Who keeps the alpha

In the early eighties, the people who ran big pension funds spent almost all of their meeting time picking managers. Every quarter a room full of trustees reviewed a dozen of them, fired the two worst, hired two more, and went home. The bigger question, how much of the fund should be in stocks, in bonds, in anything else, was overlooked.

The five papers in this chapter came out over twenty-five years, and read together they describe the whole business of managing other people’s money. Where results come from. Whether the average manager can win. What a winning manager has to possess. Why even real skill rarely reaches the person who pays for it. And what the numbers show across thousands of funds.

It is an unsentimental picture. I paid for the lessons in it with real fees before I read the papers.

Brinson, Hood and Beebower, 1986

Gary Brinson and two colleagues took ninety-one large American pension plans, ten years of quarterly returns from 1974 to 1983, and split each plan’s result into three pieces. Policy, meaning the long-term mix of stocks, bonds and cash, held in index funds. Timing, meaning the moves away from that mix. And selection, meaning the managers’ choices inside each asset class.

Two numbers came out. The policy mix explained, on average, ninety-three point six percent of how each plan’s returns moved over time. And the active part, timing plus selection, cost the average plan one point one percent a year. Timing lost sixty-six basis points, selection lost thirty-six. The trustees were spending their hours on the piece that subtracted value.

Then the paper got misquoted. You will hear that asset allocation determines ninety percent of your returns. It does not, and Brinson himself spent years saying so. The ninety-three percent is about how one plan’s returns move up and down from quarter to quarter, and most of that movement is simply the market moving. It says nothing about the level of returns, and it does not explain why one plan did better than another. When Roger Ibbotson and Paul Kaplan checked that second question in 2000, policy explained about forty percent of the differences between funds. The rest was managers, timing, and costs.

The correct reading is still worth a lot. For a diversified investor, the level of risk you choose and the mix you hold are where outcomes are made, and the manager decisions are small next to them. In most families I have seen, the hours are inverted: most of the year goes to funds and people, a few hours to the mix. Fixing that ratio is the cheapest improvement I know.

Sharpe, 1991

William Sharpe settled the question of the average manager in three pages, with no data at all. The argument is accounting.

Every dollar in a market is held by someone. Split the holders into two groups. Passive investors hold the market as it is and earn the market’s return. Everyone else is active. Since the market’s return is the average of what all holders earn, and the passive group earns exactly that average, the active group as a whole must earn it too, before costs. After costs, the active group must earn less, because active management costs more. That is the whole paper.

Notice what it does and does not say. It does not say skill does not exist. It says skill cannot be average. For one active manager to beat the market, another has to trail it, and the winner’s gain comes from the loser’s pocket, minus what both of them paid to play. That holds in every market, including the ones described as inefficient. Someone has to hold the losers there as well.

So the question to ask before hiring any active manager is the one Sharpe forces on you. Which other investors are going to lose so that this one can win, and why do I believe I can pick the winner in advance. If there is no answer, the expected result is the market minus the fee.

Grinold, 1989

Richard Grinold asked the next question. If the average manager must lose, what does an above-average one need. His answer is a formula, and it can be said in words. A manager’s risk-adjusted excess return equals skill per bet, times the square root of the number of independent bets made per year.

Skill per bet means how well the manager’s forecasts line up with what happens. Realistic skill is tiny. A correlation of two to five hundredths between forecast and outcome is roughly the ceiling that has ever been documented. Since skill is small and hard to increase, the scalable ingredient is breadth: many bets that do not depend on one another. A hundred independent decisions a year with tiny skill can produce a fine record. Ten big decisions need skill nobody has shown.

That gives you a test for concentrated managers. Take the excess return they claim, divide by the square root of their real number of independent bets, and look at the skill the claim implies. Usually it is beyond any human being on record. And the number of bets is smaller than the number of positions, because twenty stocks that all rise when rates fall are one bet, made twenty times.

For a family the law draws a line. You cannot manufacture breadth in public markets, so you hold them passively. The exception it allows is the handful of private situations where you know far more than the market does, in a business, a piece of land, a person. There, skill per bet is high and breadth is low, and concentration is what the formula recommends, provided the skill is real and you have written down why.

Berk and Green, 2004

By 2004 the industry had a paradox. Investors chased past performance as if skill existed. Performance did not persist. Most people read the second fact as proof that skill was a fiction. Jonathan Berk and Richard Green showed that both facts follow from skill being real.

Their model has three parts. Managers differ in skill. Skill runs out as a fund grows, because there are only so many good ideas and a bigger fund moves prices against itself. And money is free to flow. Put them together. A manager shows skill, money pours in, the fund grows until the extra size eats the extra return, and the investor who arrives after the record is established gets the market minus the fee. The manager, meanwhile, collects fees on a fund that is now enormous. Skill exists. It is paid for. It is paid to its owner.

I find this the most useful paper in the chapter, because it changes the question. Chasing records is a strategy for making managers rich. The question about any successful fund is what stops it from growing to the point where the edge is gone: a hard limit on size, a closed door, or terms you negotiated before the flows arrived. A family’s realistic path to keeping some of the skill rent is structural. Early money, capacity agreements, fee breaks, co-investment. Never a public track record.

Fama and French, 2010

The verdict came from Eugene Fama and Kenneth French, who took every American equity mutual fund with at least five million dollars, 3,156 funds, over twenty-three years from 1984 to 2006. In aggregate, after fees, the funds trailed their benchmarks by about eight tenths of a percent a year. Add the expense ratios back, and the aggregate was a tenth of a percent above. The managers, as a group, earned roughly their costs and passed the bill to the client.

The clever part was the test for luck. With thousands of funds, some will show wonderful records by chance alone. So the authors simulated ten thousand worlds in which no fund had any skill, produced the distribution of records those worlds generate, and laid the real distribution on top. After fees, the real funds sat to the left of pure luck almost everywhere. In the extreme right tail there was a faint trace of skill, a few funds in a hundred, and even there nobody could name which ones in advance.

The best-known live experiment ran while the paper was in press. At the end of 2007 Warren Buffett bet that a plain index fund would beat any five hedge funds anyone chose, net of fees, over the ten years starting in 2008. One firm accepted, and picked five funds of funds that between them held more than two hundred hedge funds. The hedge funds won 2008, the crisis year, and lost every year after. By the end of the ninth year, a million dollars in the funds of funds had gained two hundred and twenty thousand. In the index fund, eight hundred and fifty-four thousand. The tenth year did not change the picture.

What I do

The policy mix is written down, in round numbers, with ranges and the rule for rebalancing, and the year’s work is mostly keeping it. Public equities and bonds are held passively. When I pay an active fee, I can say in one sentence where the manager’s breadth or knowledge comes from, who is on the other side, and what stops the fund from growing past its edge. If I cannot say it, I do not pay it.

Where active management has worked for me is in real estate, private equity and other alternatives, mostly in small, underdeveloped markets with few professionals and real mispricings to be found. Few bets, and a lot of knowledge about each one. That is the part of Grinold’s law a family can use. I do not know how long it will last. Markets keep getting more efficient, and AI is making them more so, and the inefficiencies I built on may be gone by the time the next generation goes looking for them.

Three habits

Write the policy and price your attention. Every proposal that reaches your table should say whether it is a policy decision, a timing decision or a selection decision. The first kind deserves most of your hours.

Ask the Sharpe question before hiring. Whose underperformance will fund this manager’s outperformance, and how will you identify the losers in advance. No answer, no mandate.

Split every fee in two. Part of it pays for exposure you could buy passively for almost nothing; the rest is a charge for claimed alpha. Only the second part has to be earned, and the evidence has to be stronger than what luck would produce across the thousands of funds you did not pick.

Gala: get the mix right and defend it. Pay for skill only when you can say where it comes from and why it will reach you instead of stopping in someone else’s pocket.