Jason Zweig’s usually wonderful Intelligent Investor column in the Wall Street Journal had a real winner this week. He wrote about a finance professor that did some studies on diversification. The idea that diversification is essentially a free lunch is ingrained in modern finance. Investors are taught that proper portfolio construction involves owning enough companies, generally 20-30, to reduce stock-specific risk.
Mr. Zweig describes the base case for diversification when looking at aggregated data:
Don Chance, a finance professor in the business school at Louisiana State University, asked 202 students to select one stock they wanted to own, then to add a second, a third and so on until they each held a portfolio of 30 stocks.
Prof. Chance wanted to prove to his students that diversification works. On average, for the group as a whole, diversifying from one stock to 20 cut the riskiness of portfolios by roughly 40%, just as the research predicted. “It was like a magic trick,” Prof. Chance says. “The classes produced the exact same graph that’s in their textbook.”
Aggregated data, though, is tricky stuff. It’s one thing to see a pattern across a large number of cases, but what happens when you drill down into the individual portfolios that were constructed? This type of micro-analysis is often not done in finance, which can create a less nuanced view of reality. Indeed, when Dr. Chance looked at the data closely, there were some surprises.
But then Prof. Chance went back and analyzed the results student by student, and found that diversification failed remarkably often. As they broadened their holdings from a single stock to a basket of 30, many of the students raised their risk instead of lowering it. One in nine times, they ended up with 30-stock portfolios that were riskier than the single company they had started with. For 23%, the final 30-stock basket fluctuated more than it had with only five stocks.
The lesson: For any given investor, the averages mightn’t apply. “We send this message out that you don’t need that many stocks to diversify,” Prof. Chance says, “but that’s just not true.” What accounts for these results? Leave it to a professor called Chance to show that even a random process produces seemingly unlikely outliers. Thirteen percent of the time, a 20-stock portfolio generated by computer will be riskier than a one-stock portfolio.
The averages might not apply because of interaction effects within the portfolio, and because outliers are perhaps more common that we would like to believe. Even a portfolio built by throwing darts might not have the diversification that an investor is seeking.
Diversification is just one form of risk management, and it is clearly not a complete solution. Our Systematic Relative Strength accounts, for example, also break the investment universe into baskets and require portfolios to be built from a number of different baskets. Even so, portfolio volatility can change over time as new securities come into the portfolios and underperformers are eliminated. At times, risk is rewarded and portfolios might be loaded with risky securities; in fearful investment climates, portfolios might instead have a large helping of cement-like securities. Along with diversification, having a systematic investment process that adapts to the environment can be helpful in managing risk.
