Five Myths of Active Portfolio Management (Jonathan B. Berk)

[t0x1n]

Consider said paper, available here (at 6 pages it is a very easy read):
http://faculty.chicagobooth.edu/john.co ... k_myth.pdf

The crux of the paper is that manager skill exists, but it is arbitraged away by the size of assets under management. So if a manager is skilled, investors will transfer more and more money to her management, to the point that her skill is offset by the difficulty in managing large amounts of assets (impact costs etc.) back to the benchmark level. This model explains why investors chase past returns - they are trying to exploit the skill of the manager, but it is arbitraged away when everyone tries to do it. Past returns reflect skill, not returns, since the latter are affected by the amount of assets under management. Skill could then be defined as the ability to generate alpha for a given amount of assets.

The paper and the model it suggests are very interesting, but I don't see how they explains the following scenario. Suppose John is a very talented active manager, he's running a small fund and generates significant alpha for a few years. According to the model, a lot of money is transferred to his fund, until he can no longer generate alpha. So far so good. Now suppose John is fired as he is no longer able to generate alpha, and he goes to work for a small fund again. According to the model, so long as the fund remains small, he should be able to generate alpha.

I have a couple of problems with that:
1. To the best of my knowledge, manager performance doesn't persist, regardless of the size of the fund.
2. As far as I know, investors generally chase past fund returns, not manager returns. So supposedly there is an inefficiency and opportunity for profit here by investing in John's new small fund, until the market realizes that the fund is outperforming (due to its skilled manager). Under the paper's model, I'd expect investors to chase individual managers rather than funds in an efficient market.
3. A fund manager could quit the game and simply invest his own money, supposedly outperforming the market persistently since the amount of money he personally has is much lower than the amount that would impair his skill. Again, individual investors aren't known to beat the market persistently so this doesn't make sense. One could claim, though, that once the manager no longer has the resources of the fund, his skill decreases.

The only way I can explain the above is if skill (or more precisely, skill relative to other managers), can change over time. A manager could gain experience and become better, he could gain too much confidence and become worse, newcomer wildcard managers could join the game and change the picture, old managers could quit, the game itself could change thereby altering the factors that constitute skill (e.g. the advent of the internet), etc. Of course, if (relative) skill can change, then chasing it is useless anyway...

Insights will be appreciated

[allsop]

What matters for me is that I'm unable to select the overachieving active managers in the first place, after costs.

When it comes to index (passive) funds there are real skills in managing those to the betterment of the share owners.

[Rick Ferri]

My studies on the issue date back to the 1990s while researching a paper that I wrote on small-cap fund regression to the mean. The tipping point at the time for actively managed small cap was about $300 million under management. There was still some alpha up to about $500 million, and then the small cap managers were unable to add alpha because the size of the fund required changing the strategy. These numbers might be different today. I developed a mutual fund trading strategy that bought newer funds that had $50 - $100 million in assets and sold them when they reached $500 million. It worked on paper, but the cost and time to implement the strategy wasn't worth the effort.

Rick Ferri

[camontgo]

A more detailed explanation of this idea is available in the full paper.

http://finance.martinsewell.com/fund-pe ... en2004.pdf

And here is another follow-up paper by Berk extending this idea:

http://papers.ssrn.com/sol3/papers.cfm? ... id=2038108

1. To the best of my knowledge, manager performance doesn't persist, regardless of the size of the fund.

Berk doesn't measure outperformance the same way that most studies do. He does believe there is persistence (see the second paper for evidence), BUT, it isn't really the kind of persistence that investors can exploit.

Even if you believe Berk, the argument for indexing is still strong.

Example:

SkilledFund earns 2% pre-expense alpha over an index benchmark. The fund has 500M in assets. Fees are 1%, so investors net a 1% alpha. Berk calculates the active return before fees as 2% * 500M or 10M. However, only 5M in active return is earned by investors, the rest is consumed by the fund expenses...i.e. it goes to the manager.

Now assume that money pours into Skilledfund due to a few years of outperformance, assets under management grow to 5B.

Now, due to the large scale of the fund reducing the opportunity set, pre-expense alpha over index benchmark shrinks to 0.5%. Fees are still 1% so investors realize a net alpha of -0.5%. However, Berk says the fund has 0.5% * 5Billion = 25million in active return. So, by Berk's measure, outperformance has actually increased!...but it doesn't do the investors any good. All the active return (and then some) is consumed by the fees. Investors would be better off indexing.

3. A fund manager could quit the game and simply invest his own money, supposedly outperforming the market persistently since the amount of money he personally has is much lower than the amount that would impair his skill. Again, individual investors aren't known to beat the market persistently so this doesn't make sense. One could claim, though, that once the manager no longer has the resources of the fund, his skill decreases.

A manager is better off charging 1% fee on a large fund, even if alpha is negligible, than he is if he managed a smaller personal fund and generated an alpha of several percent.

For example,

If a manager could generate 3% alpha on a $10M portfolio, then his efforts would earn him an extra $300,000 per year on average. But, if he charges 1% fee on a $500M fund...regardless of whether there is any alpha... then the annual income is $5M. Why would he choose to quit this game to manage his own money?

[t0x1n]

@Rick Ferri

I developed a mutual fund trading strategy that bought newer funds that had $50 - $100 million in assets and sold them when they reached $500 million. It worked on paper, but the cost and time to implement the strategy wasn't worth the effort.

The strategy you describe sounds very interesting. However, if I understand correctly, it includes no "personal manager tracking" which is what I'm trying to refute here. Also, it will be interesting to know how such a strategy would work in tax-free accounts, where the costs are minimized (or even zeroed in the case of no-load funds?)

@camontgo

Now, due to the large scale of the fund reducing the opportunity set, pre-expense alpha over index benchmark shrinks to 0.5%. Fees are still 1% so investors realize a net alpha of -0.5%. However, Berk says the fund has 0.5% * 5Billion = 25million in active return. So, by Berk's measure, outperformance has actually increased!...but it doesn't do the investors any good. All the active return (and then some) is consumed by the fees. Investors would be better off indexing.

Looking at the absolute return, rather than percentage - an interesting approach ! It reminds me of another theory I read in some article by Larry Swedroe I believe, that the absolute amount of "alpha money" is limited. For example, each year there are only $X of actively managed money that outperform the index. However, it still doesn't explain the possible strategy of following the manager to smaller funds - unless such manager relocations never happen (from big to small).

A manager is better off charging 1% fee on a large fund, even if alpha is negligible, than he is if he managed a smaller personal fund and generated an alpha of several percent.

Agreed. However, what's bothering me is the fact that supposedly he could do it if only he desired. I don't believe that any individual, regardless of his background, can predictably beat the market.

@allsop

What matters for me is that I'm unable to select the overachieving active managers in the first place, after costs.
When it comes to index (passive) funds there are real skills in managing those to the betterment of the share owners.

We are definitely on the same team here. I'm trying to poke holes in the specific model proposed by the paper, not in passive investing in general - of which I am a great advocate.

[Rick Ferri]

@Rick Ferri

I developed a mutual fund trading strategy that bought newer funds that had $50 - $100 million in assets and sold them when they reached $500 million. It worked on paper, but the cost and time to implement the strategy wasn't worth the effort.

The strategy you describe sounds very interesting. However, if I understand correctly, it includes no "personal manager tracking" which is what I'm trying to refute here.

Actually, it did. The idea was to isolate small cap managers who were successful at one company and either moved to another or started their own firm. One problem with this strategy was that there wasn't many of these managers. A second problem was determining whether their past performance was skill or luck.

Recently the SEC is requiring manages to disclose the amount of personal assets they have in their funds. This data is encouraging as a way to sort small funds, but then fund companies can game the system if they see assets flowing to funds where the manager had a large $ position. I wrote about this idea in The Power of Passive Investing. David Swensen has also talked about it in one of his books.

In the end, indexing works just fine.

Rick Ferri

[t0x1n]

Actually, it did. The idea was to isolate small cap managers who were successful at one company and either moved to another or started their own firm. One problem with this strategy was that there wasn't many of these managers. A second problem was determining whether their past performance was skill or luck.

I stand corrected, that is indeed exactly what I had in mind. Did you catch my refined question in the edit? I wondered how this strategy would fare in tax-free / tax-deferred accounts, where costs would be minimized.

Recently the SEC is requiring manages to disclose the amount of personal assets they have in their funds. This data is encouraging as a way to sort small funds, but then fund companies can game the system if they see assets flowing to funds where the manager had a large $ position. I wrote about this idea in The Power of Passive Investing. David Swensen has also talked about it in one of his books.

I'm not sure I follow. Knowing how much a manager has invested in his own fund only tells you something about his belief in his future performance doesn't it? Why is that relevant to the distinction between skill and luck ?
Also, how would fund companies game the system? Force the managers to hold large positions in the funds? (perhaps by paying part of their compensations in fund stocks?)

Regardless, both books are on my Wishlist (assuming you mean Winning the Loser's Game in Swensen's case)

[Rick Ferri]

Actually, it did. The idea was to isolate small cap managers who were successful at one company and either moved to another or started their own firm. One problem with this strategy was that there wasn't many of these managers. A second problem was determining whether their past performance was skill or luck.

I stand corrected, that is indeed exactly what I had in mind. Did you catch my refined question in the edit? I wondered how this strategy would fare in tax-free / tax-deferred accounts, where costs would be minimized.

Terrible in a taxable account because of turnover. Also, actively managed small cap funds are more expensive than index funds. This extra cost needed to be overcome before earning any excess return. As I said, indexing looks pretty good to me now, but it was an interesting study.

Recently the SEC is requiring manages to disclose the amount of personal assets they have in their funds. This data is encouraging as a way to sort small funds, but then fund companies can game the system if they see assets flowing to funds where the manager had a large $ position. I wrote about this idea in The Power of Passive Investing. David Swensen has also talked about it in one of his books.

I'm not sure I follow. Knowing how much a manager has invested in his own fund only tells you something about his belief in his future performance doesn't it? Why is that relevant to the distinction between skill and luck ? Also, how would fund companies game the system? Force the managers to hold large positions in the funds? (perhaps by paying part of their compensations in fund stocks?) Regardless, both books are on my Wishlist (assuming you mean Winning the Loser's Game in Swensen's case)

Knowing how much a manager has invested in his own fund MIGHT tell us something about their confidence in their investment skills. There was a study by a Allison Evans at Duke who measured manager ownership relative to fund returns and found positive alphas in the funds were the managers had personal stakes over $1 million. In Pioneering Portfolio Management; 2nd edition 2009 Swensen discusses how fund companies would game the system if manager stake stated to become popular way for investors to select funds. They would beef up manager ownership with loans to make it look like managers held a larger stake then what they actually held.

Just another idea - and a reason not to do it. Those index funds are looking mighty good.

Rick Ferri

[bengal22]

I have always held the belief that due to the efficiencies of the market that the biggest difference between the active and the passive fund of similar type is the expense ratio. In other words, over time the active manager will not overcome the higher expense ratio that his fund has due to it being actively managed.

[t0x1n]

@Rick

Also, actively managed small cap funds are more expensive than index funds. This extra cost needed to be overcome before earning any excess return.

So in short, the pre-expenses alpha generated from such a strategy is not translated to after-expenses excess alpha. In other words, the market equilibrium is achieved with consideration to costs (basically weak EMH). By the way, was there any reason in particular you targeted small-cap funds?

There was a study by a Allison Evans at Duke who measured manager ownership relative to fund returns and found positive alphas in the funds were the managers had personal stakes over $1 million. In Pioneering Portfolio Management; 2nd edition 2009 Swensen discusses how fund companies would game the system if manager stake stated to become popular way for investors to select funds. They would beef up manager ownership with loans to make it look like managers held a larger stake then what they actually held.

I guess one study wasn't enough for it to become a popular factor in selecting funds. The results of the study actually don't make sense to me - they imply that the manager "knows" somehow how well she'll do, and due to that knowledge increases her stakes.

@bengal22

I have always held the belief that due to the efficiencies of the market that the biggest difference between the active and the passive fund of similar type is the expense ratio. In other words, over time the active manager will not overcome the higher expense ratio that his fund has due to it being actively managed.

No doubt. However, I'm not seeing how you can reach these conclusions in a definite fashion from the model proposed in the paper. I remain unconvinced regarding at least two of the three problems I've stated (Rick went a long way to explain #2).