Articles

Can you average fund IRRs to get a portfolio return?

Averaging a set of rates produces a number that describes no portfolio anyone owns.

No. Three funds returning 11.52, 10.12 and 19.51 per cent average to 13.72 per cent, but the money in the programme earned 11.78 per cent — 194 basis points less. The internal rate of return is money-weighted: it depends on how much capital was outstanding and for how long. The only correct aggregate is the pooled figure, obtained by combining every fund’s cash flows into one series, adding every residual value, and solving once.

Why the rate cannot be averaged

An institution holds three funds of the same vintage and strategy, with commitments of 200, 80 and 20 million. Their since-inception internal rates of return are 11.52, 10.12 and 19.51 per cent. The simple average of those three rates is 13.72 per cent. The pooled return — the rate the money actually earned — is 11.78 per cent.

The internal rate of return is a money-weighted measure. It depends on how much money was outstanding and for how long. Averaging the rates of two funds treats them as equal contributors regardless of size, which they are not. Nothing in the arithmetic of an average knows that one fund held ten times the capital of another.

Pooling is the correct method. Combine all the cash flows of all the funds into a single series, add all the residual values, and compute one internal rate of return on the combined object. That figure answers the question an institution actually has: what did the money in this programme earn?

MethodResult
Pooled — all cash flows combined11.78%
Simple average of the three rates13.72%
Average weighted by capital called11.68%
Three ways of aggregating the same three funds.

Where the 194 basis points come from

Entirely from fund C. It called 18 million, which is 6 per cent of the programme’s paid-in capital, and it returned 19.51 per cent. In a simple average it carries a third of the weight. The excellence is real and it is trivial in size, and an average cannot tell the difference.

FundCapital calledDistributionsResidual valueIRRTVPI
A180.0280.060.011.52%1.889x
B72.0104.022.010.12%1.750x
C18.045.05.019.51%2.778x
Aggregate270.0429.087.011.78%1.911x
Millions. The aggregate row is pooled, not averaged.

The capital-weighted average, at 11.68 per cent, lands below the pooled figure rather than above it. Weighting by capital called is closer than weighting by nothing, but it weights by capital and not by capital-time, so its error has no fixed sign and it widens as the funds’ durations diverge. An approximation that is sometimes high and sometimes low is not a shortcut worth taking when the exact calculation is one series and one solve.

The multiple has no such problem

Total value across the three funds is 516.0 million against paid-in capital of 270.0 million, so the aggregate TVPI is 1.911x and the aggregate DPI is 1.589x. Those are simple additions. There is no weighting question, no pooling subtlety and no averaging error, which is a genuine reason to put a programme-level multiple beside the programme-level rate. Where the pooled rate and the aggregate multiple tell different stories, the difference is duration, and duration is worth understanding.

Pooling has one honest limitation. A pooled return is dominated by the largest commitments and by the vintages where most capital was deployed. A programme returning 11.78 per cent may contain a strategy that failed and a strategy that did very well, and the pooled figure shows neither. So report it as the headline and decompose it three ways: by vintage year, which separates poor selection from expensive deployment; by strategy, which is the decomposition that supports allocation decisions; and by manager, which is the figure that belongs in front of anyone deciding whether to commit again.

What to report

Three details make pooling defensible. Convert every flow at the rate on its own date before combining, because a portfolio spanning currencies cannot be pooled any other way. Align the flows on the calendar rather than on each fund’s own timeline — fund A’s third year and fund B’s first year fall on the same dates and belong in the same period. And include every residual value, at the same measurement date, net of accrued carried interest.

Then fix the denominator. A programme with 500 of commitments, 300 called, 220 of net asset value and 200 undrawn has a total exposure of 420. An allocation framework measured against net asset value alone understates the programme’s claim on the portfolio by nearly half, and that is what produces the surprise when calls arrive faster than expected. Report exposure, not net asset value.

Never average internal rates of return. Pooling takes the same amount of work, and the difference on a real portfolio is frequently larger than 194 basis points.

The workbooks behind this article

Every figure above is a live formula in the free companion files for Private Markets Performance. Each workbook ends with a Checks sheet setting the printed figure beside the computed one. No account and no email address.

Open the companion files →

Also on this site