Sunday, October 23, 2016

Change in debt and change in GDP

If you take the ratio of two "change in" values, the change in this divided by the change in that, usually you get a graph with a few very large spikes, and everything else too small to see. I try to avoid graphs like that. But sometimes you have to look.

Here is the change in TCMDO debt relative to the change in GDP, from FRED:

Graph #1: Change in Debt relative to Change in GDP
Doesn't show me anything but one tall spike.

I've been thinking about using the Hodrick-Prescott filter to soften the changes in two "change in" series, so I can look at the ratio of filtered values.

Can't do Hodrick-Prescotts at FRED. I brought the numbers into Excel. It's quarterly data but for extra smoothing I used the bigger smoothing constant that's a default for monthly data. Here's what I got:

Graph #2
I still have that tall spike. But instead of being a momentary phenomenon it looks like it lasted 20 years. That's a little silly. What's interesting, though, apart from the spike, is that there seems to be a persistent upward trend from the 1950s to the present day. Apart from the spike and some possible wiggles.

I reduced the smoothing to a low number and took another look:

Graph #3
The spike is still there, now spread over only a few years. And there are lots of little ups and downs probably related to recessions.

Look how small the wiggle is, between 2006 and 2009. That's the crisis.

That tall spike of 2001 is really something of an anomaly. Looks like there was a particularly small increase in GDP and a good size increase in debt at the same time. In the source data I have the value 2.53 for the second quarter of 2001, 665.07 for the third quarter, and 9.25 for the fourth quarter. That third quarter number is a fluke.

I deleted the third-quarter value and left the cell blank in the spreadsheet. (The HP function reads it as zero.) And I changed the smoothing factor back to the number I started with. Here's how the graph came out this time:

Graph #4
Now we're getting somewhere. There is a definite up-trend with some wiggle to it for the whole period shown. There are exceptional lows centered on 1991 and 2009.

You will recognize the 2009 event as our crisis and the Great Recession, even though the smoothing factor pushes the peak back before 2003. Maybe we could say that the smoothing merged a normal wiggle (the 2001 recession?) with the large decline of the Great Recession. Just a thought. I try to be flexible, reading these things.

The other low, 1991, I think I know what that one is. There was a general slowdown in the growth of debt that lasted from the mid-1980s to the early 1990s. I looked at that slowdown before.

Note the rapid increase following the 1991 low. That increase was made possible by the big decline that preceded it. And the rapid increase in Debt-to-GDP provided the funds that allowed the growth of spending that created the good years of the 1990s.

So, why have we not had comparable good years after the 2009 low? I don't know yet, but I have some thoughts to explore.

1. The uptrend after 2009 starts at a much higher level. There's a lot more debt now than in the 1990s.
2. The uptrend after 2009 is less steep. So there is more of a delay before the good years this time, and those years might not be as good.
3. The uptrend after 2009 has not yet reached its peak. There is still time; the good years may yet come.

// The Excel file

No comments: