Wednesday, January 9, 2013

Art history


The first graph from yesterday's post:

Graph #1: Total Debt per Dollar of Spending Money, 1916-1989

Dated 1993, twenty years ago now. (Three ring binder holes are visible if you enlarge the image.) Printed on a dot matrix printer. Does Epson still make printers?

The graph was created using a program I wrote myself in the C language oh, so many years ago. That's how I got the vertical lines at key points, like 1933 and 1946.

See how the graph starts with 1916? That's a hint about the source data. Back when I started gathering economic data, in the late 1970s and before the internet, I would go to the library and scour the shelves. Eventually I found the Historical Statistics of the United States, Colonial Times to 1970 -- the Bicentennial Edition.

These days, you can get that online along with annual updates called Statistical Abstracts -- but they're no longer collecting data or issuing the Abstracts. You know: budget cuts.

Anyway, back then I relied on the Historical Statistics. For the Debt-per-Dollar graph, if I remember right, for debt I used Series X 393 and for money, Series X 414. The one covers the years 1915-1970 and the other 1916-1970, so I ended up with a graph that starts at 1916.

The Debt-per-Dollar graph was one I found quite interesting, because it showed just three strong trends. So I would go back to the library every once in a while, pull out the latest Statistical Abstract, and append some data to my files and my graph.

Each Stat Ab provides a few consecutive years of data, five to seven years maybe, the most recent values. It also shows older values, but only for every fifth year. So the Stat Ab for 1978 might provide values for 1955, 1960, 1965, and 1970-1976.

Somewhere along the way, they changed the way they figure the numbers. On my old graph you can see a high point at 1960, a high point at 1965, and higher numbers from 1970 and after. I have lower values for the other dates of the 1960s and for before 1960. The high points and the high later values come from editions of the Statistical Abstract. The lower values come from the Historical Statistics.

I never knew what to do about that discrepancy. I just went with the newest numbers that were available, and ended up with obvious high points on what should have been a smooth line.

Between 1980 and 1989 on the old graph you can see a similar disturbance in the values. I always figured it was due to data revisions or my own mistakes. But can see the same disturbance in the FRED graph below, along with a much larger disturbance from about 1991 to 1994.

Graph #2: Total Credit Market Debt per Dollar of Spending Money, 1959-2012
Still with me? Good.

The old graph -- I did dozens of graphs by hand (with a pencil) when I was getting started with this stuff. But the Debt-per-Dollar graph struck me immediately as significant because it showed just three strong trends: a developing problem, a period of recovery, and a golden age.

The new graph had wiggles in it that I taught myself to ignore on the old graph. Even the big wiggle there, 1991-1994. Ignored it. But then one day I was reading about periods of good economic performance and I put one and one together.

The old graph shows a long downtrend from 1933 to 1946. Economic performance was good starting the next year, from 1947 to 1973.

The new graph shows a brief downtrend from 1991 to 1994. Economic performance was good starting the next year, from 1995 to 2000.

The economy is good for a while, immediately after the burden of debt is reduced for a while. This cannot be coincidence.

The new graph provides confirmation of a hypothesis arising from the old graph. That hypothesis: Before the economy can grow again, debt-per-dollar must fall.

We must reduce total debt relative to circulating money. Still with me?

4 comments:

jim said...

Hi Art,

Have you looked into what sweeps do to your debt/dollar graph?

http://www.sweepmeasures.com/sweeps.html

The way I understand it in 1994 the banks were allowed to change the way they figured the data that goes into computing M1. Banks were allowed to shift the money from accounts that were less active from checking to savings for reporting purposes. Doing this lowered their reserve requirements.

http://research.stlouisfed.org/fredgraph.png?g=eps

The Arthurian said...

No, actually.

Philip George talks about sweeps a lot. But I never knew what it was.

Makes more sense, now that you define it for me. Thanks.

http://research.stlouisfed.org/fred2/graph/?g=epz

M1 adjusted for sweeps is higher... because M1 is figured with sweeps removed (as you point out) to make M1 balances lower FOR REPORTING PURPOSES
I get it now. Banks report the lower number in order to reduce their reserve requirement. Got it.

So really, M1 is higher (since 1994) than the M1 I've been using.

Thanks Jim. I have to look at this more. (But for the record, everything I said about BEFORE 1994 still holds up.)

The Arthurian said...

ps, Thanks for the sweepmeasures link.

jim said...

Hi Art,
My understanding is that a bank can sweep funds out of checking accounts based on the amount of transactions that the account had in a given period. It sounds like if you are not using your checking account money, the banks are allowed to use it.

According to Phillip George the swept money just disappears from monetary aggregates, but I gather from elsewhere it is used in Money Market Mutual Funds which I think would be part of M2.

It is interesting that MMMF was the very eye of the storm in the 2008 financial meltdown.

-jim