This country makes a big deal out of the Dow Jones Industrial Average. People watch its rise and fall as though it were a very important number. I rather wonder if it actually is. I doubt that most Americans make very much money at all from investment income, though a bit of googling couldn’t find me the authoritative statistics I was looking for. I also doubt that most Americans are directly employed by DJIA companies, nor even employed by DJIA companies or their subcontractors, though again I’d need more data than I have.
What I do have access to, though, are historical DJIA numbers, the historical Consumer Price Index, historical GDP, and historical population (whose source, for the life of me, I can’t find at the moment). I did a bit of rudimentary data analysis; what I found was that
- The R2 between the DJIA’s year-over-year percent return, and year-over-year percent change in per-capita real GDP, is 0.06.
- If you just look at 0/1 variables — 1 if the year-over-year number rose, 0 if it fell — then the R2 is somewhat better: 0.31. Still nothing special.
- Doing data analysis in OpenOffice sucks. I should relearn R.
I’ll admit my suspicion going in: people often falsely assume that their fates are tied to those of economic elites, because it’s in the economic elites’ interests to make people think so.
If others would like to take a crack at this data (or add any other public sources of data that I missed), I’d love to see what you come up with. Among other things, the following would help:
- Higher-frequency data (surely available for DJIA; not as surely available for CPI, GDP, or population).
- Looking at the S&P rather than the DJIA, because the S&P is a broader pool of companies that may be more representative of the economy as a whole.
- Using better statistical tools. I admit shyly that it has been eight years since I’ve done any respectable data analysis.
- Using something other than OpenOffice, which — I repeat — sucks.
P.S. (3 June 2008): Obviously I need to be using the χ2 statistic on the 0/1 variables: a two-by-two table, where each axis is “rose” or “fell” for each of the two financial measures that I’m comparing. If the rise and fall of the DJIA is statistically independent of the GDP’s rise or fall, then the χ2 statistic won’t be too large. I’m just not inclined to try to do this with OpenOffice.