How Beliefs Shape Institutional Investment Decisions

Key Insights for Investors

Institutional investors manage trillions of dollars in assets, yet we know surprisingly little about how their expectations about future returns translate into portfolio decisions. Aleksandar Andonov, Spencer Couts, Andrei Gonçalves, Johnathan Loudis, and Andrea Rossi, authors of the January 2026 study “Subjective Beliefs and the Portfolio Allocations of Institutional Investors,” shed light on this critical question by studying U.S. public pension funds and their investment consultants.

What the Researchers Examined

The authors tackled a fundamental question in finance: Do institutional investors’ beliefs about risk and return actually drive their portfolio allocations?

To answer this, they began by linking the actual portfolio holdings of 181 U.S. public pension funds (covering $73 trillion in assets from 2001-2023) with the long-term capital market assumptions (CMAs) of their investment consultants. However, they were able to match 39% of the fund years to consultant CMAs. These CMAs contain detailed forecasts of expected returns, volatilities, and correlations across six major asset classes: cash, fixed income, public equity, private equity, real estate, and hedge funds. The remaining assets that were not classified into one of these six asset classes accounted for, on average, just 2.2% of pension assets.

Using a mean-variance framework, the researchers asked: How well do the “optimal” portfolios implied by these beliefs explain what pension funds actually hold?

Key Findings

1. Beliefs Matter—But Context Is Everything

The most striking finding is that beliefs do significantly influence portfolio decisions, but much more strongly when you account for real-world frictions. When the researchers used a simple, frictionless model, consultant beliefs explained very little of actual pension fund behavior. The relationship between predicted and actual allocations had slope coefficients almost two orders of magnitude below what theory would predict (albeit often statistically significant).

However, once they incorporated realistic constraints—specifically non-negative position limits and benchmarking incentives—the picture transformed dramatically. In their full model, between 58% and 88% of the variation in pension fund allocations and 48% and 74% of active allocations (i.e., deviations from the benchmark) could be explained by belief-driven optimal allocations.

2. Diversification Benefits Are Crucial

An important innovation in this research is incorporating the full covariance matrix—not just expected returns. The authors found that accounting for how different asset classes correlate with each other substantially strengthened the connection between beliefs and allocations. This makes intuitive sense: an asset class might look unattractive in isolation, but highly attractive when considering its diversification benefits. Previous research focusing only on expected returns was missing a critical piece of the puzzle.

3. Institutional Frictions Shape Investment Reality

Two frictions proved essential for matching theory to practice:

• Non-negative weight constraints: Institutions generally can’t short asset classes, which dramatically changes optimal portfolios.

• Benchmarking incentives: Pension funds face pressure to not deviate too far from peer allocations, whether due to career concerns, formal tracking error limits, or uncertainty about their own beliefs.

The benchmarking effect was particularly powerful. The researchers calibrated this parameter so that theoretical tracking errors matched actual ones, finding it explains why funds don’t make more extreme bets even when their beliefs might justify them.

4. The Consultant-Pension Fund Belief Connection Is Real

A potential concern was whether consultant CMAs actually reflect pension fund beliefs, or if this is just reverse causality (funds selecting consultants who justify predetermined allocations). The evidence strongly supports a genuine belief connection:

• Consultant expected returns correlate positively with pension funds’ own reported return assumptions.

• Consultants serve diverse client bases (pension funds represent only 17% of clients), making it unlikely they tailor beliefs to any single client type.

• Consultant-fund relationships last an average of 16 years, and the belief-allocation link actually strengthens over time—the opposite of what we’d expect if selection drove the results.

• Consultant forecasts are unbiased predictors of future returns, inconsistent with them simply rationalizing client preferences.

5. Allocation “Mistakes” Are Modest

The researchers calculated how much pension funds would be willing to pay (under consultant beliefs) to move from their actual allocations to the theoretical optimum. The median annual value was around 0.15%—meaningful but less than one-third of typical investment management costs.

This suggests that while deviations from belief-implied optimal allocations exist, they’re not necessarily large mistakes. Some likely reflect rational considerations beyond the mean-variance framework, such as liability matching, liquidity needs, or governance constraints.

Their findings led the authors to conclude: “Our results demonstrate that risk and return expectations play a central role in institutional portfolio decisions, but isolating their effect requires recognizing the constraints and frictions that attenuate their transmission to observed allocations.”

Key Investor Takeaways