Manager persistence is actively contested as of 2026. Harris, Jenkinson, Kaplan and Stucke find little or no persistence for post-2000 buyout funds when performance is measured on the information available at the next fundraise, while BCG reports it strengthening. Even the most favorable published estimates imply most upper-quartile managers do not repeat, so diligence should test the mechanism rather than the rank.
Almost every re-up decision in private markets rests on a single assumption: that a manager who performed well in the last fund is more likely than not to perform well in the next. The assumption is so deeply embedded that it is rarely stated, let alone tested. As of 2026, it is also actively contested — with credible institutions publishing findings that point in opposite directions.
What does the research say about manager persistence?
The original finding was strong. Work on pre-2000 data found meaningful persistence in both buyout and venture, with the probability of repeating top-tercile performance in successive funds well above what random transitions would produce.
The revision is more important than the original. Harris, Jenkinson, Kaplan, and Stucke, using Burgiss cash-flow data, draw a distinction that changes the conclusion. Measured by final realized performance, the data show persistence. Measured by the information an investor would have had at the time of the next fundraise — the only information a re-up decision can use — they find little or no evidence of persistence for post-2000 buyout funds. The same work documents the upper-quartile-to-upper-quartile transition probability falling from roughly 0.42 in the 1980 to 2001 vintages to roughly 0.26 in the post-2000 cohort and finds that skipping a vintage does not improve the odds.¹
Earlier estimates point the same way. Harris and co-authors found that nineteen percent of post-2001 buyout funds that followed an upper-quartile predecessor repeated that standing. McKinsey's 2017 analysis of vintages from 1995 to 2013 put the figure at twelve percent and described a steady decline.²
What did BCG and KKR each find in 2026?
By contrast, BCG published an analysis in 2026 arguing that persistence is strengthening rather than weakening. For the 2007 to 2012 vintages, BCG reports that a manager with an upper-quartile fund had roughly a one-in-three chance of repeating and roughly a sixty percent chance of remaining in the upper half. For the 2013 to 2018 vintages, those figures rise to approximately forty-five percent and eighty percent. The explanation offered is mechanical rather than hopeful: as multiple expansion receded as a return driver, operating capability became more visible in the results, and operating capability travels across a franchise.³
KKR, writing in the same year, reports the opposite direction — nearly forty percent of above-median funds were followed by below-median performance in the subsequent vintage, and Preqin data indicate roughly thirty percent of managers sustaining upper-quartile standing by their third fund.⁴
The disagreement is not readily resolvable from outside the datasets, and three methodological choices explain most of it. Whether performance is measured as finally realized or as known at the time of the fundraise. Whether the unit of analysis is the sequential fund or the franchise across deals. And whether the sample period includes a regime in which multiple expansion supplied most returns, which compresses the apparent contribution of skill.
What holds true under either interpretation?
The useful observation is that the two camps agree on more than the headline suggests.
- Across all published estimates, many upper-quartile managers do not repeat. BCG's figures are the most favorable in the literature and still imply that more than half do not repeat.
- Where persistence is observed, it is attributed to a mechanism — operating capability, sourcing advantage, or franchise-level process — rather than to the prior fund's return itself. No study argues that the number causes it.
- The gap between measuring final performance and measuring information available at decision time is large enough to reverse the conclusion. That gap is a property of the decision, not of the data.
Taken together, those three points retire the quartile label as a diligence input. The label is not wrong. Rather, it carries almost no incremental information when used, and whatever information it does carry is a proxy for something an allocator could examine directly.
How do you test the mechanism behind a manager's record?
If the mechanism, rather than the number, is what persists, diligence should focus on identifying the mechanism and testing whether it remains intact.
- Return attribution for the prior fund. How much came from multiple expansion, how much from leverage, and how much from revenue and margin changes within the businesses? Only the third shows a repeatable capability.
- Continuity of the people who produced the record at the deal level rather than the masthead level. Franchise persistence depends on the franchise remaining the same.
- Whether the strategy being raised is the one that generated the record. A materially larger fund is a different strategy, regardless of the cover label.
- What the manager declined and why. Deal selection is easier to evaluate through rejected opportunities than through completed ones, and rejection discipline is where the current market's competitive pressure shows up first.
What does this mean for re-up decisions?
The persistence assumption is the core belief underpinning private-market allocation, and 2026 is the first year in which serious institutions have publicly disagreed about whether it holds. That disagreement is itself informative. A relationship that can only be justified by a prior quartile standing is being justified by the weakest available evidence, even under the most generous reading of it.
The allocators best positioned for the next several vintages will be those who can, specifically and in advance, state what they believe a manager does to produce return and what observation would lead them to conclude they were wrong.
For informational purposes only. Not investment advice, and not an offer to sell or a solicitation of an offer to buy any security.