In comparing a current price to something 10 years ago, it is instructive to consider the correlation of a part to a whole the correlation of ( x1+ x2 + x3 + x4 + x5 + x6 + x7 + x8 + x9 + x10) to p when the x's are uncorrelated with each other. If the correlation of x10 to p is negative then the average correlation of the other 9 x's to p must be positive, surprisingly so. And good to calculate. The correlation between the first quarters earnings and the whole years earnings is of the order of 50% assuming randomness and uncorrelation.

I had many old men very angry with me when I pointed out this fallacy in Green and Segall's and Myron Schole's work that showed that first quarter earnings were not predictive of the whole years earnings. It meant that the next 3 quarters were highly negative correlated. The same defect is relevant to the chronic bear's work, and when I took the liberty of pointing this out to him, he demurred and apologized and said something about the stochastic calculus, a trick he tried on Andy Lo with similar absurdity.

Charles Pennington writes:

I suspected that this tale had grown a bit tall with age, but here is a link to the first page of a 1966 article by Green and Segall. On that first page they do say that they can't seem to prove that the first quarter's earnings are helpful in predicting the annual earnings.






Speak your mind


Resources & Links