Wednesday, May 25, 2016

Less aimless: a closer look


In the previous post I showed household debt as a portion of total private non-financial debt. About half, but it varies, and it varied higher on the approach to crisis. Back in the normal range now, high, but high in the normal range and trending down.

I highlighted two peaks that look similar. They show similar rates of increase on the approach to peak. They show similar rates of decrease on the decline from peak -- and faster decline than increase for both.

In addition, they both peak near 0.54 on the vertical scale. And they both occur during times of superior economic performance. I'm not saying anything about cause or effect or significance; it's far too soon for that. I'm just pointing out similarities.

Graph #1: Household Debt relative to Total Private Non-Financial Debt
highlighting similarity between the 1960s and the 1990s
Looking at it a little more, the peak in 1980 shows the same pattern of rapid increase followed by more rapid decline. This peak is obviously lower than the other two, and people don't often describe the 1976-1982 period, say, as a time of superior economic performance. Still, the pattern is similar.

So I thought I'd take a look at all three peaks. I started out by bring the FRED data -- the "total credit to private non-financial sector" series is copyright BIS (but FRED is down again so I can't get their preferred copyright claim) -- bringing it into Excel and using Kurt Annen's Hodrick-Prescott code to smooth out the jiggies:

Graph #2: Ratio of Household Debt to Total Private Non-Financial
Debt (blue) and the Hodrick-Prescott Trend (red)
NOTE: This is Quarterly Data. The X-Axis Seems to Imply Monthly.
Next I changed the blue FRED data to gray so it doesn't stand out. I changed the red Hodrick-Prescott line to blue, to make it our starting point. And in red I added three subsets of the H-P data, one for each of the three peaks discussed above.

Those peaks occur at 1964Q2, 1979Q4, and 1996Q3. Each red subset shows the peak, 16 quarters before the peak, and 10 quarters after.

Graph #3: Household-to-TPNF Debt, and Three Similar Subsets
Then I pulled out the subsets to take a closer look. The blue and green lines on the next graph are the peaks in the 1960s (blue) and the 1990s (green). The two are remarkably close.

The red line is the 1979Q4 peak. It shows the same general shape as the blue and green. But the red is significantly lower, just as the October 1979 peak is lower on Graph #3.

Graph#4: The Subsets Highlighted on Graph #3
Blue=1964 peak, Red=1979 peak, Green=1996 peak
Again, this is Quarterly Data
On the X axis, Q is the peak quarter for each line. Q-16 is 16 quarters before peak. Q+10 is 10 quarters after peak.

I'll say it again: The blue and green lines, 30 years apart chronologically, are remarkably, remarkably close as a portion of total private non-financial debt. I'm not saying anything about cause or effect. But maybe there is something to be said for significance.

1 comment:

The Arthurian said...

Afterthoughts.

These graphs compare household debt to household-plus-nonfinancial-business debt. What economic situations might arise from the household share being less than half, versus more than half?

Possibly, higher consumer spending arising from higher household debt (and lower business debt) would tend to cause inflation. Oddly though, household debt goes low as inflation arises in the 1960s and '70s. So this seems unlikely.

Possibly, lower household debt (and therefore, higher business debt) could be associated with increased productivity as business investment increased. But productivity ran low from the mid-70s to the mid-90s, and that doesn't seem to fit the graph either.

On the other hand, the household share was rising through most of the 1990s, and the 1990s were pretty good. The '60s were pretty good too, and the household share was rising then, until inflation started to be a problem. So it seems that a rising household share of debt is sometimes good for the economy.

(There are limits to that, of course.)