Tuesday, November 19, 2013

Risk Premiums, Reputation, and Actively Managed Mutual Funds

In my previous post, I referenced this FAJ article by Antti Petajisto (Financial Analysts Journal, July/August 2013, Vol. 69, No. 4: 73–93) about the returns to active fund management.  The author found the typical result, that fund managers do have gross returns on average that beat the their benchmarks by about .96%, but that after fees, they lag their benchmarks by -.41%.

But, the author finds that the most active funds are the best performers.  I suspect that we are looking at the same elements that lead to persistent excess returns among low-status (small, low book value, etc.) firms, and that since the mechanisms that cause these varying risk premiums can't play out through fund NAV's, they are reflected in the level of fees relative to the funds' gross alpha.

More after the jump.

An efficient market will still have active funds

First, a parenthetical point.  It seems to me that in an efficient market, there would be some positive return to active management, reflecting the cost of information necessary to make a market efficient, and that practically all of that positive return would be earned by the practitioners.  In an efficient market, made efficient by the work of active securities analysts, active funds would have positive gross returns roughly equal to their fees.  Index funds aren't free.  They have fees from .1% to .4%.  So, the performance of these active funds seems similar to what we should expect in a functioning fund market.

Even so, one could argue that for an individual investor, buying index funds is an easy call - there is no gain to buying active funds.  That is true in the aggregate.  But, in an efficient market, there will have to be some funds actively trading in order to bring about efficiency.  Those funds will need to be owned by some set of investors, so a market in equilibrium will have some set of active funds, the prices of which will have been bid down to a point where there is little or no return, on average, compared to passive funds.  The achievement of this equilibrium won't mean that active funds go away.

Truly active funds outperform

All that being said, there are some interesting things to think about from this research regarding negative net alpha in actively managed portfolios.  The author disaggregated the funds among these categories:
  1. Closet Indexers: Low active share and low tracking error
  2. Moderately active: moderate active share and low tracking error
  3. Factor bets: low or moderate active share and high tracking error
  4. Concentrated: very active with high tracking error
  5. Stock Pickers: very active with low tracking error
Here is a table of their performance:

I see three issues behind this range of outcomes.  Some of these are discussed by the author, and I recommend reading the whole article.

First, group 4, in practice, appears to be somewhat of an amalgam of Groups 3 and 5, so I won't go into it separately here.

Group 3 represents funds that are fairly diversified, but that are creating exposure to certain macro-phenomena, some of which are manifest through exposure to common return factors, as can be seen in the results above.  These funds represent about 21% of total assets.  They have the worst returns of any group, before or after expenses.  The returns they do achieve come about from specific factor exposure (evidenced by the fact that their factor-adjusted returns are significantly lower than their gross returns).  But, the relative returns they achieve don't make up for expenses.
One could say that these funds represent a clear failure of money chasing alpha, as almost all of the expense is fruitless.  But, these funds are providing some hedging value in relation to whatever factors or macro-trends they are positioning themselves with.  The scale of that value is an empirical question beyond the scope of this analysis.  But, between Groups 3 and 4, the net cost of those factor exposures accounts for a third to half of the aggregate average relative net loss for all fund types, which puts the net return of non-factor-based funds at about the average return for a benchmark passive index fund, net of average passive fund expenses.  The bulk of the question about the value of active fund management could come down to the question of whether factor hedges are efficiently priced.

Groups 1 & 2 represent about 71% of fund assets.  These funds take some tactical positions, but they are, in effect, funds that are mostly invested in an index-tracking portfolio, with some tactical adjustments around the edges.  On average, they make the right tactical adjustments, so they are hybrid funds consisting of an active fund that, more or less, earns its fees, and an index fund that charges active fund fees.  They find useful information and act on it.  They just don't commit all of the funds' resources to it.  Note that the moderately active group outperforms the closet indexers.

Group 5 only represents about 6% of fund assets (8%, if combined with Group 2 concentrated funds).  These are funds fully committed to active management.  The average Group 5 fund is 97% active.  These 180 funds had an average annual gross return of 2.61% above their benchmarks, and 1.26% net returns above their benchmarks.  Funds fully committed to active management easily beat their benchmarks on average.

And, performance is persistent among all groups.  Last year's best funds in each category tend to be this year's best funds. Net of fees, all 5 quintiles of Closet Indexers, based on last year's performance, underperform their benchmarks.  For Moderately Active funds, only the top quintile from the previous year outperforms the following year.  For stock pickers, only the bottom quintile underperform the following year, and even those only underperform by -.26%, net of fees.  The other 4 quintiles outperform their benchmarks by .78% to 2.93%, net of fees.

When adjusted for four-factor alpha, the persistence is more shallow.  Here, all 5 quintiles of Closet Indexers provide an average net relative return of about -1% the following year.  All 5 quintiles of Stock Picker funds outperform their benchmarks the following year, on average, from .87% for the bottom quintile to 1.87% for the top quintile, net of fees.

So, the relative performance of each category of funds is relatively broad-based across the range of funds in each category.  Stock pickers are very likely to outperform and closet indexers are very likely to underperform.

Reputation is important

Two other interesting findings were:

  1. Fund age correlated negatively with returns.  Older funds did worse.  One might suspect that this is the result of massively successful funds that become too large to manage deftly, but the author found no systematic correlation between size and returns within each category.
  2. Expenses correlated with returns, with coefficients greater than 1.  In other words, funds with higher fees generate lower gross returns, even before subtracting the higher fees.

I think, taken together, all of these findings point to the importance of reputation and status.

The factors that have been found to provide persistent excess returns in individual stocks - small size, low book to market, and momentum - all reflect status to a degree.  In some ways, these are all measuring a kind of self-reinforcing herding behavior.  Expected return is inversely proportional to relative valuation.  Low status firms get low valuations, hence high returns, and if their sentiment improves, they experience additional increased valuations from that.  In some ways, this is a mathematical truism.  But, the scale of the effect is a product of human herding tendencies.  I don't mean that disrespectfully.  Social animals herd because it's the smart thing to do, for many subtle and complicated reasons.

Why shouldn't this also apply to funds, just as it does to stocks?  As I mentioned, I submit that we are looking at three different issues, represented by three different groups of funds.

1) Factor hedging:  Here, there may be legitimate premiums that are being measured as underperformance.

2) Reputational/Status funds: If we think in terms of risk premiums or discount rates, funds with status might reasonably provide a lower return because their status fetches a higher price.  In funds, this higher price can't be manifest through the NAV.  The only way for it to be realized is through fees.  These funds help to address trust issues that are endemic in such an emotional and important field as financial management, that is so dependent on agents and middle-men.  These funds represent the bulk of underperforming active funds, but the underperformance tends to average around 1% or less for any broad cross-section of these funds.

I am making a subtle point here.  My intention is not to deny that the growth of truly passive funds is a vast improvement over the old days.  Thirty years ago, practically all funds were actively managed.  Passive funds are gaining status relative to active funds, so the status premium that used to mean something to fund investors is getting bid out of this group of funds.  That is a good thing.

This development is similar to the commoditization of broker services.  It used to cost $100 to have your broker enter an order to buy or sell shares of stock for you.  Now, this service is commonly available online for less than $10 a trade.  This isn't because we used to be stupid.  It's because in the pre-internet age, reputation was difficult to parse, and it was important.  A broker had to establish experience and develop affiliations that demonstrated his ability to honestly and efficiently handle trades.  These issues can be resolved much less expensively today, so we don't pay for them.

The same thing has happened in funds.  Financial commoditization has made status and reputation less important, and so commodified funds have become more feasible.  Past investors weren't necessarily stupid.  They just depended on reputation more.  And one way reputation was signaled was by actively managing a family of funds in a responsible way.  You don't build reputation by going 95% active and risking a negative outlier year.  But, you also don't build reputation by being passive.

Along this line of thinking, the older firms with higher fees in this study achieved lower net relative returns because their risk premiums were lower, just like individual large, high market value firms fetch a higher price (lower risk premium).  Over time, these funds are being replaced by passive index funds.

3) True active funds:  These are low status funds.  As with high return stocks, they tend to be smaller.  They hold fewer positions and thus risk higher tracking error, which, again, over the long term, represents a bias for performance over reputation.  This means that the market assigns a high risk premium to them, and so their fees must be much lower than their relative returns so that the risk premium is reflected in their higher net earnings.  They are providing alpha at the fund level through active portfolio management.


It looks, at first glance, as though there is persistent underperformance among active funds.  But, this is because we are not universally acknowledging the large effect of variance in required returns at all levels of fund management.  Varied risk premiums reflecting reputation and strategic factor hedging could be associated with legitimate reductions in required returns, which can only play out through fund fee levels.  The ongoing development of efficient financial services should continue to reduce the market for Closet Index funds that earn higher fees through reputation.  Truly active funds that don't fetch a reputational premium will continue to outperform, as they have in the past.

As with individual stocks, a fund investor can persistently outperform, but systematic outperformance is achieved by taking on non-financial risk, such as reputational risk, which is frequently difficult, especially where agency issues dominate.


  1. This comment has been removed by a blog administrator.

  2. This comment has been removed by a blog administrator.