Customer Reviews

A Non-Random Walk Down Wall Street

5 star:		(6)
4 star:		(2)
3 star:		(3)
2 star:		(4)
1 star:		(1)

 

 

The other reviews are right...this book is definitely not a how-to guide for personal investors looking to "beat the market." It's essentially an academic tome, so its theme is tightly circumscribed (so they do not and should ask about all asset markets that might possibly be relevant to investors -- only the stock market over certain periods). The exposition is extremely sophisticated and makes use of cutting edge mathematical and especially statistical modeling to make the case.

The punch line has two important parts: (i) the "random walk" hypothesis is false -- day to day movements in stock prices are not random bouncing that many extant models claim they should be; and (ii) most of us will never have the capabilities to employ these modeling techniques to put the rubber to the road and find out WHICH way stock X is going on December 13.

So it's fascinating in regard to the mechanics of asset pricing, but totally useless as a practical investment guide. But that doesn't mean it's a *bad* book or that it warrants a 3-star rating (the average at the time of this review). Blame _Business Week_ if you expected something else. The book is exceptional and does no more and no less than what it claims to do.

The random walk hypothesis, considered the bedrock of financial theory and modeling, is challenged in this collection of eleven papers by the authors. They attempt in these papers to show that the financial markets do contain a certain degree of predictability, and they illustrate this by both analyzing empirical data and with the development of various mathematical formalisms. It is always interesting when a given paradigm which is entrenched in the minds of a field's practicioners, is challenged and shown to be either inconsistent or not supporting the real facts. The authors make a strong case in this book against the inherent randomness of the financial markets, and they do so in a way that is very understandable. Also, after a consideration of their results, one can construct practical trading software packages that are based on financial models not using the random walk hypothesis. Thus their study is very useful from a practical, everyday trading point of view.

After a brief overview of the efficient markets hypothesis, in the next chapter the authors go right into the analysis of the efficient markets hypothesis by using a specification test based on variance estimators. They conclude from their results that the random walk model is not consistent with the behavior of weekly returns. Interestingly, they find large (negative) autocorrelations in security prices. They do not conclude though that all financial models based on the random walk hypothesis are invalid, but rather they use the specification test to study various stochastic price processes. Since volatilities do change over time, the authors are careful not to reject the random walk hypothesis because of heteroskedasticity; the test they do employ takes into account changing variances. They also discuss the possible role that non-trading practices may have on their conclusions. For the purely mathematical reader, they include in an Appendix to the chapter proofs of the theorems they used in the chapter.

In Chapter 3, the authors employ Monte Carlo simulations to study the variance ratio, Dickey-Fuller, and Box-Pierce tests under Gaussian null and heteroskedastic null hypotheses. They also consider the power of the variance ratio test against an AR(1) process, AR(1) + random walk, and an integrated AR(1) process models of asset price behavior. The discussion is very thorough, and they conclude that the variance ratio test is a viable tool to use for inference in financial modeling. Since they do inform the reader the particular packages they use to perform the Monte Carlo simulations, their results, which they report in tables in the chapter, can be straightforwardly checked.

A somewhat esoteric but very readable account of what has been called nonsynchronous trading is given in the next chapter. They begin the discussion by employing an interesting and elementary argument that explains very well the consequences of ignoring nonsynchronicity in the sampling of multiple time series. The authors list ten consequences of the presence of nonsynchronous trading and then study the empirical evidence for nontrading effects. Also, they give a brief summary of the implications of employing Markov chains to build dependence into the nontrading process, motivating readers to perform the necessary calculations on their own.

The next chapter focuses on contrarian investment strategies; namely one that takes advantage of negative serial dependence in asset returns. The authors summarize the data on autocorrelation properties and also present a formal model of a particular contrarian strategy. They conclude, interestingly, that a large portion of contrarian profits cannot be attributed to overreaction.

The most interesting chapter in the book is the next one on long-range dependence in stock market prices, for it is here that many alternative statistical techniques have been devised to study this dependence. The R/S statistic is modified and then used by the authors to test for long-range dependence in daily and monthly stock return indices. Surprisingly, they find that after correcting for short-range dependencies, there is no evidence of long-range dependence in this data.

The authors switch gears somewhat in Chapter 7, where they discuss deviations from the capital asset pricing model. They discuss effectively the two models which attempt to explain these differences, based on risk-based and nonrisk-based alternatives. These two models are proposed as alternatives to the multifactor asset pricing models that have been employed to explain deviations from CAPM.

In chapter 8, data-snooping biases are investigated using the theory of induced order statistics and tested with Monte Carlo simulations. The authors effectively convince the reader of the impact of data-snooping biases in asset pricing models, and how these biases arise from tendencies to focus on anomalous data.

Even more practical considerations are considered in Chapter 9, where the authors show how to maximize predictability in asset returns. They use a model of time-varying premiums to estimate what they call the maximally predictable portfolio, with this model using an out-of-sample rolling estimation technique to avoid data snooping problems. Monte Carlo simulations are again used to validate the results of the models. They emphasize in their conclusions that predictability does not imply market inefficiency.

Emphasizing the discreteness of real price data, the irregular timing of transaction prices, and the conditional nature of price changes, the authors develop in Chapter 10 a model that addresses these issues using what they call an ordered probit model. They conclude, using some interesting technical analysis with their model and its comparison with empirical data, that discreteness is important in financial modeling.

Chapter 11 is very empirical, wherein the authors study transaction data on the S&P 500 futures contracts with the goal of studying price behavior in relation to arbitrageur strategies. They conclude that on the average, mispricing increases with time to maturity and is path-dependent.

The last chapter of the book discusses the October 1987 stock market crash, with the goal of analyzing order imbalances and stock returns. They conclude that there are notable differences in the returns realized by stocks in the S&P index and those that are not, interestingly.

This is a book about financial economics, not day-trading. The techniques used is advanced econometric analysis, not technical charting. The purpose is to clarify some common myth about efficient market theory and the random walk hypothesis, not to show one how to pick stocks. Just like the authors' other book ("Econometrics of Financial Markets"), this one is of the highest high quality, and does a superb job on what it set out to be.

Some readers seem to be disappointed at this book by naively assuming what the title implies, as shown by some of the reviews here. They really can't blame anyone but themselves. Just because Burton Malkiel's classic didn't show us how to day trade doesn't mean a book with the opposite title will do so, nor did the authors ever claim that, either.

Andy Lo is one of the best minds in quant investing out there right now, and I should have expected this to be a challenging read. But it was a little heavy on the calculus even so. Thankfully it's well written enough that you can take the math for granted and still get a lot out.

It is always nice to see confirmation of one's basic viewpoints in print. Those seeking a simple trading program will be dissappointed in the technical nature of the arguement. However, those not put off by a mathematical thesis will find some real jems in the text.

This is a very interesting book and would have been useful in the empirical finance Ph.D. class I took at Berkeley several years ago. Note that this is a difficult read and is orientated towards Ph.D's and their doctoral students.

There is no indication anywhere that this book was intended
either as a follow on to Burton Malkiel's A Random Walk Down
Wall Street or as a primer for day traders. Hence it is
rather disappointing to read the reviews of those who
somehow managed to reach one of the preceding conclusions.

Several statistical studies have made it clear that the
markets are not completely random as asserted by much of the
academic economics community. It is impossible to prove or
refute the Efficient Markets Hypothesis, because, as Farmer
puts it, the EMH, by itself, is not a well-posed and
empirically refutable hypothesis. This book tries to
rigorously analyze the markets as they are.

The average investor could easily reach the same conclusions
as Burton Malkiel strives for, namely that he is best off
investing in an index fund. However, they do it in a more
interesting fashion than simply asserting that, on average,
one cannot beat the average.

This book was featured in Business Week, where its very interesting premise, namely that there is just enough non-randomness on Wall Street to make money, was laid out. What Business Week fails to do is to inform the reader that, except for an interesting introduction, the essays are pure mathematics with no explanatory text for any but the cognoscenti. Lay readers beware!

It is always a good sign when the results of academic research begin to confrim what the actual practitioners already know to be true...as does this book. Although this book is clearly aimed at PHD types, rather than the investment professional, the basic theme of it is of value-and the admission of non-randomness is revolutionary.

First of all it is not an easy book; it takes the reader towards a stats demonstration that markets are NOT randomly distributed. In my opinion it should be read 10 pages per day, like a good Cognac..... Enjoy it.