Do I Have to Become a Chartist?

The Dow-Jones Industrial Average has spent 25 of the past 40 years--62%--between 800 and 1250 or between 8000 and 12500. These ranges comprise roughly 25% of the (logarithmic) total range of the Dow as a function of time.

It really looks like there may something special about the nominal value of a 1 followed by a bunch of zeros.

This is disturbing to me as an economist.

Comments

[TB]

Benford's law?

[It's more than Benford's Law. Benford's law is that numbers ought to be evenly distributed on a large scale--which means that 1/3 of the numbers should begin with "." This is much, much stronger than Benford's law. And it's not numbers that begin with "1", it's that there are a huge number of people who start selling when the market gets much above and selling when the market gets much below Dow 1000 and Dow 10000.

And there is no such pattern for the S&P, the NASDAQ, or the Wilshire...]

[Frank Dean]

Brad DeLong's next book:
Dow 100,000: The New Strategy for Profiting From Humanity's Love of Round Numbers

James K. Glassman and Kevin A. Hassett are even stupider than we thought!

[Daily Shocker]

Yay, numerology!

[dis]

benford's law indeed!

this would still be true regardless of the units used.

do the dow in british pounds and see the same result.

read on benford's law brad. it's neat and does make sense

[Jrossi]

I invest by drinking lots of Old Crow and then choosing something at random. And, hey, this year I'm beating the S &P 500!

[Eccles]

Talking of long term analysis of the Dow...

Is it just me or if you take out the dot com boom and bust of the 90s and this current crash you can smooth out (highly scientific I know) the Dow to a point roughly where we are now. Were we trying to create wealth at a rate beyond reality and we have just had a readjustment?

[Bram Cohen]

Dang, everybody beat me to mentioning Benford's law.

There's another obvious trend which people don't like to talk about - the market moves on a 40 year cycle. The periods 1900-1920, 40-60, and 80-2000 were up, and the periods 20-40 and 60-80 were awful. Of course, this is pure chartgazing, but it indicates that claims of the market providing consistent good returns are based on very little historical data.

[Ken Houghton]

Give me a chart that is obscured enough, and I will turn an economics professor into a Chartist. -- ArchiDataMinedes

[Quicksand]

> There's another obvious trend which people don't like to
> talk about - the market moves on a 40 year cycle. The
> periods 1900-1920, 40-60, and 80-2000 were up, and the
> periods 20-40 and 60-80 were awful.

On the bright side, we may already be 40% or so through the bad part!

This numerological phenomenon (1,000/10,000) is interesting to speculate on. Do rational (or irrational, for that matter) investors really make individual transactional decisions based on the psychological value of a rather arbitrary aggregate index? That would be weird.

[Andy B]

Without applying any knowledge of economics, I only see two data points here, which is (colloquially) only a coincidence, not a conspiracy. Why are you tempted to go against that general principle and reject the null hypothesis in this case?

[SvN]

I recall a serious empirical finance paper published on the round-number effect in FX markets in the 90s. They found that round numbers tended to act as repelling barriers; levels just about a round number produced abnormally high numbers of positive returns, levels just below tended to produce negative returns. These effects vanished when the reverse quotes were analyzed.

[The Fool]

I'm no statistical whiz but it seems to me you can divide that chart into 3 roughly equal time periods: a plateau, a rise, and a plateau. 3 periods would be a pretty small sample size wouldn't it? Couldn't it just be sheer coincidence that the second plateau occurred at a number starting with a 1 followed by 0's?

[The Fool]

Also, isn't that effect entirely dependent on the units chosen, i.e. the DJIA. If you chose some other index with different numbers, would the effect disappear?

[Robert Waldmann]

No you don't have to become a chartist. You could become a Decembrist or a Buchmanite or a single taxer if you prefer.

Look the chartists wanted annual elections http://en.wikipedia.org/wiki/Chartist. That is unspeakable. Don't become a fundamentalist, but have you considered Bahai ?

[Kerfluffle]

Frank Dean, you are such an optimist. In these times we should worry about a drop getting us stuck on 1000 for a decade or so.

BTW I think we need more empirical data before we can say anything for sure. How long did the Dow hover in the 100 range?

[Alex Tolley]

Try correcting the chart for inflation and use real vs nominal numbers.

[Gary]

log(125/80)=.1938... I wouldn't call that "roughly 25%."

[mn]

The DJIA index ignores dividends. Dividends are a huge part of investors returns.