# market change, correlation and estimation bias

I hear many quants sating that markets change very slowly. This "fact" is even presented as a justification of statistical arbitrage, for example, by affirming that correlations remain roughly the same for long periods, and then insight given by these correlations or by a PCA applied on the correlation matrix is valid through time.

My question is : indeed, correlation matrix does not change on a daily basis, but isn't that due to an estimation bias? When the estimation is based on last 250 days for example, any new day contribution is very small and does not dramatically change the estimator, and real correlation may be much more stochastic than the stable matrix estimated.

If this is the case, how come this artificially stable correlation matrix can give profitable trading strategies relying on it?

• This has nothing to do with banks. These are correlations implied by the options market, in which all people can trade. There is no circularity whatsoever because this calculations are $model-free$ ! They do not depend on any model!! – phdstudent Jul 27 '15 at 8:04