Comparing Two Charts — an Insight about Labor Force Participation
Thanks to FinancialRx for reminding us to check out Mike Panzer’s site for some interesting charts. Among them was a chart showing an apparent relationship between a ratio of temp workers to all workers and the S&P 500. While Mike does not make any extravagant claims about this, the text on the chart and the arrows sure make it seem like a potential decline in this ratio would be a bad sign for the market. Take a look at the chart here or below (click to enlarge).
This is a powerful type of graph to do. If there is a similarity in the pattern, one adjusts the scales to emphasize the fit. The text boxes lead the viewer to the key question. The skill in charting is worthy of Tufte (see recommended reading at right).
Since I taught the courses in this sort of analysis, I immediately had a couple of questions:
- What did the data look like BEFORE the chart’s time period?
- What sort of causal model might be at work?
- Does one market cycle offer predictive power?
Taking the second of these questions (and empahsizing again that Mike was not beating the drum about this) I am suspicious of research that begins with data instead of with theory. I have written about this approach and so has Brett Steenbarger. I wonder why this ratio is such a good predictor. Those who take courses in causal modeling approach these problems quite differently.
It is important to realize that the 1999-2000 employment era was extremely unusual. David Malpass has written effectively on this point, as has Gene Epstein, the Barron’s economic columnist. Labor force participation during this time was at an extreme, a subject which Gene covers in an entire chapter of his book, Econospinning, also in our recommended reading.
To summarize briefly this era, remember that there was a great fear about the consequences of the Y2K problem. Companies could try to update their existing computers and software. Many chose to advance purchases of computers, operating systems and software. COBOL programmers were solicited. It was all part of the fuel for a massive expansion of employment, drawing in marginal labor force members. It is possible that this peak of labor force participation will never again be reached, so comparisons should be to longer-term trends, something we have covered in past insights.
I mentioned my curiosity about the data series in our office, and our newest staff member picked up the ball and ran with it (despite the distractions of her wedding, tomorrow — Best wishes Renae!) She went to the BLS website and got the entire data series, beginning with the first tracking of the temporary workers. Here is what she found:
You can see that the series did not fit so well, using Mike’s scale, in the earlier time period.
Did something change in this relationship? There is not enough history to form a good conclusion. That is often the answer when we are looking at apparent relationships. We do not have enough data for many things we would like to explore.