Thursday, May 25, 2017

Just to be sure

This graph (from yesterday's post) shows average growth rates for periods beginning in 1930. But the graph is a close-up. It only shows the years since 1952:

Graph #1
So the first point on the red line represents average growth for the 1930-1952 period. This seems misleading, as the years before 1952 are not shown. I want to re-do the calc, figuring average growth rates for the period beginning in 1952.

I'll keep the green line from the first graph, the "average annual growth" calculation from BEA. I'll toss the running average line, and make a new red line using the BEA calc on data from 1952 and after:

Graph #2

The new red line shows a lot of variation in the early years. That's to be expected, as there isn't much to average against. The green line has a backlog, 20 years of data from before 1952, acting like an anchor to prevent the green line from moving when the blue line moves.

In the early years the red line has no such backlog, so big changes in the blue line create big changes in the red. After a dozen years or so, the red has a backlog of its own. Then it is not so much influenced by changes in the blue. And then we see the red and green run side-by-side.

By the time the red and green run side-by-side, both are more influenced by their past than by each new change in the blue line. So if somebody says average RGDP growth in the 1947-2017 period is 3.21%, it tells us more about the past than it does about today.

On the other hand, if the side-by-side years show a general downward trend, it means there are so many below-average years that they are dragging the anchor down. I conclude, then, that the important information in these long-term averages is not that the average value is 3.21% or 3.3% or whatever.  The important information is that growth has been going downhill since the 1960s.

I go back to the first graph and this time keep the red line, the running average since 1930. I get rid of the green line and create a new one showing the running average since 1952. This time we compare running averages for two different start-dates.

Graph #3
This time the red line has an anchor, two decades of data from before 1952. The green line doesn't. So the green line responds more to changes in the blue than the red line does. But again, after a dozen years or so, the "since 1952" line has its own backlog, and we see red and green show the side-by-side behavior.

We get the same behavior for running averages as we got for the BEA calculation. It turns out that the "backlog" is more significant than the calculation that makes use of it. To me this says the long-run average isn't worth much, except it shows that new growth keeps dragging the average down.

To finish up, I'm replacing both lines on the first graph with the "since 1952" data.

Graph #4
Red and green follow the same path: decline since the 1960s.

// The Excel file

No comments: