Thanks for sharing. Let me quickly reply to each of the criticisms:
1. Concentrated portfolios have more risk: Absolutely agree and no quarrels with that.
2. Skill will be harder to deteft: this argument is based on the so-called fundamental law of active management that says Information Ratio = Information coefficient x SQRT(breadth). This equation is empirically invalid. First it assumes that a manager has universal information coefficient that is constant independent of how many bets she makes. The research on high conviction ideas shows that this is exactly NOT the case. Second it assumes that more bets (more stocks chosen) increase breadth because these stocks are not correlated with each other. This is not true in real life because fund managers tend to select stocks along a theme or in a specific sector where breadth is hardly growing at all if you pick more stocks. Furthermore in a crisis/bear market correlations go towards 1 thus reducing any breadth effect to nil just when you need it most.
3. Trading costs will rise: Huh? The example they give is so far removed from how fund managers invest that it is hard to take seriously. Yes, if you have larger positions in any single stock your trade size increases and if you have a mega fund of several billions it might be a problem. But we have solved that problem a long time ago. How do you think fund managers in small cap and micro cap funds do it. They have had that problem forever. Furthermore trading costs are also a function of liquidity, not only trade size. And in large cap stocks the index funds and noose traders provide so much liquidity that I have a hard time seeing trading costs increase more than one basis point.
4. Probability of active underperformance: correct. But that is the whole point. The probability of underperformance increase for all managers when the portfolios become more concentrated but it increases relatively more for managers with no skill than form managers with skill. And that is exactly what I am advocating for. In statistical terms, they argue that your Type II error increases while ignoring the fact that at the moment we have a massive Type I error which would decrease much more than the increase in type II error.
So overall I find this note not very convincing and I fact self serving. After all S&P is an index provider who makes money out of investors benchmarking their portfolios against their indices.
This is a really interesting question and it triggered a long thought process in me that kept me from sleeping (which is a good thing). I will write a short blog post about it next week, but I think I will also develop it into a full academic paper, because the math turns out to be really interesting. So please give me a week until the post with my response is published.
And thanks very much for the challenge. This is awesome :-)
JK - I am interested to learn what you think of this take on "high conviction" (concentrated) portfolios. Best, Pete
https://www.spglobal.com/spdji/en/documents/research/research-fooled-by-conviction.pdf
Thanks for sharing. Let me quickly reply to each of the criticisms:
1. Concentrated portfolios have more risk: Absolutely agree and no quarrels with that.
2. Skill will be harder to deteft: this argument is based on the so-called fundamental law of active management that says Information Ratio = Information coefficient x SQRT(breadth). This equation is empirically invalid. First it assumes that a manager has universal information coefficient that is constant independent of how many bets she makes. The research on high conviction ideas shows that this is exactly NOT the case. Second it assumes that more bets (more stocks chosen) increase breadth because these stocks are not correlated with each other. This is not true in real life because fund managers tend to select stocks along a theme or in a specific sector where breadth is hardly growing at all if you pick more stocks. Furthermore in a crisis/bear market correlations go towards 1 thus reducing any breadth effect to nil just when you need it most.
3. Trading costs will rise: Huh? The example they give is so far removed from how fund managers invest that it is hard to take seriously. Yes, if you have larger positions in any single stock your trade size increases and if you have a mega fund of several billions it might be a problem. But we have solved that problem a long time ago. How do you think fund managers in small cap and micro cap funds do it. They have had that problem forever. Furthermore trading costs are also a function of liquidity, not only trade size. And in large cap stocks the index funds and noose traders provide so much liquidity that I have a hard time seeing trading costs increase more than one basis point.
4. Probability of active underperformance: correct. But that is the whole point. The probability of underperformance increase for all managers when the portfolios become more concentrated but it increases relatively more for managers with no skill than form managers with skill. And that is exactly what I am advocating for. In statistical terms, they argue that your Type II error increases while ignoring the fact that at the moment we have a massive Type I error which would decrease much more than the increase in type II error.
So overall I find this note not very convincing and I fact self serving. After all S&P is an index provider who makes money out of investors benchmarking their portfolios against their indices.
Thanks. So let me ask, what are your thoughts on Henrik Bessembinder's findings in regard to the concentration argument?
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3537838
Cheers,
Pete
This is a really interesting question and it triggered a long thought process in me that kept me from sleeping (which is a good thing). I will write a short blog post about it next week, but I think I will also develop it into a full academic paper, because the math turns out to be really interesting. So please give me a week until the post with my response is published.
And thanks very much for the challenge. This is awesome :-)