Given the time series for a particular stock market, what are the statistical weapons one can bring on to prove, or disprove that random walk hypothesis?
$\begingroup$ As a note, most stock markets tend to increase over time as they compete w/ interest. Or are you talking about a "biased" random walk? $\endgroup$– user59Feb 10, 2011 at 16:00
$\begingroup$ @barrycarter, I'm talking about random walk with a constant drift term, like Wiener process $\endgroup$– GravitonFeb 10, 2011 at 16:03
Test your historical time series for both randomness and independence. Understand that a time series may be random and independent; non-random and independent; random with dependencies; and non-random with dependencies. A mistake would be to limit dependency tests to autocorrelation. The most general test I know of is called the differential spectrum by Sherry, which works like this:
- histogram the price changes in your time series
- if the price changes are independent, they should be symmetric about 0
- use Pearson's $\chi^2$ test with one
sign as "observed" and the other as
"expected" for a quantitative measure of symmetry.
However, when you find something with this test, you've still got to hunt for the dependency. But at least the test can tell you if you've got a dependencies or not.
The most important point is that whatever tests you work with, that market can change in the future. So I wouldn't say a market "is" or "isn't" a random walk. Rather, it may phase in and out of random-walkiness for indeterminate amounts of time.
1$\begingroup$ Hi I don't know what it's worth but as it seems quite easy to implement you might try this one : arxiv.org/abs/1007.4259 $\endgroup$ Feb 10, 2011 at 13:41
The first chapter of the book Econometrics of Financial Markets by Campbell, Lo and MacKinlay discusses this very well.
2$\begingroup$ I think you mean the 2nd chapter. But yes, describes tests for it as well. $\endgroup$ Feb 10, 2011 at 20:47