Behavioral Investment Management: An Efficient Alternative to Modern Portfolio Theory. 2012. Greg B. Davies and Arnaud de Servigny.
Reviewed by Mark S. Rzepczynski
Behavioral finance — everyone talks about it and continues to add to the compendium of flaws in our decision making, but little has been done to improve the overall state of investment management on the basis of our growing knowledge of how people actually act. In Behavioral Investment Management: An Efficient Alternative to Modern Portfolio Theory, Greg B. Davies and Arnaud de Servigny attempt to address this void in a unified investment management framework based on portfolio optimization with a behavioral component that uses advances in utility theory.
If you are looking for the secrets of better decision making, you will be disappointed. This book is not for the nonquant investors who want to improve their portfolio decision making and minimize the flaws in their logic. But if you are open to a method for incorporating a better-defined utility function into modern portfolio theory (MPT), this book is worth reading. Behavioral Investment Management will not cause a complete reexamination of MPT. Rather, this work is an extension of portfolio theory through practical tips for using utility theory and dynamic adjustments to improve out-of-sample estimation. Davies and de Servigny call this approach “behavioralizing finance.” It is all rather ambitious given the amount of work developed around behavioral finance, but there are enough nuggets of useful information to elevate the quality of many investment managers.
Although the authors overreach in stating that the current investment paradigm is broken and can be fixed with their approach, they develop an explicit framework to address behavior through a more complex utility function. In its simplest form of historical, static mean–variance optimization, MPT is a straw man that can be easily knocked down, and the authors do an excellent job of developing the theory’s limitations and offering solutions. Their presentation, which incorporates complexities beyond the quadratic utility, offers a simple way to adjust to downside risk and higher movements of the return distribution.
The quadratic function — the simplest form of utility function used in finance—does not properly describe investor behavior. It has been a methodological expedient that does not fully reflect risk, loss aversion, and the impact of wealth. Most importantly, a quadratic approach does not allow for the full impact of return distribution beyond mean and variance. Formally, Davies and de Servigny try to incorporate the descriptive theory of choice embodied in cumulative prospect theory by using an exponential utility function, which can incorporate different levels of risk tolerance in a manner that better reflects current utility theory.
The authors contrast the conventional approach of MPT with the advancements of the Black–Litterman Bayesian approach and show the former’slimitations. Through testing various in-sample and out-of-sample approaches to modern portfolio optimization, they outline the empirical problems of historical, static mean–variance optimization. The problem of out-of-sample results comes from the dynamic breaks and conditional return behavior of asset prices. The impact of return changes associated with both estimated means and changing covariances distorts the weights that are appropriate for hitting a target volatility and return. Moreover, non-Gaussian distributions lead to more extreme values that cause further distortions in optimized weights. Although real, this estimation uncertainty is not a true behavioral problem and is well known to most users of optimization techniques.
Davies and de Servigny behavioralize portfolio theory by using an exponential function that can account for higher movements of the return distribution. By incorporating skew and kurtosis into a utility risk measure and discounting excess return by what they call the behavioral variance, the authors account for the unique risks of an asset class, risks that cannot be incorporated into a simple mean–variance framework. Investors prefer more of the odd moments of the distribution—mean (first) and real (third)—and less of the even moments: variance (second) and kurtosis (fourth). An asset with similar variance but more kurtosis will be less desirable. Behavioralizing will incorporate this difference. Adding these complexities to excess return creates a desirability index for a given asset. Lower risk tolerance will be affected by the entire return distribution. The risk–return trade-off for a fixed tolerance can now be easily measured.
Although return distributions may become more normal over longer periods, non-normal distribution effects drive short-term behavior. These behavioral measures capture shorter-run asset behavior. The impact of loss aversion and reference dependency is also accounted for through shape adjustments of the utility function. These functions can be adjusted to the level of risk tolerance. By accounting for downside risk or regret as well as reference returns, the authors’ model allows for framing and loss aversion, two of the most important developments in behavioral finance.
Davies and de Servigny address the issue of shifting asset returns from regime changes or changes in the business cycle by incorporating the exponential moving average for distribution means. Although a dynamic approach to return and covariance estimation can be more useful than a static historical approach, it is a long way from eliminating decision biases and behavioralizing the portfolio management of return expectations. The authors incorporate a momentum heuristic bias, which is helpful in the context of a static mean but does not fully incorporate the wealth of advancements in portfolio theory toward conditional modeling.
The authors link the effect of estimation uncertainty with differences in the utility function to provide an inclusive framework that results in their behavioralizing of investment management. Their work shows the impact of including these two themes to present a compelling case for an enhanced approach that leans on our knowledge of behavioral finance. Of course, the issue of determining the risk tolerance of an investor and parameterizing it is still suspect. A cardinal ordering can be made, but the exact level of tolerance must still be measured.
Davies and de Servigny do an effective job of bringing advances in MPT into a framework that adjusts to some key behavioral problems with respect to the traditional risk–return trade-off. Although there are better books on how best to use optimization techniques, Behavioral Investment Management represents a clear advance by embedding investor behavior via explicit utility functions. The book neither addresses nor solves all the behavioral mistakes that can be made; in fact, it avoids issues of decision biases and errors altogether. But it contains enough useful material for even veteran portfolio managers and quants to adjust their thought processes.