The 93.6 Percent Number That Everyone Misquotes: Brinson, Hood and Beebower

There is one statistic that has been repeated in more investment brochures than any other: asset allocation explains 93.6 percent of portfolio performance. It comes from a 1986 paper by Gary Brinson, Randolph Hood and Gilbert Beebower, and it has been used for four decades to justify everything from index funds to target date products to charging fees for allocation advice.

The number is real. The paper is real. And the industry's interpretation of it is, in an important respect, wrong. Understanding both the finding and the misreading is one of the more useful things a quant can learn, because it is a beautiful case study in how a precisely stated statistical result gets mangled the moment it leaves the journal.

The problem: where does a pension fund's performance actually come from?

A large pension plan makes three broad kinds of decision:

Investment policy. The long-run strategic mix: we will hold roughly 60 percent equities, 30 percent bonds, 10 percent cash. This is the boring decision made once and revisited rarely.
Market timing. Deviating from that mix tactically. Equities look expensive, let us go to 55 percent for a while.
Security selection. Within each asset class, picking which specific stocks and bonds to hold.

Almost all of the industry's energy, fees, and glamour go into decisions 2 and 3. Brinson, Hood and Beebower asked a simple question: how much do each of them actually matter?

The key idea via analogy: the recipe versus the seasoning

Think of a portfolio as a meal. The policy mix is the recipe: what ingredients, in what proportions. Market timing and security selection are the seasoning: small adjustments the chef makes on the day.

The authors built a clean framework to separate the two. For each of 91 large US pension plans over the years 1974 to 1983, they constructed a hypothetical benchmark portfolio: one that simply held the plan's own long-run average asset mix, but invested passively in index returns for each asset class. That benchmark represents "what you would have gotten from your recipe alone, with no chef."

Then they compared the plan's actual returns against that benchmark and attributed the difference to timing and selection.

The headline result: across those plans, the passive policy benchmark explained on average 93.6 percent of the variation in total plan return over time. Timing and selection together accounted for the small remainder, and in the aggregate they subtracted rather than added value.

Two conclusions followed, and both are genuinely important:

The strategic mix dominates. How much you have in stocks versus bonds swamps the effect of which stocks you picked.
Active decisions did not pay. On average, across these large, professionally managed, well-resourced plans, market timing and security selection collectively destroyed value relative to just holding the policy mix passively. Not by a lot, but the sign is the point.

Now the part everyone gets wrong

The claim "asset allocation explains 93.6 percent of performance" is not what the paper found. The paper found that asset allocation policy explains 93.6 percent of the variation in returns over time within a given plan. Those are very different sentences, and the difference matters enormously.

Here is what the result actually says. Take one pension plan and look at its quarterly returns over ten years, bouncing up and down. Now ask: how much of that bouncing is explained by the fact that it holds a mix of stocks and bonds, and stock and bond markets bounce? Answer: almost all of it. When the market goes down 10 percent, the plan goes down roughly in line with its equity weight, whatever its manager was doing that quarter.

But that is a statement about movement over time, and it is close to a tautology once you say it plainly: if you hold a lot of equities, your returns will move when equities move. Of course they will. The finding does not say that asset allocation determines how much money you end up with relative to another plan, and it certainly does not say that 93.6 percent of your return level comes from allocation.

Later authors, most notably Ibbotson and Kaplan, made this decomposition explicit by asking three separate questions that the original number is often wrongly used to answer at once:

How much of the variation in a fund's returns over time does policy explain? Roughly 90 percent. This is the Brinson result.
How much of the variation in returns across different funds does policy explain? Far less, something closer to 40 percent, because different funds have different managers making different active bets, and those bets are what differentiate them.
What percentage of a fund's return level comes from policy? Roughly 100 percent on average, essentially because active management is a zero-sum game before costs.

Three different questions, three different answers, and the marketing brochures cheerfully use the first answer to make claims about the second.

Why it mattered

It made asset allocation the main event. Before this paper, the profession's attention was overwhelmingly on stock picking. After it, the strategic mix became the central decision, and it is now the first thing any serious investment policy statement addresses.
It provided early, credible evidence against active management. These were not naive retail investors. These were 91 large pension plans with professional managers and real resources, and their timing and selection decisions collectively subtracted value. That result has been replicated in many forms since.
It built the foundation of performance attribution. The framework of splitting returns into policy, timing, and selection is now the standard language of institutional performance reporting.
It is a permanent lesson in reading statistics carefully. The gap between "explains variation over time" and "determines your outcome" has probably supported more misleading marketing than any other misreading in finance. Knowing exactly what the number means, and does not mean, is a genuine professional skill.

The honest limitations

The headline statistic is routinely misused, as above. The single most important thing to know about this paper is what its central number does not say.
The sample is narrow. Ninety-one large US pension plans over ten years, from 1974 to 1983. These are big institutions with similar mandates, similar constraints, and fairly similar policy mixes. That homogeneity is part of why cross-fund differences look small.
The plans had similar allocations, which suppresses the measured importance of allocation. If every plan in your sample holds roughly 60/40, then allocation cannot explain much of the difference between plans, almost by construction. Run the study across a set of funds with wildly different mixes and the numbers change.
The active management result is an average. That some managers subtract value on average does not mean none add it, though the broader literature is not encouraging on that front.
"Variation explained" is an R-squared, and R-squared is a slippery thing to build an investment philosophy on. A high R-squared tells you that two series move together. It does not tell you which decisions were valuable.

The one-line takeaway

Brinson, Hood and Beebower showed that a pension plan's long-run policy mix explains the overwhelming majority of how its returns move over time, and that timing and selection added nothing on average, but the famous 93.6 percent figure has been widely misread as saying that asset allocation determines how much money you make, which is a different claim the paper never made.