Liquidity stress testing

Two duties with different perimeters

FUND 3.6.3R(2) requires regular liquidity stress tests under normal and exceptional conditions for each AIF other than an unleveraged closed-ended AIF. FUND 3.7.5R(2)(b) separately requires the risks of each position and their overall effect on the portfolio to be identified, measured, managed and monitored on an ongoing basis, including through appropriate stress testing procedures, and carries no closed-ended carve-out.

A closed-ended unleveraged buyout or infrastructure fund therefore runs no liquidity stress test and still owes the FUND 3.7 procedures on its positions. FUND 3.4.3R(5) names both duties as sources of results the AIFM must provide to the FCA, so a firm running only the liquidity test files a return with a visible gap.

The asset-side assumption

In its 2023 multi-firm review the FCA found a wide range of approaches, some methodologies insufficient to assess the actual liquidity of the portfolio, and assumptions that were not appropriately conservative. The defect it named was a modelling choice: some firms' models assumed they would always sell the most liquid assets, without regard to the liquidity of selling a vertical slice.

The objection is narrower than a ban on waterfall selling: a model that never tests the alternative has not tested the fund's real liquidity, because the horizon it reports can be delivered only by treating the remaining investors worse. A firm using it should produce both numbers and say which its dealing practice follows.

Scenario severity

The FCA's 2025 review of liquidity risk management at wholesale trading firms found that some firms described shocks as extreme even though previously observed volatility had reached similar or greater levels. Stating each scenario alongside the worst comparable historical episode for that asset class makes severity a claim the governing body can check rather than a label. The same test applies to reverse stress testing: where the level of redemptions the fund could not meet is lower than something that has already happened in its own market, the governing body should be told in those terms.

The redemption being modelled

The 2023 review found that firms typically had governance and organisational arrangements to meet large one-off redemptions, but not to oversee cumulative or market-wide redemptions capable of having a significant impact on a fund. That distinction changes the price side of the model. A single large redemption is a fund-level event with a market to sell into. Cumulative redemptions over several dealing dates erode the buffer without triggering a threshold. Market-wide redemptions mean every holder of the same assets is selling on the same day, so the price the model assumed is not available. A test built on the first case and applied to the third reports a horizon the fund does not have.

Evidence on realisation horizons

The closest UK datapoint is a wind-down rather than a stress, which makes it a floor. M&G announced the closure of its Property Portfolio in October 2023, holding £565 million across 25 properties with 17.24% in cash, and estimated up to 18 months to realise. All assets were confirmed sold in mid-May 2026, roughly 29 months, with no redemption queue and no forced timetable. It had run from £2.1 billion at the end of 2019 to £565 million at closure on persistent outflows rather than a single event: the liability side moved first, over four years, and no shock scenario would have caught it.