Diversification is a type of protection for an investor’s portfolio against some risks. But, we should ask: what are these risks we are trying to protect the portfolio from? In terms of portfolio returns, there are only two types of risks: excessive volatility and underperformance. I do not discuss volatility as, for long term investors (multiple years), time takes care of Mr. Market. Underperformance is the risk to worry about. Underperformance includes total loss of value: no one would argue that a -100% return is not underperformance.
So, what number of stocks constitutes appropriate diversification? I argue that it depends on the source of underperformance that the investor is trying to shield the portfolio from. This depends on the investment strategy. All return focused investment strategies largely fall between two camps:
buying/selling stocks within a pool of stocks that has a certain statistical probability of yielding a superior return
buying/selling stocks based on conviction
Diversification that protects against deviations from expected return
If the investment strategy is to trade stocks within a pool of stocks that has a certain statistical probability of yielding a superior return, the purpose of diversification is to reduce the tracking error.
The right level of diversification would be the number of stocks sufficient to approximate the return of the reference pool without having to trade the entire reference pool of stocks. Too low a diversification level would risk randomly selecting a percentage of “losers” greater than in the reference pool, thus yielding inferior portfolio results. If this pool of stocks has a statistical advantage over the general market, there would be no reason to gamble accepting lower than optimal diversification with the risk of excessive losses. An excessive diversification would have operational disadvantages: it could be impractical to analyse all stocks in the reference pool or it could be suboptimal to trade them all.
Some examples of types of investment strategies where this approach to diversification makes sense could be: trading net-nets or buying a selection of companies in an industry that in aggregate is likely to outperform.
Diversification that protects from investor’s mistakes
If the investment strategy is based on the analyst’s conviction that specific stocks will deliver superior returns, the purpose of diversification is to protect the portfolio from the analyst’s mistakes.
For the perfect investment analyst, the right portfolio is 1 stock, no diversification at all.
The ultimate job of an investment analyst is to forecast the stream of cash flows that the company will be able to distribute and the discount rates from now until end of the world. The perfect analyst, that possesses a perfect crystal ball, will be able to see this future with 100% accuracy, including wars and meteorites. Therefore, no diversification is needed for this blessed man or woman.
Clearly the perfect investment analyst does not exist, but this digression sets the lower boundary of diversification and the starting point for this argument. For any real-world analyst, the future of a business will always differ in some ways from the one forecasted. The question that any analyst could try to answer is: once a stock has entered the portfolio what is the probability that the analysis was broadly right and the stocks performs broadly as it was forecasted?
Since we are discussing investments, any deviation from the analyst’s forecasts on the upside should not be too much of a problem for the analyst (unless it leads to investment omission mistakes). Deviations on the downside are definitely the most worrisome and total loss of value is the greatest risk.
Let’s say that, based on an analyst’s track record, there is a 10% probability that any one stock in his portfolio loses 100% of its value every year. Let’s also say that this track record is a good predictor of the analyst’s future performance (not necessarily always the case). How is the risk of total loss of value protected with a 10 and a 5-stock portfolio?
10 stock portfolio. 0.001% probability that half of its stocks lose all their value in one year (equivalent to once every 100,000 years). 0.046% probability that one third of its stocks lose all their value in one year (equivalent to once every 2,154 years).
5 stock portfolio. 0.3% probability that half of its stocks lose all their value in one year (equivalent to once every 316 years). 2.2% probability that one third of its stocks lose all their value in one year (equivalent to once every 46 years).
These examples show how even concentrated portfolios provide good downside protection against the analyst’s greatest risk (i.e., total loss of value). Naturally, the more accurate the analysis, the lower the probability of making mistakes and the lower the diversification needed. The overall portfolio performance would depend on how well the remaining 9 or 4 stocks perform. Which also depends on the accuracy of the analysis.
My conclusion is that, if the analysis is sound, based on deep understanding of the business behind a stock, and if there is a meaningful track record on the quality of the analysis output, excessive diversification is useless. In fact, it is detrimental. Sound analysis takes a lot of time and time is limited (the most precious resource we have). A too large portfolio reduces the soundness of the analysis thus increasing the probability of making mistakes. One could argue that the increased probability of making mistakes is counterbalanced by the higher diversification. My counterargument would be that even if this was the case, we would lose analysis accuracy not only on the downside but also on the upside, which could lead to equal downside protection but inferior overall performance. Focus is key in an investment strategy based on the analyst’s conviction. Diversification is vital, as it reduces risk but it also reduces focus; therefore, should not be overdone.
August 2020