A Puzzle About Demand Deposits
Whenever I graph M1 at FRED, it zig zags. I have just been taking that in stride, not really thinking about why it looks like that.
For my recent “inflationary big one” article, I charted demand deposits, which exhibit the zig zag pattern. So I finally decided to think through why the graphs look like that.
Since the zigging and zagging happen on a monthly scale, I figured it had to do with workers’ paychecks. For example, at the start of a period businesses pay workers with checks, and a lot of the workers literally “cash their paychecks.” So demand deposits go down, and currency held by the public goes up. Then during the course of the month, as the workers spend their cash, currency holdings go down while demand deposits go up.
Yet if this were the explanation, I would have expected the charts of currency and demand deposits to zig and zag in opposite directions, which doesn’t seem to be the case:
So is my theory right, and FRED’s aggregate of “Currency Component of M1” counts currency held in the hands of the public and in bank vaults? (In that case, it wouldn’t move around when being shuffled from the workers into bank holdings.) Or is my theory wrong?
What is the resolution of the underlying data? Is it daily, weekly or something else? By connecting two points on a graph it implies that there is a proportional transition between them over time, but this is almost never fully accurate. Is any data point the average or median of a set of individual readings?
By looking at the actual data we might find the answer to the zig zag.
Weekly, and it’s not seasonally adjusted.
As far as I know, hardly anyone gets paid monthly anymore. Either every two weeks or twice a month.
This probably isn’t correct, but is there any chance that some companies are borrowing in order to make payroll, increasing M1? The reason I do not think that this is a correct explanation is that I do not see why companies would regularly have a liquidity crisis at the same time every month.
The zig zag is meaningless: it has to do with how the data is collected and shown. Look at for example Chinese Current Price Gross Domestic Product: http://research.stlouisfed.org/fredgraph.png?g=1OP. Do you think this is a representation of reality or a feature of how the data is collected and presented?
Martin I’m assuming the Chinese GDP figure is an artifact of the reporting, I agree. But are you sure that’s what’s going on with demand deposits?
http://research.stlouisfed.org/fred2/series/WDDNS
http://research.stlouisfed.org/fredgraph.png?g=1P8(one year version)
http://research.stlouisfed.org/fred2/series/DEMDEPNS
http://research.stlouisfed.org/fredgraph.png?g=1P7 (one year version)
Here, compare the weekly to the monthly and see how many ups & downs there are.
It is a stock measure, so it matters when it is collected. If you have an up trend as is the case with this monthly measure the next down will probably be higher than the previous up and as a result you will have less peaks.
As you can see from the weekly, the fluctuation is fairly the same with an amplitude of about 80 billion. If you now have a look the way the data is collected:
“In order to avoid double counting those deposits that are simultaneously on the books of two depository institutions, the demand deposit component of M1 excludes cash items in the process of collection (CIPC) and Federal Reserve float. Demand deposits due to depository institutions in the United States and the U.S. government, as well as other demand deposits and CIPC are reported on the FR 2900 and, for institutions that do not file the FR 2900, are estimated using data reported on the Call Reports. Demand deposits held by foreign banks and foreign official institutions are estimated using data reported on the Call Reports. Federal Reserve float is obtained from the consolidated balance sheet of the Federal Reserve Banks, which is published each week in the Federal Reserve Board’s H.4.1 statistical release.”
This makes sense.
This is just my theory. I don’t work for the Fed so I don’t know whether this is correct.
I briefly thought about the groceries-theory or the salary-theory, but that doesn’t make much sense to me as companies would also be paying from demand deposits and grocers would be receiving on their demand deposits. The distribution within the stock would change, but the stock itself would remain the same.
Yes, I gave some thought to the stock over time as well as reporting. Seems plausible. I understand that if person X writes a check, all that is really happening is money is being moved from one institution to another. However, loan payments are cyclical, and when loans get paid the money supply does shrink.
You may be right that it is an issue of reporting. But, as my second option I would look at the cyclical nature of loan payments and their effect on money supply. But, I do not think, as Bob does, that it is due to people “cashing” their checks rather than depositing the funds. I just do not see that action as being significant or prevalent enough to cause such a drastic and cyclical change.
Wow. That is strange. If you count the peaks there are roughly 12 cycles per year.
Bob, maybe you aren’t looking at the right side of the equation. Instead of looking at payrolls, take a look at receivables (i.e. bills getting paid). I think that you may be looking at the wrong side of deposits, especially those that would be cyclical.
That’s what I’m saying though, Joseph. As people pay their bills through the month, businesses keep depositing the proceeds in their bank accounts. Then they give big checks to their workers, many of whom withdraw them as actual currency.
The paycheck theory doesn’t even pass the smell test. “Big checks” would have to be coordinated to always be during the same window each month.
The pattern becomes clear when you look at the % weekly change. The monthly drop in demand deposits is nearly always between the 6th and the 11th. In months where Monday falls on the 5th and 12th, there are drops on both.
Looking at non-M1 M2, total savings is nearly always up more than demand deposits are down during the drop week. There’s something else going on that likely relates to corporate balance sheet maintenance.
As to the M1 Currency component, it is “FR Notes, net” from H.4.1minus FR Notes held by Treasury (disclosed by Treasury) minus FR Notes held in bank vaults from H.3.
EB wrote:
The paycheck theory doesn’t even pass the smell test. “Big checks” would have to be coordinated to always be during the same window each month.
To repeat, I’m not saying I’m sure I’m right, but your guys’ criticisms are goofy. You’re objecting, EB, on the grounds that people don’t tend to get paid at the same time during the month?
Yes. Based on my admittedly anecdotal experience, people get paid at different times all throughout the month. Bi-weekly, monthly, weekly, and also on different days of the week. I work for numero uno, so maybe I’m out of the loop.
Your theory seems to suggest that there is a critical mass that gets paid regularly only once a month. Even if that is true, why would it always be in the same, narrow window?
I don’t know if this helps at all but the number goes up almost every time for the 15th of the month. For example, 1/10-1/17 it went from 510.1 to 532.2 and 2/14-2/21 it went from 526.5 to 534.1.
It also rises every time this year on the last reported day of the month from the previous week. On 1/24 it was 531.0 and on 1/31 it was 555.2, also on 2/21 it was 534.1 and on 2/28 it was 548.5
I guess that would be the opposite of your theory above. It seems like when people get paid on the 15th and end of the month the DD rise. Maybe it means that when people get paid they deposit their checks and as they spend their money from there the DD start falling. Maybe businesses aren’t immediately depositing the money when people are buying their stuff and instead tend to deposit it just before they pay their employees.
You kind of threw me off with the statement about workers actually “cashing their checks”. I am under the impression that it is your opinion that the cyclical nature of the graph is due to cash balances. Is this the case? While I concede that this is true in some cases, a great example being the spike in late 2008 when many people went into cash (and the dollar rallied), I do not see that as being prevalent today. I think that overall deposits are more indicative and that the cycle shown can be explained by such. As for the general upward-sloping trend, that is merely deposits being multiplied.
The above comment is to Bob Murphy.
I’m not saying I’m sure what’s going on, but just to be clear: If a worker cashed a check, the the red line would go down. The employer’s bank balance would drop, and the worker’s wouldn’t go up (because he would be holding the money in the form of paper currency in his wallet).
It is hard for me to accept that it is the cash withdrawals of “worker’s” that would cause the downward-trend in the 1st interval in the cycle. Rather than looking at the cause of the downward trend in the interval, I would look at the cause of the upward trend. In my understanding, the only thing that could create this upward trend in each interval of the cycle is an increase in deposits.
Sure, cash tends to expand with deposits over time, but it is the deposits that allow this to happen. If demand for cash was the prime driver of the trend, then wouldn’t the graph trend downward?
How about this…
The employers pay their workers. The workers checking accounts go up (not currency holdings.)
The various accounts where the businesses have their funds swept go down. While in reality
they funds are in checkable deposits, they aren’t measured as checkable deposits. So, you should find the opposite fluctuations in weekly not seasonally adjusted — repurchase aggreements, money market savings accounts, etc. Those things being much larger, the fluctuations are less noticable.
And, of course, we don’t get weeking estimates of repurchase agreements.