PE/VC performance is one of the most-debated empirical topics in finance. The headline numbers are flattering; the honest numbers, after fees and benchmarking against a public alternative, tell a more nuanced story.
The five metrics every LP knows
IRR annualises returns and is the industry's default metric — but it is sensitive to deal pacing and can be inflated by subscription-line financing that delays the first capital call. TVPI (total value to paid-in) is the cleaner total-return multiple. DPI (distributions to paid-in) measures realised cash returns and is the LP's true scorecard once a fund matures. RVPI (residual value) is the unrealised mark, contestable and increasingly so since 2022. Net versions of each subtract fees and carry — that is the number LPs actually earn.
A useful sanity check: a fund with a 25% gross IRR and a 17% net IRR is paying an 8-point spread to the GP. A fund with 25% gross and 12% net is paying a 13-point spread, which is a lot.
PME: the public-market yardstick
The Kaplan-Schoar Public Market Equivalent (PME) test asks the most important question in the asset class: did the LP do better in this fund than in a public-equity index over the same period? A PME of 1.0 means the fund tied; above 1.0 means it added value; below 1.0 means it didn't. The empirical result, across many studies, is that the median buyout fund has a PME of ~1.0–1.1 — barely positive — while top-quartile funds have a PME of 1.3+ and bottom-quartile funds destroy value relative to the index. Manager selection matters more than asset-class allocation.
Dispersion and persistence
PE/VC returns show the widest dispersion of any institutional asset class — top decile to bottom decile in venture often spans 30+ IRR points. There is also persistence: top-quartile managers in one vintage are more likely than chance to be top-quartile in the next, a result first documented by Kaplan-Schoar and confirmed in subsequent updates. This is why LP manager-selection effort is concentrated at the top of the distribution.
Biases in the published data
Reported PE/VC benchmarks suffer from several construction biases worth naming. Survivorship bias: failed funds may not report. Selection bias: voluntary reporting tilts toward better outcomes. Mark-to-model bias: unrealised values are GP-determined and tend to be smoothed. Vintage timing bias: fund-year cohorts have different underlying market exposures. Honest LP analysis adjusts for all four.