Confusing - no? One might be tempted to see it as chaotic, random, formless, or a lot of fury signifying nothing. Let's do what I've done with percent change in GDP data so many time and try to find whatever order may be lurking in the chaos. The visual problem here is that the size and number of the spikes disguises whatever underlying pattern might exist. One way to simplify is to take a moving average.
This chart is rescaled and has the same data in the background, but zeroes in on the 21 period moving average.
Now we have something we can make sense of. The straight blue line is the best fit. The green line is one I constructed connecting the first and last peaks. (I'm always amazed by the amount of respect lines like this receive from presumably random data.)
We see -
1) A positively sloping trend line.
2) Steadily increasing peaks.
3) An essentially level series of bottoms, at the 0.00 line.
All of this is consistent with my assertion that the most recent decade has been a remarkable profit generator. Not the best decade in the entire data set, though. Quite surprisingly, that honor falls to the 70's! The two most moribund decades since the Great Depression account for the greatest growth rates in corporate profits. Make of that what you will.
4) The peak to peak period is not constant. In fact, it seems to be increasing quite regularly. I'll have a closer look at that when I have more time. For now, it's just an oddity.
The increasing slopes over time we saw here really do mean something. There are trees in the forest and a careful observed can identify them.
.
![[Most Recent Quotes from www.kitco.com]](http://www.kitconet.com/charts/metals/gold/tny_au_xx_usoz_4.gif)
5 comments:
It seems we were both working on the same idea again! While you were preparing your post I was doing a similar one.
Real Pre-Tax Corporate Profits per Capita
I see twin bubbles.
Agreed: The apparent randomness in your first graph is not evidence of actual randomness. And your second graph shows it.
As an aside, I really like that second graph.The blue-on-gray makes the plotted line stand out, and the faint image from your first graph there in the background ties the two together and emphasizes your reason for making the second graph.
Glad you pointed out the 1970s-was-best thing. The area under that part of your 21-period line looks to be twice the size of the area under the 2000s part. (And this is how I get in trouble: by eyeballing things, as opposed to doing another graph of them as you are doing here. Hmm.)
Perhaps the 1970s performance is related to the high inflation of those years? (Mark's graph shows a much smaller 1970s, after adjustment for inflation among other things.)
You often observe a change in trend around 1980. Putting one continuous trend line on the graph tramples your observation. Showing two trend-lines here (before & after 1980) might show a saw-tooth pattern... But then, how would we justify interrupting the trend line at that particular point?
I have a new chart up that uses the fall in the dollar index to explain some of this.
Corporate Profits vs. Dollar Index
Came over from Illusion of Prosperity.
If you want to do the trendline analysis right, I think you might need to start and end at the same point in a cycle. Otherwise wouldn't you get a trend bias just from the above-trend or below-trend portion of the last segment?
WS -
Welcome!
I'm not getting what you mean. Sorry.
There are two ways I know to do trend lines, and I did them both here. One is a best fit, which hives you a slope. The other is to connect the tops and/or bottoms. This is the way trend channels are defined. Connecting tops, frex, does give you he same point in each cycle, it that's what you meant.
I didn't bother with a bottom line, since the y-axis provides that quite nicely.
Cheers!
JzB
Post a Comment