Thursday, January 06, 2005

Past returns and future performance

The true-life history of stock market returns over two different cycles.

I. The presidential election cycle.

1940 to date

Average annual change in the Dow Jones Industrial Average during the ...

First year of presidential term: +3.9% (8 loss years of 16)
Second year of presidential term: +4.9% (7 loss years)
Third, pre-election, year of term: +17.2% (0 loss years)
Fourth year of term, election year: +6.1% (5 loss years)

Conventional wisdom: Presidential Administrations stoke the economy before the election to enhance their re-election prospects, then after the election they impose restraint to wring the excesses out. If one looks at the current Administration's budget for discretionary spending for the years before and after the election, it sure seems to fit the pattern.

Ergo, this is a poor year to invest in the market. And a good number of investors take this quite seriously.

II. The years ending in 0-to-9 cycle.

1884 to 2003

Average annual change in the Dow Jones Industrial Average during years ending in...

0: -7.2%
1: -0.2%
2: +2.2%
3: +6.2%
4 : +7.4%
5 : +30.7%
6 : +6.0%
7 : -3.2%
8 : +18.5%
9 : +9.2%

[all data via No-Load Fund Investor]

Hey, this is the best year to invest in the market!

Except we disregard cycle II because there's no logic behind it as there is for the first cycle. So so we conclude it's just random chance, an example of "data mining". Some one of the ten years must do better than the rest, so if you look for one you'll find one. You'll find that with any cycle, of course, or any mix of variables one examines to find a correlation with market performance. It's just chance.

But if we add a plausible explanation, does that make it anything more than chance? In the first cycle there's a good one-in-four chance that the post-election year will have the worst return -- so does the explanation of it there actually explain anything, or is it merely a "just so" story that we made up to see logical order where there isn't any?

After all, if Administrations stoke the market to optimize their chance of getting re-elected, why isn't it the election year that sees the biggest gain, rather than the pre-election year?

And if that easily explained, visible, empirical relationship is just chance, how many other fundamental investment relationships are just chance correlations that might dissipate tomorrow? Why, I'm told that for well over 100 years it was considered an investment fundamental that because stocks are more risky than bonds they had to pay a cash dividend rate that exceeded the interest rate on bonds, and they always did -- and then one day they didn't, and they never did again.

On the other hand, that 0-to-9 cycle covers a good 120 years ... maybe there is something determining it that we just don't understand. Maybe that spooky action at a distance nobody fully grasps in quantum mechanics has unappreciated deterministic effects on investments -- and we're all passing on a sure thing just because we don't have a made-up story to explain it to ourselves...

Aw, geeze ... it's a lot easier just to buy the whole dang Wilshire 5000 and forget about it.