Softpanorama (slightly skeptical) Open Source Software Educational Society

May the source be with you, but remember the KISS principle ;-)

Beware that annual rebalancing might hurt returns

Review of Intelligent Asset Allocator by William Bernstein

The Intelligent Asset Allocator: How to Build Your Portfolio to Maximize Returns and Minimize Risk

• Hardcover: 206 pages
• Publisher: McGraw-Hill; 1 edition (September 22, 2000)
• Language: English
• ISBN: 0071362363

March 11, 2006

I am a programmer and usually do not write financial books reviews. Also it is easy to criticize a book that was published 6 years ago (it was published September 22, 2000 when smell of the forthcoming dot-com crash was already in the air). At the time of publication it was a very good book that definitely saved a lot of pensions from complete decimation due to heavy doze of technology-related stocks (or funds), an allocation typical for the level of greed (or appetite for risk taking) that was typical during dot-com bubble.

But now we need to take recommendations more skeptically and at least check how they faired during six years since the book publication before jumping into action.  And if we are talking about investment book it is prudent to judge its value on the base of performance of its recommendations in five years or even a decade after the publication.

Thinking about optimizing my 401K portfolio before possible this (or next) year stock slide due to an inverted yield curve I dusted off the book and simulated the performance for one of the recommended portfolios for the period from Jan, 2000 to Mar, 2006. The initial investment was assumed to be zero and monthly contributions assumed to be $1K. Rebalancing was done at the end of each year.

Please be aware that recommended on page 154 portfolio using the period and contribution mentioned above with the end of the year rebalancing produces almost zero return on the capital from Jan 2000 to March 2006 (7% total or ~1% annualized excluding taxes and fees). Taxes and mutual funds fees will drive return lower. For example Vanguard additionally charges annually $10 for any fund in which you have less then $10K. That means that any Vanguard money market fond provides better return with much less risk for this period. BTW without rebalancing the same portfolio produces 31% total or ~5% annualized beating S&P500 (17% total return with the monthly investment of 1K mentioned above).  It looks like this "rebalancing-induced returns drop" is the most pronounced when you start with zero initial capital. If the initial capital distributed into the "slots" the first month  is comparable to the subsequent investment (say $50K or $100K, while subsequent monthly investment for the period is 76K) "rebalancing-induced returns drop" is less pronounced and is approximately 50% of total return. BTW the book recommended the initial capital for this portfolio in $100K-200K range which helps to avoid some mutual funds fees inherent in splitting small initial capital into 9 slots, so the case above should be viewed as an extreme version of the "rebalancing-induced returns drop".

Unless my calculations are wrong ( I strongly recommend not to take my results for granted; anybody can redo the calculations using Excel; it's not that difficult) it looks like the idea of periodic end of the year "blind" rebalancing that the author preaches is open to review. Such rebalancing is just a blind trade of your assets determined by a very primitive algorithm that was never validated on large samples of data and different historical periods.

The idea of "split" stock and bond allocation (with possible additional diversification into less correlated asset subclasses) is an old and sound idea, but diversification should probably have some limits. When you start doing hair splitting into too many subclasses of each major investment class the portfolio performance probably became more dependent on the performance of the weakest component as year after year rebalancing will move money to the worst performer(s). If it is always the same worst performer you probably will eventually approximate the return from it. And this probability rises with the increase of the number of components.

Excessive complexity of portfolio might not produce better returns. For example 9-component portfolio that I tested above for 2000-2006 returns is a way too complex and nothing suggests that it can produce statistically better results than classic "age based" binary stock-bond allocation using, for example, Total Stock Market Index and Total Bond Market Index with, say, (100 - your age)% in stocks and rest in bonds, or, depending on the tax situation, with some small but rising with age percentage parked in money market (actually in March, 2006 one year CDs return more then 5% and this year might outperform most bonds). Those multi-component portfolios, which are supposedly cooked by experienced financial cooks and thus can be called "borsch allocations" might just increase taxes and mutual funds fees without improving returns. It makes sense to keep allocation simple.

Like alchemists in the past tried to convert mercury into gold, Bernstein tries to find magic mix of several mutual funds representing different classes of assets adding to the idea of (static) diversification the idea of weak correlation. Actually correlation is a very tricky thing for time series to rely on it in your money allocation decisions; basing your decisions on calculation of correlation for previous periods expose you to a similar risk as buying mutual funds based on past performance.

In reality it might well be impossible to get a good asset allocation with low risk on purely technical basis without right insight into the direction of world economy and major currencies. Stocks are inherently risky. Each allocation whether based on some magic calculations or not is essentially a prediction of future market direction which might be true or not. For example any 60:40 split of stocks and bonds is essentially a 60% prediction of solid stable growth of the economy. It does not work well for prolonged recessions like Japanese. Investment in gold is essentially a prediction of the forthcoming devaluation of dollar against basket of major currencies. And so on. From this point of view "frozen" in the book "year 2000 vision of the future" is nothing to boast about as in proposed allocations the author almost completely missed the possibility of dollar slide (currency risk), as well as dramatic raise of commodities, especially precious metals and energy.

It would be also interesting to make simulation runs based on the data from 1995 version of his book which permits checking a decade of performance but I do not have it.

Last modified: February 27, 2008