Volatility laundering is the understatement of private‑market risk that stems from infrequent, appraisal‑based pricing. Because private assets are marked quarterly against the prior quarter’s mark, their reported returns are serially correlated and artificially smoothed — and a variance‑based model interprets smoothness as low risk. Unsmoothing the data roughly doubles measured buyout volatility and nearly triples it for real estate.
The efficient frontier has governed institutional allocation for seven decades, and for most of that time the debate has focused on its shape: where the curve bends, how far alternatives push it up and to the left, and which mix sits at the tangency point. These remain useful questions. They are also the wrong place to focus attention now, because the frontier has always been only as honest as the risk figures fed into it, and the part of a modern institutional portfolio where those figures are least honest is precisely the part that has grown to its center.
Properly understood, the history of portfolio theory is not a story about better curves. It is an ongoing debate about how to measure risk without being misled. Each major advance revised the measurement, conceded that the prior input was shakier than the geometry implied, and shifted the unsolved part of the problem elsewhere. The current chapter is no different, and an allocation committee that understands the through‑line will read today’s frontier with the skepticism it deserves.
What is the efficient frontier, and what did Markowitz assume?
The original bargain
The efficient frontier is the set of portfolios offering the highest expected return for each level of variance. Harry Markowitz formalized the terms in 1952. His paper “Portfolio Selection,” expanded into a book in 1959 and recognized with the Nobel Memorial Prize in 1990, made two moves that still structure the discipline. It defined the risk of a holding not by the holding in isolation but by its covariance with everything else in the portfolio, and it defined risk itself as variance. From those two definitions, the efficient frontier follows as a matter of arithmetic: for any level of variance, one combination of assets carries the greatest expected return.
The 1960s extended this logic outward. Building directly on Markowitz, Tobin introduced the separation theorem, and Treynor, Sharpe, Lintner, and Mossin independently arrived at the Capital Asset Pricing Model, which reduced an asset’s risk to a single term: its beta relative to the market. The notions of alpha and beta that allocators use every day descend from this lineage. What every step shared was the founding assumption, inherited without much re‑examination, that variance and covariance were adequate proxies for risk, and that the numbers plugged into them were real.
Why does mean‑variance optimization break down?
The corrections
Those inputs were not always reliable, and the field spent the following decades saying so. Mean‑variance optimization carries a well‑documented flaw that its elegance conceals: it behaves as an error‑maximizer. Small mistakes in the estimated returns and covariances that feed the model produce portfolios that are wildly concentrated and confidently wrong, because the optimizer loads onto whatever the estimation error happens to favor. The dependable‑looking curve rests on inputs that the available data windows are usually too short to pin down.
Behavioral finance dismantled a second assumption: that the investors on the other side of every price are rational. Factor research dismantled a third: that a single-market beta captures systematic risk, replacing it with a richer set of exposures. None of this overturned Markowitz; it refined the honesty of the inputs, which is a more durable kind of progress. Each correction was the discipline conceding that its risk measure had been too clean.
What is volatility laundering, and how does it distort the efficient frontier?
The measurement problem the frontier was never built for
As private markets grew from a satellite holding into a core allocation, they imported a measurement problem the frontier had no equipment to handle. Private assets are not marked continuously; they are appraised quarterly, and each appraisal leans on the prior quarter’s mark. The reported return series that results is serially correlated and smooth, and smoothness, within a variance‑based model, is interpreted as low risk.
Cliff Asness of AQR gave the effect its memorable name: volatility laundering. The mechanism is not a moral failing of any manager; it is the arithmetic consequence of pricing something infrequently. A stale mark carried forward makes each new observed return partly an echo of the last, which suppresses measured volatility and dampens measured correlation to public markets.
The risk did not shrink. The camera simply stopped taking its picture.
The magnitude of the distortion is nontrivial and nonuniform. Researchers have unsmoothed private‑market return series by stripping out the serial correlation, and the correction is uneven across strategies. In one 2024 study reported by Morningstar, adjusting for lagged betas roughly doubled the measured volatility of small and large buyout funds and of late‑stage venture, tripled it for real estate and secondaries, and moved early‑stage venture from 29 percent to 87 percent. Private credit barely moved at all.
Reported vs. unsmoothed volatility by strategy
PIMCO’s own work puts the true economic volatility of private equity near 30 percent against a headline figure closer to 10 percent and estimates that smoothing lowers the apparent probability of a severe drawdown to almost nothing against a real probability closer to one in six over three years.
Fed smoothed inputs, the efficient frontier draws a curve that flatters private assets and quietly steers the optimizer toward them. That is the trap beneath most published versions of the private‑markets efficient‑frontier argument, and a committee that has read the unsmoothing literature discounts those curves on sight.
What is the Total Portfolio Approach — and does it fix the measurement problem?
The institutional response, and its limit
The frontier’s heirs understand this, and the most consequential governance shift in the institutional world is a direct response to it. In November 2025, the board of the California Public Employees’ Retirement System voted to abandon strategic asset allocation, the framework of fixed per‑asset‑class targets that has organized institutional portfolios since Markowitz, in favor of the Total Portfolio Approach. It is the first United States public pension fund to do so, across a fund of roughly $626 billion, with the new model live as of July 2026 and performance judged against a single 75/25 reference portfolio. Canada’s CPP Investments, Singapore’s GIC, and Australia’s Future Fund reached comparable conclusions earlier.
The Total Portfolio Approach asks a better question. Rather than filling an asset‑class bucket to its target, it weights each investment by its marginal contribution to the whole portfolio’s risk and liquidity budget. But it does not resolve the measurement problem underlying the frontier; it relocates it. As one skeptical account in Institutional Investor put it, the approach is the same human framework, just more sophisticated. KKR, an advocate, is candid that it is no substitute for skill, and that outcomes still turn on the quality of forecasting, risk modeling, manager selection, and implementation. The unsolved part of the problem has simply moved from “what is our target allocation” to “whose marks, whose risk model and whose managers do we trust to fill it.”
What should an allocation committee do about smoothed marks?
What survives
The history of portfolio theory pays off here with something usable. When the data are unsmoothed, the naive case for private markets, the one that sells lower volatility and easy diversification, weakens considerably. The dispersion case grows stronger. The gap between an 8‑percent input and an 87‑percent input is not a gap between public and private as categories; it is a gap between strategies, and beneath that, between managers running the same strategy. McKinsey’s 2026 assessment of a maturing asset class reaches the same conclusion from the opposite direction: returns will increasingly be shaped by deliberate choices about sourcing, pricing discipline, and operational value rather than by exposure to the asset class itself. Alpha, in their phrasing, will increasingly be made rather than found.
The frontier, in other words, still shifts. The honest version of that shift is earned by the structures and managers a committee holds, not by the mere fact of holding private assets.
That reframes the question worth asking in this piece. The sharper question for an allocation committee to ask is not what its target private‑markets allocation should be. It is whether a given holding would still sit where the model places it on the frontier if it were marked honestly, and which managers would survive that test. It also suggests one signal to weight differently going forward.
Serial correlation in a manager’s reported returns is not evidence of stability. It is evidence of stale pricing, and it belongs in the risk column, not the comfort column.