Key takeaways
AQR's outflows continued for two years after performance recovered. The outflow cycle, once triggered, operated on a separate timeline from returns.
Asness published continuously through the drawdown in academic journals and on a public blog. The content was built to be challenged, not to reassure.
Institutional allocators were measurably stickier than the retail base. Massachusetts PRIM maintained roughly $1 billion across two AQR funds throughout the entire period.
AQR never recaptured its $226 billion peak. Strong communication infrastructure still left a roughly $50 billion permanent shortfall.
A firm that shrank by half while its founder published more than ever
AQR Capital Management peaked at $226 billion in assets under management in 2018. By the end of 2022, the firm managed roughly $95 billion. Cliff Asness, the firm's co-founder, attributed about one-third of the overall decline to performance. The rest was investors pulling capital out. During that entire stretch, Asness was publishing blog posts, academic papers, and public arguments for why the strategy would recover. Not quarterly. Continuously.
The flagship Absolute Return fund eventually posted 43.5% in 2022, its best year since the 1998 launch. Firm assets climbed back to approximately $175 billion by October 2025. But AQR never recaptured its peak. What happened during those years reveals something specific about which LPs are reachable through communication during a crisis, and what kind of content actually reaches them.
Two-thirds of the AUM decline was behavioral, not mathematical
The 2018 to 2020 “quant winter” hit AQR's value-heavy strategies hard. The flagship Absolute Return fund declined more than 30% from peak to trough. That was painful. But the outflows were worse than the performance losses.
Investors started pulling money in 2018 and kept pulling through 2023, even after returns had already turned positive. The mutual fund book alone lost more than $25 billion. AQR cut roughly 40% of its workforce, closed strategy lines, and liquidated several funds. By late 2021, total firm assets had fallen almost 40% from the peak. By end-2022, total firm assets sat near $95 billion.
The timeline shows that performance began recovering in late 2020. The asset base kept shrinking for two more years. The outflow cycle, in this case, operated on a completely separate timeline from the returns.
Publishing into the drawdown at academic depth
Where drawdown communication often defaults to cautious quarterly updates, Asness went the other direction. While AQR was closing funds and reducing headcount, he was producing public intellectual output that ran directly at the hardest questions facing the firm. Specific, dateable examples:
July 2019: A “Cliff's Perspectives” post titled “Quant Cassandra”, referencing a presentation he had shared with clients for years predicting that factor investing would eventually be hated. He wrote that forecasting the drawdown was “obvious” but getting people to internalize it was something else entirely.
May 2020: A public defense of value investing during the largest growth-stock surge in a generation, arguing that diversified global value strategies looked historically cheap.
Ongoing: A 23-page paper for the Journal of Portfolio Management, “The Less-Efficient Market Hypothesis,” building the case that markets had become less efficient over the course of his career, which, if true, meant the value factor was riskier but likely more rewarding long-term.
The content was analysis, not reassurance. It was published in venues where it could be examined, challenged, and cited independently. Asness described the internal process: “We went through hundreds of hypotheses generated by the outside because the world's very good at explaining why whatever's been happening is just and true and will happen forever.”
AQR investigated whether value was permanently broken, stress-tested the alternatives, and published the conclusions. That degree of openness was unusual even in normal conditions. AQR's research culture runs opposite to secretive peers like Renaissance Technologies. During the drawdown, Asness leaned further into it.
Bridgewater Associates built a similar function differently: the Daily Observations and Principles framework gave LPs an interpretive lens assembled over decades, before any crisis hit. Asness built his in real time, under pressure, in public. Different mechanisms, same structural purpose: giving institutional investors a framework for evaluating performance that extended beyond the return number.
Who stayed, who left, and what separated them
Michael Trotsky, CIO of Massachusetts Pension Reserves Investment Management, kept approximately $1 billion invested across two AQR funds throughout the drawdown, maintaining a 15-year relationship with the firm. His assessment of Asness: “Cliff absolutely is a survivor. He and AQR believe in value, and being a long-term believer in that style is a rarity these days.”
Meanwhile, the retail and advisor channel drove the majority of AQR's outflows. While AQR didn't waver on its thesis, many mutual fund investors had already bailed before the rebound, with fund assets falling from a $51.4 billion peak to $26.9 billion. The firm eventually pruned its fund lineup and exited entire strategy areas.
Lock-ups and portfolio construction constraints partly explain why institutional allocators were stickier. But institutional LPs tend to make redemption decisions through governance processes: committees, board reviews, consultant input. Those processes create space for a thesis to be presented and evaluated.
Output built for institutional scrutiny
Because Asness's analysis was published openly, in academic journals and on a public blog, it carried a different weight inside institutional review processes than typical manager communications.
Independently citable, falsifiable, and built to engage with counterarguments, the work was usable in ways that a quarterly letter reassuring LPs about long-term conviction never could be. The retail channel had no way to process that kind of output. Asness's publishing held some of the institutional base. It was never going to hold the rest.
What the recovery showed and what it left unresolved
The performance story is strong:
The flagship Absolute Return fund rallied 43.5% in 2022, its best year since the 1998 launch.
It followed with an 18.5% gain in 2023.
AQR launched its most aggressive multistrategy vehicle, Apex, in 2020 during the trough, and it gained 16.2% in 2023.
Launching a new product during a contraction, while closing others, signaled something a quarterly letter can't: operational conviction backed by resource commitment at a moment when resources were scarce.
But the recovery has clear limits. AQR's roughly $175 billion in late-2025 assets still sits about $50 billion below its peak. That shortfall represents capital that left and never returned, even after four consecutive years of strong results. The firm emerged leaner, with fewer products, fewer people, and a different organizational shape.
The counterargument: value rotated because macro conditions changed. Interest rates rose. Growth multiples compressed. Asness's communication may have helped retain the institutional base that captured the rebound, but it didn't cause the rebound. The LPs who stayed were rewarded by market conditions. How much of their staying was driven by Asness's output, and how much by lock-ups, portfolio construction, or independent conviction in value, is impossible to isolate cleanly.
Bottom line
Asness's case is one of the few documented examples of a hedge fund founder treating public intellectual output as a core firm function during a sustained drawdown. The practical takeaway: Publishing credibly under pressure requires a channel and a track record that already exists before the pressure arrives.
Communication infrastructure that can survive a multi-year drawdown needs to be producing analytical content before the drawdown starts. Asness could publish credibly during AQR's worst period because he had been publishing for 20 years. The LP base had already been conditioned to evaluate thesis-driven arguments, not just return streams. He just kept using it.
Most IR programs aren't tested until the drawdown arrives. The AQR case suggests that that's too late to find out whether they work. If you're thinking about what your communication infrastructure looks like under stress, that's a conversation we have regularly.
