Based on manucastle's suggestions, I borrowed (from Las Vegas public library!) a copy of Matson and Hardy's Data Driven Investing. Here is a short review. Become a Complete Fool
Data Driven Investing is sort of two books in one, with quite different flavors. One part consists of statistical analysis of some historical data of the S&P 500 index and some other economic factors such as the US 4-year election cycle and the Federal Reserve's monetary policy in an attempt to find correlations that can be exploited in the future. The second part is more descriptive with very detailed (excessively, in my opinion) descriptions of psychological biases (this seems to be a "hot topic" these days, with many sources of good information), and a description of trading strategies employed by the authors.
While the statistical part of the book raises some interesting possibilities for future profit, I think that the conclusions are highly suspect for the most part due to poor statistical significance of the conclusions. The authors looked at annual (calendar year) returns of the S&P 500 index for 52 years from 1952 and 2003 and compared them to the 4-year election cycle and certain periods of Federal Reserve monetary policy. The problem here is that statistical uncertainty with so few data points is going to be high. Dividing the 52 annual returns into 4-year cycles leads to only 13 groups of data, for example. The statistical significance of any conclusion from this is going to be weak. However, without having calculated any probabilities and significance values, I would guess that there is something of value here, in that S&P 500 returns during the last two years of an election cycle (2007 and 2008 for the current cycle) really do show consistent and substantially higher returns than the first two years. At least there are 26 data points in one set, and 26 in the other. The difference in returns is big and consistent enough that my qualitative feeling is that this is a possibly statistically significant result of this work.
However, the statistical rigor falls rapidly from here. When combined with Federal Reserve monetary policy (which is a fuzzy definition to begin with), the number of data points for each case become so small that I can guarantee there is no statistical significance to any of the conclusions. Many of the categories, such as early years of election cycle and Aggressively Tight money policy, only have a small number of data points, typically half a dozen to a dozen.
The second part of the book begins with an extremely detailed listing of psychological biases affecting investing decisions. This area is receiving a lot of attention these days, and there are many sources of good information, many of them better than Data Driven Investing. This book's problem here is that the psychological biases are not tied back to specific investing actions in an explanatory or testable way. It's mostly just a long list of detailed descriptions. In contrast, the other recent book I read, Evidence-Based Technical Analysis by David Aronson, does a very good job of describing the relevant psychological biases in an investing context.
After the discussion of biases, the authors present a set of techniques they use for trading. Along the theme of the book being really two different parts glued together, this section is disappointing because there is no data provided to help the reader understand which methods actually work, or why they were selected. There are rules such as "Expect large moves on light volume to reverse themselves" and "Be less aggressive buying when you already hold a sizable position" and "Sell Friday afternoon and buy Monday morning." These would be great rules if they were trustworthy, but the reader has no basis on which to make that decision.
All in all, it is good to see a little statistics applied to investment. We have a great deal of data available, but no more statistical expertise to use it effectively than we did before. The first part of Data Driven Investing suffers from significant lapses of statistical rigor and, I believe, offers really only one possible valid conclusion: that the last two years of the US election cycle provides better returns than the first two years. I would hold all the other conclusions highly suspect.
In contrast, the next book I read after Data Driven Investing, Aronson's Evidence-Based Technical Analysis, is a wonderfully well developed tour through a statistically rigorous method to identify whether TA rules have statistical significance. To me, the TA part of the book is not that important (for TA fans it should be, however), but learning the methods used to establish statistical significance is very much so. In comparison, Data Driven Investing falls flat.
Join the best community on the web! Becoming a full member of the Fool Community is easy, takes just a minute, and is very inexpensive.
Based on manucastle's suggestions, I borrowed (from Las Vegas public library!) a copy of Matson and Hardy's Data Driven Investing. Here is a short review.
Become a Complete Fool