One comment on the benefits to diversification though — there’s a common misconception about what is diversified away. Your ‘risk’ is defined as volatility of returns (lumpiness month to month) which does decay around 15-30 stocks. However what decays much more slowly is how different the returns of a 25th percentile lucky vs a 75th percentile lucky portfolio is, when you increase the number of stocks in a portfolio. In other words, you do not diversify single stock risk until you have something like 50-100 stocks. This paper is the only place I’ve seen this discussed: https://ndvr.com/pdf/documents/NDVR-HowManyStocksShouldYouOwn-November2022.pdf
Thanks a lot Michael! Thanks for the resources. Will definitely look into it!
I agree that the idiosyncratic "luck" component remains high even beyond 20 stocks. Especially if you take timing luck, rebalancing frequencies, non-linear return profiles of sort portfolios, etc. into account. Of course this luck cuts both ways. Having a Russian stock entering your screen in early 2022 or a NVDA enter in early 2023 can break or make your whole year if you are concentrated. I like to delude myself to think that using an integrated multifactor approach mitigates some of the specification risks by diversifying and smoothing the signal. In my opinion finding the balance between the risks and chances is somewhat a personal question with no definite answer.
Nice write up and I do the same including some made up qualitative factors: https://www.detrituscapital.com/small-illiquid-unknowable/
One comment on the benefits to diversification though — there’s a common misconception about what is diversified away. Your ‘risk’ is defined as volatility of returns (lumpiness month to month) which does decay around 15-30 stocks. However what decays much more slowly is how different the returns of a 25th percentile lucky vs a 75th percentile lucky portfolio is, when you increase the number of stocks in a portfolio. In other words, you do not diversify single stock risk until you have something like 50-100 stocks. This paper is the only place I’ve seen this discussed: https://ndvr.com/pdf/documents/NDVR-HowManyStocksShouldYouOwn-November2022.pdf
Thanks a lot Michael! Thanks for the resources. Will definitely look into it!
I agree that the idiosyncratic "luck" component remains high even beyond 20 stocks. Especially if you take timing luck, rebalancing frequencies, non-linear return profiles of sort portfolios, etc. into account. Of course this luck cuts both ways. Having a Russian stock entering your screen in early 2022 or a NVDA enter in early 2023 can break or make your whole year if you are concentrated. I like to delude myself to think that using an integrated multifactor approach mitigates some of the specification risks by diversifying and smoothing the signal. In my opinion finding the balance between the risks and chances is somewhat a personal question with no definite answer.