Friday, January 1, 2016

Debt in Context: Counterfactuals and the Common Trend


Graph #1: M1 Money (blue) and GDP (red) as Context for TCMDO Debt
I've been looking at this graph for a while now, and I'm starting to see things. In fact, I briefly considered the title Hallucinations for this post.

No no, just kidding.

The graph shows total (public and private) debt relative to GDP (in red) and relative to circulating money (in blue).

What I see is significant similarity between the two lines. Oh, sure, from the mid-1960s to the mid-1980s the red line runs flat, and the blue line rises all the while. But then in the mid-1980s the red line does a rapid catch-up. And after that, the red and blue run parallel again. Just like before the mid-1960s.

Similarity fails again in the early 1990s when the blue line falters. But then the blue does a rapid catch-up. By 2001 the red and blue run parallel again.

In this graph I see two lines remarkably similar, except when something temporarily pushes one of them away from the pattern both lines seem to desire.

I know: Lines don't "desire" patterns. Still, one or the other of these two lines is always on the gradually increasing rise to peak; and whichever line departs that trend returns quickly when opportunity presents.


Most of the 30-odd years before 1990 are taken up with the red line's failure to run parallel to the blue. Half the years from 1990 to the crisis are given over to the blue line's failure to run parallel to the red. Even I can see these things. There are only five or six years before 1965, four years between 1986 and 1991, and half a dozen years from 2001 to the crisis when the red and blue seem to run together. I can see.

The strength of a common trend is not always shown by how well the lines cling to that trend. In this case it is shown by how aggressively the lines return to trend, even after a prolonged absence.

I've been looking at this graph for a while now.


The common trend is due to the increase of debt. The departures from trend are due, or mostly due, to the context variables -- GDP and circulating money.

Knowing that much, and knowing some of what happened during the years shown on the graph, we can make at least one observation with confidence. First, though, take a moment to look at some of what happened in those years.

The years where the red line runs flat and the blue line doesn't -- the mid-1960s to the mid-1980s -- are often referred to as "the Great Inflation" by people who know about such things. Allan Meltzer dates the Great Inflation as the period 1965-1984.

Source: https://research.stlouisfed.org/publications/review/05/03/part2/Meltzer.pdf
I don't know about Meltzer's analysis of the origins of the Great Inflation, but I am happy to use his start- and end-dates. Here's Graph #1 again:

Graph #1 Repeated: M1 Money (blue) and GDP (red) as Context for TCMDO Debt
The very small peak in the red line, visible in the mid-1960s, is the point after which accumulated debt no longer increased faster than GDP. That very small peak occurs in the last quarter of 1964. The effect of inflation is first seen on the graph the following year, 1965.

Meltzer's start-date for the Great Inflation is 1965. The dates match.

After the 1982 recession, the red line runs almost perfectly flat for a bit, then rises rapidly. That flat bit ends in the first quarter of 1984.

Meltzer's end-date for the Great Inflation is 1984. The dates match.

The departure of the red line from trend between 1965 and 1984 was due, or mostly due, to GDP. And the reason for the departure from trend, we can say with confidence, is that GDP grew unusually fast in those years because of the inflation.

Because of the inflation. Each year during the Great Inflation, GDP was higher than the year before because prices were higher. (Real growth contributed to rising GDP also, of course, but during the Great Inflation, the price increases were unusually large.)

Each year during that inflation, GDP was higher because prices were higher. New spending on credit was higher for the same reason. This made new additions to debt larger than in the past. But existing debt, prior year debt, was not similarly enlarged. Only a part of the total debt was enlarged by inflation. Only part of the debt, but the whole of GDP was enlarged by inflation.

During the Great Inflation, then, GDP grew faster (relative to debt) than it did before or after that inflationary time. The unusual growth of GDP kept the red line below trend until the end of the Great Inflation.

So much for the red line's departure from trend. As for the blue line, well, the blue line shows debt relative to the quantity of circulating money. So the fall in the blue line is the result of a slowdown of debt growth or a speed-up of money growth, or both. And the blue line's return to trend is due to a speed-up of debt growth or a slowdown of money growth, or both.

Note, however, that a significant change in debt growth would affect the red line as well as the blue. So if you see a change in the one line but not in the other, then a change in debt is not the primary cause of the change that you see. The cause is a change in the context variable.


A few days before I found Simon's thoughts on "context" at Mainly Macro, I was looking at this graph:

Graph #2: Circulating Money relative to GDP, with an Interruption in the Downtrend
Graph #2 shows M1 money relative to GDP. It is just what you'd get if you took the data from Graph #1 and divided the red by the blue. But #2 is a crappy, clunky graph and I don't like it.

It is a most interesting graph, because it shows how much spending-money we have, compared to the amount of new stuff we buy every year. The amount of spending money went down for a long time. It fell from more than 25 cents for every dollar's worth of stuff we bought in 1960, to less than a dime just in time for the crisis. After the crisis, the trend changed and the line started going up.

It is a most interesting graph, because money also went well above trend in the 1980s and '90s. Above trend, and going up instead of down.

It is interesting that we see uptrend twice on Graph #2, and that at least one of those uptrends is part of the policy solution to a big economic problem. Actually, I see both uptrends as parts of policy solutions. I don't know whether uptrend was conscious policy in the 1980s and '90s or just an accidental result of policy. Either way, though, it led to the "exceptional growth" of the latter 1990s.

But that dull red line, the line I put there to show the down-trend and to emphasize the departure from that trend, that awkward line makes the graph clunky and crappy. So that's a graph that didn't get used till now.

Anyway, this is where my mind has been -- looking at disturbances in trends, and trying to see the trend despite the disturbance. This line of thought fascinates me.

I want to make a better graph.


Back to Graph #1:

Graph #1 Repeated: M1 Money (blue) and GDP (red) as Context for TCMDO Debt
I wanted to generate some fake numbers -- sorry, some "counterfactuals" -- for the red line, where the blue is rising on trend but the red runs flat and then does a quick return to trend. For the blue line also, where it departs and returns to the common trend.

Thought about it a bit, and decided to use the blue numbers as a proxy for the red where the red line misbehaves. Take the blue numbers and scale them to fit the red line and plug them in, to create counterfactual data. Likewise, use the red line to fake some on-trend numbers for the blue line. I got the FRED data for Graph #1 and set to work.

One thing was certain: I didn't know where I wanted to start or end the counterfactuals. So I set up cells where I could type in start- and end-dates for my made-up values. So that if I just change the dates it would change where the counterfactuals are used, and the graph would change.

That idea worked. If you're interested in such things, you can download the Excel file and take a look. It's all done with worksheet functions. No Visual Basic.

Here is the resulting graph:

Graph #3: The Actual (faint) and Counterfactual Ratios
Now it is easy to see the common trend.

Contemplate this graph, and you will observe the following facts:

1. The fall from trend of Debt-to-GDP (red) was due to the Great Inflation.
2. The fall from trend of Debt-to-M1 (blue) was due to a slowing of debt growth after 1985 combined with a significant bump-up of money growth in the 1990s.

The facts lead to these conclusions:

1. Debt-to-Income ratios typically show little increase in the years before 1980. This is not because debt was growing slowly, but because inflation enlarged the denominator.
2. The exceptional growth of the 1990s followed from the financial cost savings arising after the fall from trend of the blue line.

If you read this post as many times as I proofread it, it will make good sense.


Access Counterfactuals and the Common Trend.xls via Google Drive.

1 comment:

The Arthurian said...

Some three and a half years back, Joshua Wojnilower observed that "debt growth has remained pretty stable over time" and pondered "why the rate of debt growth would be so consistent over time despite changing conditions".

He sees it, too.