A statistical read of $233 billion in CalSTRS and CalPERS private equity commitments — asset class performance, manager persistence, and the range of outcomes LPs actually experience.
Every year, California’s two giant public pension systems — CalSTRS (teachers) and CalPERS (state and local employees) — publish the fund-by-fund performance of their private equity portfolios. Because these are public records, they give us something rare in private equity: a large, transparent, apples-to-apples dataset spanning nearly 30 years and hundreds of managers. We pulled both reports, cleaned roughly 940 individual fund commitments, and ran the numbers. Here is what the data says about how this asset class actually behaves, not how it is marketed.
About the dataset
This analysis draws on two public disclosures: the CalSTRS Private Equity Portfolio Performance Report (as of June 30, 2025) and the CalPERS Compiled Returns — Private Equity report. Together they list 943 individual fund commitments — 477 from CalSTRS totaling $81.3 billion committed, and 466 from CalPERS totaling $152.3 billion committed — spanning vintage years 1998 through 2025. After normalizing repeat fund names into general partner families (for example, grouping “Blackstone Capital Partners VI, VII, and VIII” under a single GP relationship), the combined dataset represents 556 distinct GP families.
Both source reports were published as scanned, image-based PDF tables rather than structured data files, so every figure in this report was rebuilt programmatically: committed capital, contributed capital, distributed capital, remaining value, and IRR were parsed for each row and checked against the digit-grouping of the underlying dollar figures. A small share of records (under 4%) needed manual correction where a fund name wrapped awkwardly across a page break; those corrections were verified by hand against the original report text.
Because IRR is only a meaningful signal several years into a fund’s life, this report treats funds with a 2020-or-later vintage year as immature and excludes them from most return statistics — leaving 310 CalSTRS funds and 159 CalPERS funds (469 combined) in the “mature fund” analysis set referenced throughout. One further caveat: CalSTRS’s own report notes that its IRR methodology may differ from CalPERS’s, from the general partners’, and from industry norms — so cross-plan comparisons here are directionally informative rather than strictly apples-to-apples.
Executive summary: return by strategy
Before getting into vintage-year cycles, manager persistence, and fund size, here is the headline view: how each private equity strategy performed on average, and just as important, how wide the gap was between its best and worst fund. The table below covers every strategy category with at least five mature funds in the combined dataset.
The min/max and standard deviation columns are the real story here. Every strategy — even the steadier ones — contains at least one fund that lost money, and most contain at least one fund that returned 25% or more a year. Growth equity carries the widest standard deviation of any category (22.1 points) on just six funds, so its 19.2% mean should be read as directional rather than statistically robust — a single outstanding or disappointing fund moves the average a lot when the sample is that small. The same caution applies to healthcare/life sciences (five funds). Buyout/corporate PE, the largest category by far at 323 mature funds, is the one strategy where the sample size is large enough that its 14.0% mean, 11.0-point standard deviation, and -19.1%-to-91.3% range can be taken as a reasonably reliable picture of what the broader buyout market has actually delivered. Notice also that credit/special situations combines a below-average mean with the second-lowest standard deviation — the clearest evidence in this dataset that it behaves like a genuinely different risk profile, not just a lower-returning version of buyout.
Reading the fine print, briefly
A few terms recur throughout this piece. IRR (internal rate of return) is the annualized return a fund has generated so far, accounting for the timing of cash flows in and out. Vintage year is the year a fund started investing — think of it as the fund’s birth year, which matters enormously because funds born into a recession or a bubble inherit very different entry prices. TVPI (total value to paid-in capital) is a multiple: for every dollar an investor put in, how many dollars has the fund returned or is currently worth. And a fund is not considered mature enough to judge until roughly six to eight years after its first capital call — before that, an IRR is closer to a rumor than a result.
Asset class performance over time
Averaged across both pension systems, mature private equity funds (vintage 2019 and earlier) returned a mean IRR of 13.3% and a median of 12.7%, with meaningful year-to-year swings. The chart below tracks the average IRR of each vintage year’s crop of funds at both plans. Two patterns jump out immediately.
First, the two plans move together. Even though CalSTRS and CalPERS run independent portfolios with different managers and different commitment sizes, their vintage-year curves rhyme — both dipped through 2005–2007 and both climbed into the 2016–2018 stretch. That is a market effect, not a manager-selection effect: entry price and macro conditions at the time of investment explain a large share of what any given vintage will return, no matter who is running the fund.
Second, the 2005–2008 "financial crisis vintages" are visibly the weak spot in the whole 27-year dataset. Grouped into five-year buckets, funds born 2005–2009 averaged just 9.1% mean IRR versus 15.9% for funds born 2015–2019 — a six-point-plus gap driven almost entirely by when the fund started buying companies, not by manager skill. If there is one truism private equity investors relearn every cycle, it is this: the price you pay at the start of the ride matters more than how well you drive.
The range of outcomes is wide — and that is the point
Private equity is often sold on its average return, but the average hides a genuinely wide spread of individual outcomes. Among mature, sizeable commitments ($50 million or more), the best performers cleared 30–80%-plus annualized, while the worst lost money outright. The standard deviation across the combined mature dataset is roughly 10.7 percentage points around a 13.3% mean — in plain terms, a one-standard-deviation swing is the difference between a solid double-digit return and a fund that never got out of single digits, or worse.
Notice the pattern in the loss column: energy, clean-tech, and 2006–2008-vintage funds dominate the bottom of the list. That is not a coincidence — it is the same "entry-price" story from the section above, compounded by a sector (energy) that has had its own multi-decade cycle of boom and bust layered on top of the broader private equity cycle.
Does manager performance persist?
This is the question every allocator actually cares about: if a manager’s last fund was excellent, does the next one tend to be excellent too? We tracked every general partner family that raised three or more funds across both plans, ranked each fund against its own vintage-year peer group (top, second, third, or bottom quarter), and then looked at what happened from one fund to the next fund from the same manager.
The correlation between a manager’s current fund IRR and their next fund’s IRR is 0.47 (statistically significant at well beyond the 99.9% confidence level, with a sample of 230 fund pairs). That is a moderate, real relationship — not proof that yesterday’s winner is tomorrow’s lock, but clear evidence that performance is stickier in private equity than in most public asset classes, where a manager’s past return tells you almost nothing about their next one. This is precisely why access to a small set of proven managers is one of the more durable edges in this business, and why relationships with general partners — not just capital — are a genuine competitive asset for an allocator.
Where the capital concentrates: repeat GP franchises
Both pension systems have, over time, gravitated toward a relatively small set of managers they keep re-upping with. We identified 82 GP families that raised three or more funds across the combined dataset. Collectively, those repeat franchises absorbed $89.2 billion — 38% of all capital committed — even though they represent a small fraction of the 556 distinct manager relationships in the data.
There is also meaningful overlap between the two pension systems themselves: 58 of CalSTRS’s 264 distinct manager relationships — 22% — are managers that CalPERS has backed as well. Two of the largest, most sophisticated institutional allocators in the country, running fully independent due-diligence processes, are converging on a similar shortlist of managers. That convergence is itself a signal: in a market with tens of thousands of private equity firms, the pool that giant, patient capital actually trusts with repeat commitments is small.
The mega-fund paradox: what the academic research actually says
The pattern above raises an uncomfortable question. A wide body of academic and industry research has found, fairly consistently, that smaller private equity funds outperform larger ones — yet the largest institutions in the world, including the two in this report, keep re-upping into an ever-larger set of mega-franchises. Both things appear to be true at once. Here is the research, then our own dataset’s answer to whether the theory holds up in practice.
What the research shows
The most rigorous evidence comes from Lopez-de-Silanes, Phalippou, and Gottschalg, published in the Journal of Financial and Quantitative Analysis (2015), which examined 7,500 individual private equity investments over 40 years. They found scale to be one of the strongest predictors of returns at the deal level: the median gross IRR was 36% in the lowest-scale decile of investments versus just 16% in the highest-scale decile — a more than two-to-one gap the authors attribute to organizational diseconomies of scale, not manager skill. Cambridge Associates benchmark data has repeatedly shown a similar pattern at the fund level, with buyout funds under $1 billion generating higher median and top-quartile returns than mid- and mega-cap funds over the 2010–2020 period. McKinsey & Company has put a number on the gap directly, estimating that small- and mid-cap buyouts have outperformed large-cap buyouts by up to 4 percentage points of IRR. Research cited by Bernstein in 2025 similarly found that both IRR and total value multiples have run higher for smaller funds in aggregate, with the gap most pronounced among top-quartile managers. And PitchBook data shows that roughly 18% of first-time funds clear a 25% IRR, compared with about 12% of funds from more established, serially-raised franchises.
The proposed mechanisms are intuitive rather than exotic: smaller funds write smaller checks into less competitive, less picked-over deals; they deploy capital faster instead of straining to put ever-larger sums to work; and their economics lean more heavily on carried interest than on management fees, which sharpens the alignment between the manager and the investor. Institutional Investor has reported on research from the consultant StepStone showing the same pattern from a different angle: managers who grow their fund size by more than double from one vintage to the next see returns fall by an average of 5.8 percentage points, while managers who shrink their fund size see returns rise by 4.4 points — a finding public pension systems like the Texas Municipal Retirement System have begun calling “fund size discipline” in their own manager selection process.
Testing the paradox against our own data
So does the theory hold up here? We isolated the 15 GP families that raised the most cumulative capital across CalSTRS and CalPERS — essentially the mega-franchises institutions keep re-upping with, including Blackstone Capital, TPG, Apollo Investment, Thoma Bravo, Carlyle, CVC Capital, and Hellman & Friedman — and compared their fund-level returns against everyone else in the same buyout strategies.
Two things are true simultaneously, and they explain the paradox rather than resolve it. First, brand does carry a modest, real edge: mega-franchise names outperformed the rest of the buyout universe by roughly two points of mean IRR in this dataset, which lines up with the manager-persistence finding earlier in this report. Second, size itself is still a drag: within that same buyout universe, funds raised at $1 billion or larger averaged nearly 4 points lower IRR than funds under $500 million — consistent with Lopez-de-Silanes, Phalippou, and Gottschalg’s deal-level finding almost exactly. The mega-franchises are not winning because they are large; they are winning, to the modest extent that they do, in spite of being large — on the strength of manager quality and deal access that partly offsets a real scale penalty.
Now to the direct question: does consistent institutional support actually put these franchises at the top of the list, consistently? Not really — ranking every mature fund against its own vintage-year peer group, mega-franchise funds landed in the top performance quartile 32.5% of the time — better than the 24.0% rate for everyone else, but far from a lock. And 40% of the time, a mega-franchise fund still landed in the bottom half of its peer group. The flagship vehicles make this concrete: Blackstone Capital Partners V (2006, a $1.6 billion CalSTRS commitment alone, one of the largest single commitments in either dataset) and TPG Partners V (2006, a combined $1.6 billion commitment across both plans) — the largest funds raised by two of the most heavily re-upped franchises in this entire dataset — both landed in the bottom half of their 2006 vintage-year peer group. Apollo Investment Fund VIII (2013) and Onex Partners IV (2014) did the same. The brand name kept the capital coming; the flagship-sized vehicle itself did not consistently deliver top-of-the-list results.
This is not an argument that mega-fund investing is irrational — institutions have real reasons beyond raw IRR to concentrate with a known set of managers, including capacity to deploy at scale, co-investment access, governance rights, and lower manager-selection risk. But the data is a useful check on the assumption that consistent commitment automatically buys consistent outperformance. It mostly buys consistency of relationship, not consistency of results.
Strategy matters: dispersion by asset class within private equity
"Private equity" is not one thing — buyout, venture, growth, credit, and energy strategies behave differently, both in average return and in how much that return varies fund to fund.
Buyout and corporate private equity — the largest single category by capital deployed — delivered a 14.0% mean IRR with a standard deviation around 11 points, a reasonably efficient risk-return trade. Credit and special situations strategies produced a lower but far steadier 11.6% mean with roughly half the volatility (a 6.1-point standard deviation), which is exactly what you would expect from a strategy built around contractual cash flows rather than equity upside. Venture capital, unsurprisingly, showed the widest dispersion of any category relative to its size — a reminder that venture is a "power law" business, where a handful of enormous winners carry the average and most individual bets fall well short of it. Energy and infrastructure funds were the one category with a below-average mean (7.1%), consistent with the commodity-cycle-driven losses we saw in the bottom-performer table above.
Bigger is not always better: fund size and returns
One popular assumption is that larger, more established funds are the "safer" choice. The data pushes back on that a little. Across mature funds with a reported commitment size, the correlation between commitment size and IRR is slightly negative (-0.10), and funds under $250 million in commitment size averaged better median returns (12.4–12.7%) than funds of $500 million or more (11.3–11.5%). This does not mean small funds are safer — the smallest funds also carry plenty of the losses we saw above — but it does undercut the idea that writing a bigger check to a larger, more institutional vehicle reliably buys a better outcome. Manager selection and vintage timing appear to matter more than fund size on their own.
What should retail investors expect from the new wave of PE access products?
Everything above describes institutional-share-class private equity: pension funds writing nine- and ten-figure checks directly into a limited partnership, with no sales intermediary and management fees negotiated at scale. Over the past few years, a very different access point has opened up for individual investors — interval funds, non-traded business development companies (BDCs), and tender-offer funds sold through broker-dealers and wealth platforms, often with a minimum investment as low as $2,500–$25,000. These products are not a free lunch version of what CalSTRS and CalPERS get; they carry a materially different, and materially higher, fee stack.
What the fee stack actually looks like
Retail-channel private equity and private credit vehicles typically layer three types of cost on top of one another. First, an upfront sales load: broker-sold share classes commonly carry charges of 3.5% up to 5.75% of the amount invested, paid to the selling broker-dealer, plus an ongoing annual servicing/distribution fee (often 0.25–0.85% of assets) for as long as the shares are held — the "trailing" or tail fee that keeps paying the distribution channel year after year. Second, an asset-based management fee, typically 1.25–2% of gross (not net) assets annually — and because many of these vehicles use 30–45% leverage, the fee is charged on a larger asset base than the investor's actual equity. Third, an incentive or performance fee, commonly 12.5–20% of profits above a 5–7% hurdle rate. Morningstar has specifically flagged that these incentive fees, though structured to look conditional, are in practice collected almost regardless of manager skill once a fund is modestly leveraged, because the hurdle is calculated on borrowed money's low cost of capital rather than on the fund's true equity return. Stacked together, industry data shows average all-in fees of roughly 5.15% of net asset value annually for non-traded BDCs generally, versus about 3.3% for the newer "perpetual" structures — both multiples of the 1–2% all-in cost of a typical institutional share class, before any sales load is even considered.
The platform layer: iCapital and CAIS
A separate cost sits between the retail investor and the fund manager: the technology platforms that RIAs, private banks, and broker-dealers use to package institutional-style private funds into feeder vehicles for individual clients. The two dominant players are iCapital and CAIS, and neither typically charges the end investor a visible, separate line-item fee — instead, the platform’s cut is embedded inside the feeder fund’s own expense structure, which is part of what makes the true, all-in cost difficult for an investor to see without asking directly. Envestnet’s PMC research group has described the standard iCapital-distributed structure as a 40-to-50-basis-point platform access fee layered on top of the underlying fund’s own economics, before the advisor’s separate fee is even added — a combined stack that one 2026 industry review pegged at north of 3% a year in aggregate, before any performance-based carry is deducted. CAIS has run a similar embedded-fee model, historically charging as much as 20 basis points on the custom feeder funds it builds for advisors; in 2024, CAIS said it would cut that fee to as low as 5 basis points on new feeder funds, with its chief executive writing in an internal memo that the total cost of alternatives to end investors, across asset managers, custodians, reporting firms, platforms, and advisors, "must be considered if we are to truly democratize alternative investments," and estimating the fee cut could save end investors hundreds of millions of dollars a year in aggregate. That both platforms are compressing fees under competitive and regulatory pressure is a genuinely good sign for retail investors. But the starting point still matters: an extra 20–50 basis points of platform fee, compounded annually over a typical 7-to-10-year hold, is one more layer between the gross return a fund manager reports and the net return an individual investor actually receives.
A real-world illustration
This is not a hypothetical. The Ares Private Markets Fund (Class A) — a real, currently-offered retail interval fund — generated a 12.44% net return in 2025 after its 5.14% annual expense ratio was already deducted. Layer in the share class’s sales charge of up to 3.5%, and the return an actual investor realized dropped to 8.50%. That is a nearly 4-point-plus haircut from fees and load alone, on a fund that was already performing well. Broader data points in the same direction: one recent industry analysis found that U.S. private equity funds, across the market and including fees, generated 5.8% annualized returns between 2022 and September 2025, versus 11.6% annualized for the S&P 500 over the identical period — a period in which the same underlying illiquidity and complexity that public pensions have long been compensated for accessing did not, net of the modern fee stack, translate into better outcomes for the investors paying it.
Does the fee load erode the edge even for top performers?
Based on the evidence available, the honest answer is: largely, yes. Consider the mega-franchise buyout funds in our own institutional dataset, which averaged a real 15.4% mean IRR net of standard institutional fees (no sales load, roughly 1.5–2% management fee, 20% carry above an 8% hurdle, all negotiated at scale). Layer a retail-channel fee stack on top of even that above-average result — a 3.5–5.75% one-time load, a 3–5% all-in annual expense ratio instead of the roughly 2% institutional rate, and an incentive fee calculated on a leveraged base — and several points of annual return are consumed before the investor sees a dollar. A gross return in the mid-to-high teens can plausibly land in the high single digits to low double digits net of the full retail fee stack, which is close to, or in some periods below, what a simple, nearly-free public index fund would have delivered over the same holding period, without the multi-year lockups, quarterly-only liquidity, and valuation-lag risk that come with the private structure. Retail investors evaluating these products should ask for the all-in expense ratio (not just the headline management fee), the exact sales load and share class, whether the incentive fee hurdle is calculated on levered or unlevered returns, and how the product’s net, fee-inclusive return has actually compared to a comparable public market index — not just to a gross private markets benchmark that no investor ever actually receives.
What this means for how we think about private equity
- Timing is a real, measurable risk factor. Vintage year alone explains several points of expected return — committing steadily across cycles, rather than concentrating in any single year, is a straightforward way to manage that risk.
- Manager persistence is real, but not deterministic. A 0.47 correlation and a near-doubling of top-quartile repeat rates is meaningful evidence in favor of backing proven teams — but a strong past fund is a tilt of the odds, not a guarantee.
- Strategy-level diversification changes your risk profile, not just your return. Credit and special situations strategies offer a genuinely different volatility profile than buyout or venture, worth weighing against an investor’s actual tolerance for a bad year.
- Access to a concentrated set of trusted managers is a competitive asset. The overlap between CalSTRS and CalPERS, and the capital concentration among repeat franchises, both point to the same conclusion: relationships and reputation, earned over multiple fund cycles, are hard to manufacture and hard to buy.
None of this replaces the work of diligencing an individual manager, an individual strategy, or an individual moment in the cycle — but it does give a data-grounded starting point for the conversation. As Ben Franklin might have put it if he ran a pension fund: an ounce of vintage-year diversification is worth a pound of after-the-fact regret.
What patterns are you seeing in your own portfolio or manager lineup that either confirm or challenge this data? And if you have access to a proprietary track record you would like benchmarked against this public dataset, we would enjoy comparing notes — send us a note with your questions or your own read of the numbers.
