Why traditional investment committees can amplify group biases

Investment committee meetings, a governance cornerstone at almost every asset owner organisation, run the risk of amplifying group biases and social dynamics, and can push the IC towards more extreme investment positions collectively than the average of their individual views, according to a senior ADIA investor.

Bernhard Scherer, head of portfolio implementation at ADIA, argued in a recent paper that traditional IC meetings tend to introduce drifts in opinions “aligned with the salient story of the day, thereby increasing the correlation of errors”.

“A traditional IC meeting is a monthly ritual. Members arrive with thick slide decks, most of which are lightly rebranded broker reports, copied charts from last week’s strategy notes, and valuation tables whose axis labels are either missing or incompatible,” he wrote in a paper for the CFA Institute Research Foundation.

“The meeting begins with a macro overview. Someone speaks first — usually the CIO, sometimes the most confident strategist — and sets the tone.

“Dissent is polite but rarely decisive. Nobody writes down an explicit portfolio. Nobody specifies how much risk they would actually like to take if they were accountable for the outcome.”

This kind of “conversation-driven” IC meeting presents several structural “pathologies”, he argued, including information sequencing and cascades, where the IC gravitates towards a strong view presented by a dominant member early on to avoid disagreement; group polarisation, where the IC lands on more extreme positions incentivised by the majority of views; and the disposition effect, where teams tend to delay loss realisation and defend past consensus decision.

The paper recalled literature from behavioural psychologist Daniel Kahneman and suggested the “hygiene” of decision-making could be improved with several methods. One of them is running a “noise edit” where an IC, before deciding that it “has a view” on certain issues, lets its members independently assess the same macro scenario and risk constraints and measure how far apart they are from reaching a consensus.

Another exercise breaks down complex problems into manageable, explicit dimensions. For example, an M&A transaction can be separated into valuation, downside risk, and management quality. Each IC member will score these dimensions separately, before any group discussions.

For organisations looking for alternatives to traditional IC meetings and decision-making, Scherer proposed an approach named “anonymous portfolio-vector averaging”. Each IC member will submit a portfolio weight vector, before all vectors are scaled to a common ex ante risk level and are averaged. The resulting aggregate is then rescaled to a target tracking error and implemented.

“If asset owners are going to have ICs anyway, they should at least use them to diversify opinions in a disciplined, algorithmic way, rather than to amplify noise and provide post hoc narratives,” he said.

Lessons in the past

Chasing consensus can damage not only fund-level but even industry-level investment results, Scherer said, recalling the liability-driven investment (LDI) crisis that enshrouded UK pensions in 2022. The gilt market was sent into a tailspin when the UK government announced a series of unfunded tax cuts and triggered a selloff in government debt, forcing yields sharply higher.

The UK’s defined benefits schemes used LDI strategies which consisted of derivatives exposed to the price of gilts as a way to hedge liabilities. To meet increasingly urgent margin calls, they sold long-dated gilts as the market plummeted, which only exacerbated the sell-off and eventually triggered central bank intervention.

“This LDI episode is a reminder that it is insufficient for committees to adopt what looks like a sophisticated consensus. If everyone adopts the same structure, the system behaves like one highly centralised IC portfolio,” Scherer said.

“Most schemes had trustee boards and ICs, often advised by consultants and LDI managers. The committees embraced a single risk management narrative: hedging liabilities through derivatives to stabilise measured funding ratios.

“Leverage and collateral liquidity risk were underappreciated at the IC level; many trustees did not fully understand the mechanics of collateral waterfalls and stress scenarios.”

Scherer also cited the 2008 liquidity crisis among US university endowments as an example where the IC’s buy-in of CIO dominance and investment narrative became structural risks. During the GFC, many US university endowments that followed the Yale model, which dictated heavy allocation to illiquid assets, were hit with large drawdowns and had to cut costs or raise expensive liquidity.

“In practice, the governance bargain was simple: Committees bought into the CIO’s long-horizon illiquidity narrative and did not insist on a granular, quantitative treatment of liquidity and survival constraints,” Scherer said.

“Once a single story (illiquidity premia, synthetic hedging, liability matching) dominates the committee, information cascades, group polarisation, and free riding kick in.

“The committee stops being a diversified provider of independent views and becomes a rhetorical support act for the CIO.”