Dec

11

About 15-20 years ago a number of academic articles were published (by Fama-French and Shiller among others) claiming that aggregate stock market returns are partly predictable using variables such as P/B, dividend rates, the term structure of interest rates and so on. This is the ‘market predictability’ literature. More recently a number of articles have appeared that claim this ‘predictability’ is weak and not useful from a practical point of view. This article by Giot & PetitJean, entitled International Stock Return Predictability: Statistical Evidence and Economic Significance, belongs to the strand of the literature that grants that predictability may have existed in the past but doubts that it will continue and/or finds that it is of no practical value. I’ll call it the ‘predictability is pretty useless’ school. There is also another line of criticism to which you can subscribe, that is more radical and I will call the ’statistical malpractice’ school, that believes the apparent predictability does not exist even in the past data and is an artifact of flawed statistical procedures such as overlapping multi-year stock returns and/or too few degrees of freedom for proper inference when the procedures for overlap-adjustment are known to give bad result in small samples, etc., etc..

The two little known Belgian authors examine five variables: dividend yield, earnings/price ratio, short term interest rate, long term interest rate, and interest rate spread. They look at the stock markets of 10 major countries, and they use an ‘Out of Sample’ methodology in which the relationship is estimated over a period and then applied to the following period, i.e. a predictive study. They use a statistical test I am not familiar with, due to McCracken (2004), which is apparently a variant of the Diebold-Mariano test that I discussed with Professor Diebold the other day.

The conclusion in their words:

The short-term interest yield and, to a lesser extent, the long government bond yield are the best out-of-sample predictors of stock returns. However, the out-of-sample predictive power of these variables does not appear to be economically meaningful across countries and investment horizons.

This is all I could glean from a quick and partial reading of the article.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search