Key Takeaways
- • A new study finds that measuring investment skill at the total portfolio level is unreliable because pension fund portfolios diverge from simple stock-and-bond reference portfolios, making benchmark-relative returns highly noisy.
- • Even after 10 years, a genuinely skilled pension fund underperforms its reference portfolio 15 per cent of the time due to noise alone; it takes more than 20 years for that probability to fall below five per cent.
- • A fund with no skill has a 20 per cent chance of outperforming its reference portfolio by one per cent annually over 10 years purely by luck, and it takes more than 20 years to bring that false-positive risk below five per cent.
- • Diversification is identified as the noisiest source of value-add and the main driver of tracking error that can obscure other sources of skill.
- • Proposed remedies include disclosing false-positive and false-negative probabilities, using a diversified benchmark, and placing greater weight on independent ex ante reviews of governance, portfolio construction and risk management.
Summary
Pension funds that have lagged simple portfolios of listed stocks and bonds in recent years have prompted questions about why boards pay professional asset managers rather than investing directly. A new study, as described by Benefits Canada, argues that this conclusion assumes investment skill can be reliably inferred from past returns, and that this assumption fails at the total portfolio level.
A core benchmarking principle is that a portfolio should closely track its benchmark, meaning low tracking error. Pension funds, however, pursue opportunities beyond public stocks and bonds and make active asset allocation decisions, so their portfolios diverge from the simple "reference" portfolio of listed stocks and bonds. That divergence makes benchmark-relative performance noisy, and skill becomes difficult to distinguish from luck.
The report quantifies the effect. A genuinely skilled fund still underperforms its reference portfolio 15 per cent of the time after 10 years purely from noise, a "false negative." The probability falls below five per cent only after more than 20 years. Conversely, a fund with no skill has a 20 per cent chance of beating the reference portfolio by one per cent a year over 10 years through luck, a "false positive," again requiring more than 20 years to reduce to five per cent.
The article proposes three steps. First, boards should recognize the uncertainty in point estimates of value-add and funds could disclose the probability of false positives and negatives. Second, benchmark design should be revisited: because diversification is the noisiest source of value-add, boards could compare performance to a broadly diversified benchmark that already captures diversification benefits, distinct from policy benchmarks reflecting strategic asset allocation. Third, boards should emphasize ex ante assessments of investment processes, such as governance, portfolio construction and risk management, conducted by independent experts, since these help distinguish good decisions with bad outcomes from the reverse.
The author concludes that benchmarking methods designed for narrow active strategies, such as small-cap value funds, are poorly suited to evaluating a pension fund's total portfolio. Outperforming the reference portfolio does not necessarily prove skill, and underperformance does not necessarily prove its absence.
A core benchmarking principle is that a portfolio should closely track its benchmark, meaning low tracking error. Pension funds, however, pursue opportunities beyond public stocks and bonds and make active asset allocation decisions, so their portfolios diverge from the simple "reference" portfolio of listed stocks and bonds. That divergence makes benchmark-relative performance noisy, and skill becomes difficult to distinguish from luck.
The report quantifies the effect. A genuinely skilled fund still underperforms its reference portfolio 15 per cent of the time after 10 years purely from noise, a "false negative." The probability falls below five per cent only after more than 20 years. Conversely, a fund with no skill has a 20 per cent chance of beating the reference portfolio by one per cent a year over 10 years through luck, a "false positive," again requiring more than 20 years to reduce to five per cent.
The article proposes three steps. First, boards should recognize the uncertainty in point estimates of value-add and funds could disclose the probability of false positives and negatives. Second, benchmark design should be revisited: because diversification is the noisiest source of value-add, boards could compare performance to a broadly diversified benchmark that already captures diversification benefits, distinct from policy benchmarks reflecting strategic asset allocation. Third, boards should emphasize ex ante assessments of investment processes, such as governance, portfolio construction and risk management, conducted by independent experts, since these help distinguish good decisions with bad outcomes from the reverse.
The author concludes that benchmarking methods designed for narrow active strategies, such as small-cap value funds, are poorly suited to evaluating a pension fund's total portfolio. Outperforming the reference portfolio does not necessarily prove skill, and underperformance does not necessarily prove its absence.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.