Wednesday, February 8, 2017

Getting a look at the result -- that's the prize


Sometimes one thought crashes into another and makes a little spark in my brain. Those are always intriguing. They always make me want to write. You never know at the start how the writing will turn out. Maybe there's nothing there. Maybe it was a hiccup, not a spark.

For the millionth time I was proof-reading my Fact-checking thingie the other day, and looking at that graph again:

The Long-Term Decline of Real GDP Growth
I suddenly thought I can subtract the trend-line from the data-line and see how it looks then. And that was all it took to get me going.

The graph is old. It only goes to the start of 2013. So I got new data from FRED and sat down to make a new graph, similar to the old one. That's when I noticed all the revisions. Revisions make me uncomfortable. I'm all for getting the numbers right, but they're still changing numbers from 69 years ago. There's a point where it just becomes ridiculous. There's a point where you start to stop trusting the revisers.

In general, we might expect estimates of nominal GDP to change for a few years after the initial estimate, as new data become available, and to then stay fixed.

Oh, well. If I didn't do graphs, what would I do?


I made a new graph, similar to the old one but using the current data:

Graph #2: RGDP Growth and Trend

Then I modified it: Showed more decimal places in the Trendline Label, because I'll need the numbers; Removed the dates from the X axis and went with numbers in sequence, because I need that for the calculation too; Added a column to the spreadsheet where I use the Trendline Formula and the X Axis values to calculate values to fit the Linear Trendline; Showed those numbers as a Green Line on the graph, to make sure my calculated trend line matches Excel's trend line; And showed you the result:

Graph #3
Now I have numbers that represent the trend line. Now I can subtract those numbers from the original data and look at the result.


If I take the linear trend line, subtract it from itself, and graph the result, I get a perfectly flat line at the zero level. Obvious.

If I take the red line and subtract the linear line from it, I get a version of the red line that is centered on the zero line and runs flat, not downhill. That's what I want to see.

Graph #4
That's what I wanted to see: The high in the 1990s is lower than the high of the 1960s. And the low after 2008 is lower than the low of the 1950s. Things are lower now than they were before. That's what I was wondering about.

Now the only thing I don't know is if "things are lower now" is true, or if it's just that the linear trend was a bad choice.

Not a spark. Just a hiccup.

No comments: