Diversification, or Buying the Median

Diversification behaves differently inside private markets than across asset classes. Because returns are skewed, each added manager pulls a program toward the median rather than the mean. Addepar research points to three to six commitment-based positions per portfolio objective, with more risking return dilution. Manager count should be an output of underwriting capacity, not a policy set in advance.

Diversification is the least contested idea in institutional investing, so it is rarely examined. In public equities, the logic is settled: adding positions reduces idiosyncratic risk without reducing expected return, so breadth is close to free. Most private markets programs assume the same logic carries over.

It does not carry over cleanly because the distribution's shape is different. In a symmetric distribution, the average is the typical outcome. In private markets, a small number of results at one end drive the average, so broad exposure converges toward the median rather than the mean.

Key takeaways

A typical private equity fund holds roughly a dozen portfolio companies over its life, so each position carries meaningful weight in the outcome.
PitchBook data indicate that one to three investments in a venture portfolio of twenty to thirty generate between 50 and 80 percent of total return, with the remainder clustered between zero and two times capital.
Addepar research concludes that three to six commitment-based fund positions per portfolio objective balance cash-flow diversification against return dilution, and that portfolios of six or more risk over-diversification.
CBRE Investment Management reached a similar conclusion in core real estate, finding three to five funds sufficient and observing that larger portfolios dilute the information ratio as managers converge toward the average.
The counterargument deserves weight. Verdad Advisers found that observed private equity dispersion resembles what randomly constructed portfolios of about twenty levered micro-cap companies would produce, with managers trailing a factor-matched benchmark by two to four percentage points annually.
Manager count is an output of underwriting capacity, not a policy. Concentration is rational only to the extent a selection process produces information the market does not already price.

Why does a skewed return distribution change the math?

A typical private equity fund holds roughly a dozen portfolio companies over its life.^¹ In venture, PitchBook data indicate that one to three investments in a portfolio of twenty to thirty generate between fifty and eighty percent of total return, with the remainder clustered between zero and two times capital.² Cambridge Associates data place the spread between upper and lower quartile venture returns above thirty percentage points in most vintage years, roughly three times the equivalent buyout spread.²

That structure has an uncomfortable implication. A program designed to reduce the probability of holding a poor manager also reduces the probability of achieving the outcomes that produce the asset class average. Breadth does not merely trim the left tail. It trims both, and in a skewed distribution, the right tail is where the allocation rationale lives.

The published work on optimal fund count falls well below where most programs sit. Addepar's research on fund diversification concludes that three to six commitment-based fund positions per portfolio objective balance cash-flow diversification against return dilution, and that portfolios with six or more positions risk over-diversification, reducing the chance of capturing manager-specific return while adding operational burden.^³ CBRE Investment Management reached a similar conclusion in core real estate, finding three to five funds sufficient and observing that larger portfolios dilute the information ratio as stronger and weaker managers converge toward the average.^⁴

Why do private markets programs end up holding so many managers?

Few allocators set out to hold thirty managers. Programs arrive there because each individual commitment is defensible. A new vintage requires deployment. A long-standing relationship expects a re-up. A committee is more comfortable approving a smaller position in an unfamiliar manager than a larger one. A specialist strategy fills a perceived gap. None of those decisions is unreasonable in isolation, and the cumulative result is a portfolio whose expected outcome is the market median, less two layers of fees.

The drift is rarely visible because standard reporting measures nothing to show it. A program can report a respectable pooled return while holding a structure that, by construction, has eliminated its ability to produce an exceptional one.

What is the strongest case against concentrating?

The case for concentration deserves its strongest counterargument, and it has one. Verdad's analysis of private equity dispersion found that the observed spread between private equity funds is roughly what you'd expect from randomly constructed portfolios of about twenty levered micro-cap companies. Private equity managers trailed a simulated factor-matched benchmark by two to four percentage points annually, a gap consistent with the fee load.^⁵

If that analysis holds, dispersion in private markets is largely a mechanical consequence of concentration and leverage, not evidence of differentiated skill. An allocator who concentrates on that basis has not improved expected return. That allocator has increased variance around a mean already reduced by fees, while believing the opposite.

The two findings are not in conflict. They describe the same distribution from opposite ends. Dispersion creates the opportunity to capture something above the median, and it also punishes concentration undertaken without a basis.

How should an allocator decide how many managers to hold?

The useful reformulation is this. Concentration is rational only to the extent that a selection process produces information the market does not already price. Absent that, concentration is leverage applied to noise. With it, breadth is the more expensive error, because each additional position dilutes the very signal the process is meant to capture.

This means the number of managers in a portfolio is an output, not a policy. It should follow from an honest assessment of how many managers the program can underwrite to a standard that justifies holding them. Four diagnostics make that assessment concrete.

For each manager held, what specifically was underwritten beyond a track record and a reference set, and would that evidence hold up if the manager produced a poor next fund?
How many positions could the investment team re-underwrite from first principles in a single quarter? Positions beyond that count are held on inertia.
What is the program's pooled expected return relative to the median of its opportunity set, and does the difference justify the fee and governance cost of the structure?
Which positions exist because they were selected, and which exist because a slot needed to be filled?

When is breadth the right choice?

Diversification within an asset class is not the same decision as diversification across asset classes. Across asset classes, it reduces exposure to a single risk. Within a skewed, high-dispersion, high-fee asset class, it buys the median and charges private market pricing for it.

That can still be the correct choice. An allocator without a defensible selection process is better served by breadth than by concentration and should say so plainly rather than present the position as a conviction. What does not survive examination is holding thirty managers while describing the program as a manager selection.

Frequently Asked Questions

How many private funds should an advisory firm hold?+

Published work sits well below where most programs land. Addepar research points to three to six commitment-based positions per portfolio objective, and CBRE Investment Management found three to five sufficient in core real estate. Both conclude that additional positions dilute manager-specific return while adding operational burden.

Does diversification work the same way in private markets as in public equities?+

Not cleanly. In public equities, adding positions reduces idiosyncratic risk without reducing expected return. In private markets the distribution is skewed, so a small number of results drive the average and broad exposure converges toward the median rather than the mean.

What is over-diversification in a private markets program?+

It describes a portfolio wide enough that manager selection no longer changes the result. Breadth trims the left tail and the right tail together, and in a skewed distribution the right tail is where the allocation rationale lives. The expected outcome becomes the market median, less fees.

Is concentration in private markets a better strategy?+

Only conditionally. Verdad Advisers found that observed dispersion is consistent with randomly constructed portfolios of levered micro-caps, meaning concentration without a genuine selection edge increases variance around a mean already reduced by fees. With a defensible process, breadth becomes the more expensive error.

How can an allocator tell whether a program is buying the median?+

Standard reporting will not show it. Useful diagnostics include what was specifically underwritten for each manager beyond a track record, how many positions the team could re-underwrite from first principles in a quarter, and which positions exist because a slot needed filling.