Practice chapter · understanding valuation
By Niko Hatziiosifidis · last reviewed 23 Aug 2026
The NAV of a semi-liquid fund is not a market price but a model-based estimate of fair value — in the language of accounting, almost always “Level 3”: valued using assumptions that no market directly confirms. That estimate is governed by rules (IPEV, IFRS 13, AIFMD), but it reacts to market movements with a lag and in muted form — which is why the smooth NAV curve shows less variation than is economically taking place. If you want to read NAV figures professionally, you always read them together with the more honest signals: secondary market prices, realised sales against the last carrying value, and the behaviour of the redemption queues.
With an ETF the exchange supplies the price — with a semi-liquid fund there is no exchange. The NAV is an estimate, not a market price: the hypothetical sale price (“fair value”) a buyer would pay for the holdings today. So that this estimate is not a matter of taste, virtually the entire industry values according to a common set of rules, the IPEV Guidelines (last fundamentally updated in December 2025). Which method is used depends on the asset class — the two most important approaches are multiples and DCF.
A company is priced using valuation multiples, usually EV/EBITDA: what are investors currently paying for comparable firms? The yardsticks are listed peer companies (the “liquid market”) and recent sales of comparable firms — the latest exits and takeovers among peers. The peer multiple is never applied mechanically but adjusted for size, growth and margin; EBITDA is adjusted for one-off effects. If valuations on the stock market rise or fall, that has to feed through the multiples into the fund NAV as well.
A wind farm or a fibre network has what a mid-sized company lacks: cash flows that are predictable for decades and often fixed by contract (concessions, power purchase agreements, regulated tariffs). That is why the discounted cash flow method dominates here: all future payments are estimated and discounted back to today at a risk-adjusted rate — roughly 5–7% for regulated networks, 7–9% for contracted assets, 10–13% for projects with market or construction risk. The flip side: interest-rate sensitivity. As a rule of thumb, a discount rate one per centage point higher costs around 10% of the value.
A direct loan is valued through its interest rate relative to the market: at purchase, the risk premium (spread) contained in the coupon is recorded; at each valuation date it is checked whether market rates, the spreads on comparable loans or the borrower’s credit quality have changed — technically another DCF. A loan that is being serviced properly therefore usually sits close to its face value; only when payments run into trouble is it written down, through a higher discount rate and the value of the collateral.
This is where valuation is most heavily formalised: external, independent experts determine the market value, mostly using the income capitalisation approach, and internationally in line with the RICS “Red Book”. German open-ended real estate funds must have their properties revalued at least quarterly; from a property value of €50m upwards, two independent valuers are mandatory, with a prescribed rotation of appraisers.
How well do these estimates match reality? There are three honest yardsticks:
For many years private equity holdings were on average sold above their last carrying value (2010–2021 roughly a +29% “exit premium”) — in the bull market the marks were on the conservative side. Since 2022 that has reversed: most recently sale prices have been below the last NAV on average. Conservative valuation is not a law of nature.
Sell fund units early on the secondary market and you do not get the NAV, you get what buyers will pay: most recently around 92–94% of NAV for buyout, ~91% for private credit, but only ~75% for venture and ~70% for real estate. That discount is the market’s view of how reliable the reported NAV is.
In 2022 the Nasdaq lost 33% — buyout funds reported only single-digit losses. Part of that is a genuine difference (different companies, less tech concentration), part is valuation inertia: models follow the market with a lag. That makes the NAV line look smoother than the economics of the investment actually are.
The accounting rules (IFRS 13, and its US counterpart ASC 820) sort every fair value measurement according to how much of it the market supplies and how much comes from a model. Level 1 is the ideal case: an identical instrument quoted on an active exchange — the price is the value. Level 2 means: no direct exchange quote, but observable market data from which the value can be derived almost mechanically. Level 3, finally, means: material valuation assumptions are not observable in the market — the value comes out of a model plus professional judgement. Practically everything that sits inside semi-liquid private markets funds — company stakes, direct loans, wind farms, office buildings — is Level 3. That is neither a blemish nor a seal of quality but a measurement instruction: Level 3 means that two honest, competent valuers can arrive at different numbers for the same asset, both within a defensible range. That is precisely why the IPEV guidelines require calibration to the last real transaction: the model is calibrated against the entry price actually paid. You will find the levels in black and white, incidentally: annual reports must disclose a fair value hierarchy table (IFRS 13, para. 93 et seq.). For you as a reader, Level 3 means above all this: a NAV is a point estimate drawn from a range — a deviation of a few per cent from the price later realised is not manipulation but the normal measurement uncertainty of this asset class.
One property follows from the valuation process that you have to know before you compare NAV curves: smoothing. It arises without any ill intent, out of three mechanisms. First, the frequency: a full valuation is usually done quarterly, with the value rolled forward in between. Second, the lag: by the time new company figures and market multiples reach the model, the quarter is over. Third, caution: valuers demonstrably anchor on the last established value — documented in property research since Geltner (1991) as “appraisal smoothing”, and modelled for fund returns by Getmansky, Lo and Makarov (2004): the reported return is a moving average of the true return. The measurable consequence: volatility too low, correlation with equities too low — and Sharpe ratios calculated from them too high.
2022 supplied the object lesson: the Nasdaq lost 33%, the S&P 500 around 19% — buyout funds, on MSCI Burgiss data, stood practically at zero over the year (MSCI, 2023). In property the same picture: the unlisted Blackstone fund BREIT reported around +8% for January to November 2022, while the listed MSCI US REIT index lost around −25% over the same period (Morningstar). Part of that difference is real — a different portfolio, active value creation. A substantial part is a measurement artefact. That is exactly what set off the “volatility laundering” debate: since January 2023 Cliff Asness (AQR) has accused the industry of now marketing infrequent valuation as an advantage — investors, he argues, are buying measured calm, not real calm (Asness, “Volatility Laundering”). The other side counters that long-term investors should not be trading on interim marks in any case, and that smoothing prevents panic selling. Both positions can be defended; what is not in dispute is the measurement finding: the NAV volatility of a semi-liquid fund is not a valid risk metric.
A second point belongs with that: fees hang on the NAV. The management fee is calculated on the carrying value — and many evergreen funds also charge the performance fee on unrealised NAV gains, not only on holdings actually sold. That is legal and disclosed in the prospectus, but it is a structural conflict of interest: the manager earns on a valuation that its own organisation substantially prepares. The important mitigations are the high-water mark, the hurdle rate and clawback — we show these mechanics on every fund page.
If the NAV is an estimate — where do you find harder information? In three places. First, in the secondary market: that is where fund stakes change hands for real money. In 2025 buyers paid an average of 87% of NAV for LP stakes — buyout 92%, private credit 91%, venture/growth 78%, property 70% (Jefferies, FY 2025). What is most revealing is the movement: in the stress year 2022 the average fell to 81%, venture to 68%. The discount is not “the true value” — it also contains the buyer’s liquidity premium — but its change is a market verdict on how robust the reported NAVs currently are (more in the chapter Understanding secondaries). Second, in realised transactions: every exit is a backtest of the last valuation. The research also shows that the marks are not always equally conservative: Brown, Gredil and Kaplan (Journal of Financial Economics, 2019) find that weaker managers in particular tend to set their NAVs too high during fundraising phases — and are punished for it by the market, because investors see through it. Third, in the behaviour of liquidity itself: gates that have been imposed and growing redemption queues are the market’s verdict made visible. The object lesson is BREIT: from November 2022, redemption requests exceeded the limits for 15 months while the reported NAV went on showing gains — the queue was the more honest signal, not the curve (see the chapter on gating).
Behind “the NAV” stands a chain of roles. The valuation of the individual assets is prepared by the manager’s valuation team — functionally separated from portfolio management under Article 19 AIFMD. The industry benchmark is the IPEV guidelines (December 2025 version, to be applied for reporting periods from Q2 2026): fair value as an exit price, calibration at every valuation date — and the express note that for evergreen structures with continuous dealing in units, a merely quarterly valuation cannot be sufficient (IPEV, 2025). Many managers additionally have the marks reviewed by external valuation specialists; an external valuer appointed under the AIFMD must be professionally recognised and is liable to the manager for negligent errors. The fund administrator works the NAV per unit out of all this, the depositary monitors the processes, the auditor tests the valuations in the annual accounts, and the supervisor tests the organisation. For ELTIFs, the technical standards (Del. Reg. (EU) 2024/2759) additionally tie redemption frequency, notice period and liquidity buffer to one another. Important for your expectations: this chain of controls tests process and defensibility, not “the one right number” — even a properly audited NAV remains a Level 3 estimate with a range around it.
IPEV in detail. The Guidelines (December 2025, to be applied from Q2 2026) are consistent with IFRS 13/ASC 820 (the “exit price” principle) and no longer recognise a blanket illiquidity discount — illiquidity belongs in the inputs (multiple adjustment, discount rate). Newly regulated or tightened: documentation discipline, the valuation of hybrid instruments and complex capital structures (liquidation preferences!), discounts for contractual transfer restrictions, and, for the first time, guardrails on the use of AI in valuation. Since 2018 the price of the last financing round is explicitly not a default fair value but only a calibration point whose informational value decays over time — relevant above all for growth and venture holdings with structured rounds.
Calibration is the real discipline. If you bought in at 10x EBITDA while peers were trading at 12x, you have to carry that relative discount forward — or justify unwinding it. IPEV also recommends backtesting: holding realised exits systematically against the last fair-value mark. Questions to put to the reporting: is backtesting disclosed? How often do exit proceeds differ from the last mark by more than 10–20%?
Governance under AIFMD Art. 19. The valuation function must be functionally independent of portfolio management — internally with Chinese walls, or as an external valuer (who is liable, must be professionally recognised and may not sub-delegate; the fund administrator, who merely adds up the NAV, does not count). Open-ended vehicles must value more often than annually, as their assets require. The ELTIF 2.0 standards additionally require reliable, up-to-date valuation data for each asset at every valuation date, and tie redemption frequency, notice period and liquidity buffer to one another.
Name the incentive question openly. In classic closed-end funds carried interest only arises on realised exits; many evergreens, by contrast, charge the performance fee on unrealised NAV gains — sometimes even on the immediate write-up of secondary positions bought at a discount. That is legal and disclosed, but it is a structural conflict of interest: worth examining are the high-water mark, the hurdle, and the question of whether the fee is clawed back if valuations fall later (clawback).
Putting “volatility laundering” in context. The term (coined by Cliff Asness/AQR) describes how infrequent model valuation systematically depresses reported volatility and correlation. For portfolio construction that means: NAV-based Sharpe ratios and correlations of semi-liquid funds are not comparable with daily exchange-traded time series; anyone modelling allocations should un-smooth them or add listed proxies. And as a practical indicator: large secondary market discounts and gates actually being applied are often more honest stress signals than the NAV itself.
De-smoothing in practice. Use smoothed series in risk models and you import the smoothing into your allocation: volatility and correlation that are too low lead mechanically to inflated private markets weights in any optimiser. Two standard corrections: Geltner unsmoothing treats the reported return as an AR(1)-weighted blend of the true return and the previous period’s return and works back from there; the model of Getmansky/Lo/Makarov generalises this to moving averages of higher order. Both typically raise volatility and equity beta markedly and lower Sharpe ratios — the exact values depend heavily on the data series and the assumptions; de-smoothing delivers ranges, not truth. For venture, Woodward (2011) shows that “stale values” depress measured risk — the true risk is often twice the reported figures. A pragmatic alternative: add listed proxies (listed PE, REITs) as a measure of risk, not of return.
Governance questions for the manager: Who values externally, how often does the mandate rotate, and does the valuer merely sanity-check or produce a valuation of its own? Is backtesting disclosed (realised exits versus the last mark)? How quickly and how far were the marks adjusted in 2022/23? What does the performance fee accrue on (realised versus unrealised), with what high-water mark, hurdle and clawback? Who sits on the valuation committee, and can portfolio management be outvoted there? Does the valuation frequency fit the redemption frequency — or are units redeemed monthly on the basis of quarterly marks?
NAV timing: the quiet transfer of wealth. In a semi-liquid fund every investor deals at the established NAV — not at the price a market would set today. If the NAV is still too high after a market fall (smoothing!), the redeeming investor is paid out more than their stake is economically worth; the difference is borne by the investors who stay. In mirror image, a new subscriber buys into a portfolio that is valued too highly. In rising markets both effects reverse. This is not fraud but a structural property of every vehicle that combines continuous dealing with delayed valuation — and it explains why gates, notice periods and the IPEV call for a higher valuation frequency in evergreens also protect the body of investors who remain. Further reading: Getmansky/Lo/Makarov (2004); Geltner (1991); Brown/Gredil/Kaplan (2019); Asness (2023); IPEV guidelines (2025).
Under the AIFMD rules: the valuation function must be independent of portfolio management; in practice that usually means a full quarterly valuation with monthly roll-forward. The IPEV guidelines (2025) urge a higher frequency for evergreen structures; the ELTIF standards additionally tie redemption frequency, notice period and liquidity buffer to one another.
The NAV is the best rules-based estimate of the sale value — but an estimate with a range around it (“Level 3”), not a market price. Realised sales regularly deviate by a few per cent; in the secondary market, buyers paid an average of 87% of NAV in 2025 (Jefferies).
Because it is valued infrequently and with the help of models: quarterly marks plus roll-forward smooth the curve, and valuers adjust to changes in value only with a lag. In 2022 the Nasdaq lost 33% while buyout funds stood almost at zero on MSCI Burgiss data — partly a genuine difference, in large part valuation inertia.
The effect whereby reported NAV returns are a moving average of the true performance: less measured volatility, lower measured correlation, inflated Sharpe ratios. Described in the research by, among others, Getmansky/Lo/Makarov (2004) and, for property, by Geltner (1991).
The bottom tier of the fair value hierarchy under IFRS 13: the value rests on material assumptions that are not observable in the market — model plus judgement instead of an exchange quote. Practically all the assets of semi-liquid private markets funds are Level 3; annual reports disclose the share.
The room for manoeuvre is tightly constrained by IPEV rules, independent valuation, the administrator, the auditor and the supervisor — but it is not zero. The research (Brown/Gredil/Kaplan 2019) finds inflated marks above all at weaker managers during fundraising phases — who are punished for it by the market. Watch whether the performance fee accrues on unrealised book gains: then the manager earns on its own valuation.
It follows with a lag and usually in muted form: the multiples and discount rates in the models only catch up with the next quarterly valuations. In that phase the NAV can sit above the realistic sale value — secondary market discounts and redemption queues show the stress earlier.
Last reviewed: 23 Aug 2026 · Methodology · Not investment advice.