and depositors to withdraw their money, which were out of the Fed’s hands.
A model that sees the money supply as something controlled by the Fed, rather than influenced by the Fed, largely misses the point.
Money is a natural phenomena. It will arise with or without government. In its broader definition, it is created when negotiable debt is entered into, and is destroyed when negotiable debt is paid off. Reserve requirements of commercial banks and the tendency of banks to push up against them strongly influences the tendencies of the money supply, but money supply creation and destruction is a far more democratic process than the popular business press would suggest.
M1 is a narrow definition of money: It includes all coins and currency held by the public, traveler’s checks, checking account balances, NOW accounts, automatic transfer service accounts and balances in credit union accounts.
Viewed definitionally, the December surge in M1 is hardly surprising. At Christmas, bonuses are paid and people divert money from non-M1 definition money sources to buy stuff. Some of that money is borrowed or comes from cashing out investments. Spending happens. Debts are paid off, unspent money diverted from investments is reinvested. The world returns to normal.
Indeed, it is quirks like the December surge with its January hangover that have made broader measures of the money supply (like M2 and M3) more popular for policy purposes. The April mini-surge is probably related to cashed out investments or borrowed funds to pay taxes, and to the receipt of refunds that aren’t immediately disposed of through consumption or investment.
M1 isn’t irrelevant, but as a true measure of the inflation driving money supply, it is a bit like trying to measure net worth using cash in your checking account and wallet as the sole proxy variable. While the two aren’t unrelated and indeed in each case the one goes into the calculation of the other, in both cases they aren’t the most important or tend indicating components.
M1 per capita is only weakly linked to inflation because it measures only a small share of the total amount of money circulating in the economy and heuristically, the value of the unit of currency is basically the total money supply divided by the total value of goods and services circulating in the economy.
Its uselessness in estimating inflation doesn’t make it useless. M1 per capita is essentially a measure of anticipated short term expenditures, a concept that has considerable value in day to day trading activities and short term economic forecasting.
Wow this is a foolish post. Yes there is no interest rate change in Q4 because thats why the money supply was increased. Demand for money increases in Q4, in order to meet this demand and keep the interest rate (price of money) constant the Fed increases the supply. All your data shows is the Fed does its job well.
This isn’t really Megan McArdle coming to have a laugh is it?
@ohwilleke,
If there is no relationship between interest rates and the money supply, as Mike urges, does the Fed even influence the money supply?
Good observation about the effect of holiday bonuses on M1.
Tao J,
1. The Fed doesn’t entirely control anything. Think of it this way… the Fed is like the owner of a Ferrari. He can step on the accelerator, but and that will generally make the car run. But the degree to which it does so depends on a number of factors, ranging from the amount of gas in the car, road conditions, how tuned up the vehicle is, and whether its chained to the nearest tree. As to the difference in “control” between M0 and M1… don’t underestimate the power of jawboning.
2. I want to hold off on explaining why M1 and not M0. Suffice it to say that M1 is more related to economic growth. FWIW, we do touch on the reason in a footnote in the book. I want to eventually get around to something more in depth in blog form eventually, but its on a big to do pile.
3. Do you remember where you saw data for M0 & M1 going back to 1929? I have not seen that data and would like to have it.
ohwilleke,
1. I’m not assuming that the Fed controls any of the variables completely, except the FF target. (See the explanation to Tao Jonesing above.)
2. I don’t follow what you are trying to get to about measuring inflation with M1. Obviously, the Fed doesn’t have perfect knowledge of what inflation is at any given time, but they’ve got a good enough idea that they aren’t off by much.
Rob,
To quote the NY Fed again:
“The rate has tended to move to the new, preferred level as soon as the banks knew the intended rate”
Notice that they said nothing about the rate waiting until a few hundred million bucks worth of bonds get bought or sold. In fact, even a casual observer can see a frenzy of activity in the markets which begins literally within seconds of a Fed announcement, whether the NY Fed’s trading desk is doing anything or not.
So… what you are saying contradicts a) what the folks in charge of the open market operations that are the biggest piece of the money supply equation believe and b) what happens in the markets whenever the Fed makes an announcement. Given that, I think its incumbent on you to trot out a bit more evidence in support of your position.
Tao J,
Of course the Fed has a huge influence on the money supply. The NY Fed’s trading desk pulls out and puts in huge sums with its open market operations. Then you have required reserve ratios. You have jaw boning. A friend of mine, then at the FDIC, used to attend some of the monthly meetings and used to tell me about them. Suffice it to say, when the Fed says, “you know, we think its probably a good idea if banks in the Kansas City region cut back on making loans by 0.7%, the banks get the message instantaneously. And when they cut back on making loans or increase their loans on the “suggestion” of the Fed that also has a big impact on the money supply.
No all that’s saying is that expectations matter and the Fed has credibility. So markets move to adjust to the new Fed target right away. All these facts fit squarely within monetary theory. If the Fed didn’t have credibility adjustment would take longer and would take heavy lifting by the Fed.
Also this assumes that the only way the Fed can get the money supply to increase is by their bond market transactions. But by having financial balance sheets adjust to new target rates money can be created or destroyed.
All of this is difficult to deal with because neither money demand nor velocity is constant and so money supply will bounce around.
Rob,
“Also this assumes that the only way the Fed can get the money supply to increase is by their bond market transactions.”
No it doesn’t. See my comments upthread. In general, these operations are merely the most important of the tools available.
“All of this is difficult to deal with because neither money demand nor velocity is constant and so money supply will bounce around.”
Exactly. So imagine how difficult it would to use money – whose demand also fluctuates every month (the Social Security check thing) and seasonal – to keep a targeted FF rate.
Rob,
Sure, the Fed has credibility. Everyone knows if the rate doesn’t go where they want it, they will force it there. Which in turn means they don’t have to use the tools they would otherwise use to force the FF to where they want to move the FF. They can use it for something else.
There is nothing else. Monetary policy is the interest rate, full stop.
Why is it hard? Again they can do it because they are willing to do it and are expected to do it. Same way they maintain exchange rate bands.
Mike,
You can probably get the older M0/M1 data from Steve Keen, who uses it for one of the charts (Figure 11) in this post:
http://www.debtdeflation.com/blogs/2010/07/03/are-we-it-yet/
He’s pretty good in providing the underlying data when you ask for it (he actually uploaded some data at my request within a matter of hours).
It is interesting that FRED has M0 back to 1918 but M1 only back to 1959.
You can see charts of the M1 data in Friedman’s and Schwartz’s “A monetary History of the U.S.” but I’m not sure how helpful that is . . .
A 1990 Fed paper found that banks extend credit 9-12 months before the corresponding reserves are deposited. Perhaps that explains why interest rates move immediately instead of : banks determine money supply through their lending practices, and any changes in the money supply caused by changes of the target rate by the Fed won’t show up for 9-12 months.
http://www.minneapolisfed.org/publications_papers/pub_display.cfm?id=225
Steve Keen’s cogent analysis:
http://www.debtdeflation.com/blogs/2009/01/31/therovingcavaliersofcredit/
Mike,
Let me rephrase, does the Fed’s setting of interest rates have any influence on the money supply? It seems that your answer would be no. “Non-relationship” implies both correlation and causation are missing.
FYI — In addition to the spreadsheet data that Steve provided me re: GDP and debt (that is what I asked for), he pointed me to the Census website for the official data, which I was able to find and confirm the accuracy of his spreadsheet. He is pretty scrupulous about using official U.S. data, so I trust that he can help you track down the ultimate source of his money supply data.
Thanks. I’ll get ahold of him later this week when I’m less pressed for time. But I’m a bit concerned if it originates with the Census. The Census publishes and kind-of-official-kind-of-collaborative-with-academics tome on Statistics of the US; its sort of a historical offshoot of the Statistical Abstract of the US that they publish each year (which I find is an amazing source).
The historical statistics book, though, is largely ex-post extrapolations. Now, their ex-post extrapolations are great for most purposes, but I tend to avoid using them where possible. There’s enough of a “trust-me” to contemporaneous data that I get very nervous about series that were filled in decades later. I think the quality of most of these series in the Historical book are probably pretty good for broad strokes, but that’s as far as I’d throw ’em without seeing precisely how they’re computed.
Tao J,
My answer would be no, within a range and we’ve been in that range for most or all of the period for which we have data. Now, if the Fed were to decide to move the FF to 12.5% tomorrow (notice where the decimal point is) things would be different.
Would you care to elaborate on your assumptions regarding B2*x2+B3*x3+B4*x4….
Whoops, you don’t want to go there. I love how an ill-defined model with negative correlation is sufficient to disprove causation.
I got a model that would fit the Presimetric theme. Regress population growth during a presidency against a binary variable defined by whether the president’s last name begins with A through M or not.
We do understand that the whole December (and January) thing has to do with people putting cash in their pockets to shop? Taking cash money out of banks means that banks need cash money, which the Fed provides. M1 is a narrower definition of money than M2, with cash a much bigger share of M1 than M2, so Fed efforts to provide cash for shopping leads to a spike in M1.
The Fed does not aim to loosen policy in December. That makes your choice of vocabulary a little suspect. You says the Fed aims to “loosen” money supply rather than “increase” it. Increase is objectively, narrowly true. “Loosen” is typically used to describe policy, rather than the amount of M1 in the system, so using it to describe a purely technical, seasonal adjustment by the Fed seems a bit misleading.
You are exactly right that the Fed isn’t attempting to use rates around Chrstmas to “loosen the monetary spigot” because their is no spigot, and there is nothing to “loosen”. The Fed is trying to balance cash supplied against cash demanded so that banks don’t run out of cash. It’s just a cash thing, not a policy thing. Policy is not “looser” in any meaningful sense around the holidays. Your statement that in January, “the Fed doesn’t always drain out the money supply as quickly as it might otherwise like” partakes of the same mistaken thinking. There is a bunch of shopping in January, and that means people want cash in their pockets, so the Fed leaves cash available.
And yes, the whole bit about announcement effect is true, too. That doesn’t mean the Fed doesn’t use bond purchases and sales to achieve its monetary policy goals. It simply means that announcing a funds rate target makes policy more transparent, and that the Fed is pretty good at keeping its reserve maintenance activities consistent with the funds rate target. Bill and coupon passes are common during normal times, when the Fed is easing or maintaining steady rates, and reverses are common when the Fed is tightening. Leaving out observations regarding reserve maintenance doesn’t mean reserve maintenance doesn’t matter. It just means you left it out. This post seems mostly a confusion over language and an incomplete examination of Fed operations.
For a look at the Fed’s reaction to holiday shopping from 20 years ago, take a look at “Does the Fed Cause Christmas?”
http://clevelandfed.org/Research/commentary/1990/0101.pdf
Mike, I find this post less than useful because it sort of confutes short- and long-term effects — or more accurately ignores long-term effects. Do interest rates follow M1 (averaged over some period) some x lag period(s) later?
It’s like driving a supertanker…
Jay,
What model? What negative correlation?
I am pointing out that there is no correlation between the FF and the money supply. I am also pointing out that the NY Fed puts out a book that, when discussing its own trading desk, says that their trades (which are the main tool for affecting the money supply) do not affect the FF.
To me, those two pieces of evidence seem enough to conclude that when you move the MS, you aren’t necessarily moving FF, and vice versa. If you believe the opposite, explain how, put up some data, and then get on the horn to the NY Fed’s trading desk because there’s a lot at stake here.
kharris,
I was too loose with the term loose. Point taken.
That said, I don’t think its all an accommodation story. I don’t want to make the argument for money supply affecting economic growth here since I think I can make it much better with a post I have in mind.
BTW… I did read the Cleveland Fed paper a long time ago. It has a flaw which I’ll try to remember to cover when I write the post described in the previous paragraph.
Steve,
I was playing with data… nothing formal, very quick. But… I looked at the one month change in real GDP per capita and how it correlated with the real FF rate and the one month change in real FF.
I looked at the correlation for the same month, 1 mo lag, 2 mo lags, 3 mo lags, 6 mo lags, 12 mo lags, 18 mo lags, 24 mo lags and 36 mo lags, all of these in both directions. Nothing.
Then I did the three month change in real M1 per capita and real FF. Ditto. Then the six month change, and then the 12 month.
The best I’m finding is a 40% correlation… between the real FF rate and the 12 month change in real M1 per capita, with the latter lagged 36 months.
Now, this was quick and dirty, so maybe I’m making a mistake (drop me a note if you want the spreadsheet) but this really is not showing any evidence of changes in the money supply affecting interest rates over any reasonable time frame.
And frankly, I really don’t see why it should. Yeah, I had the same theory classes as everyone else, but it doesn’t make any sense. This isn’t like corporate bonds and money the corporation raises.
Mike I know this is hard to grasp, but you are looking at a complex system. You will need more than 1 RHS variable and I doubt the relationship is linear to begin with.
Jay,
I know all of that. I also know that the folks who have been implementing the system since its inception – the trading desk of the NY Fed, seem to feel the same way I do, as per the quotes above.