Performance Measurement

Performance Measurement: How to Evaluate Managers

Assessment of asset management outcomes is a complex task riddled with pitfalls. Raw returns do not provide a complete picture; it is necessary to account for risk, benchmark, time horizon, and the impact of luck. Correct attribution analysis allows for separating skill from luck and assessing sources of results.

Basic Return Metrics

Time-weighted return (TWR) measures returns by eliminating the impact of cash flows. This is the standard for comparing managers, as it isolates the investment decision from decisions about subscription/redemption.

Money-weighted return (IRR) takes into account the timing and size of cash flows. This is the return actually received by the investor.

TWR and IRR can differ significantly — a good manager may have a poor IRR if investors entered at the peak.

Gross vs Net returns: gross — before fees; net — after fees. Comparison of managers should be on the same basis. Net returns show the real result for the investor.

Risk-Adjusted Metrics

Sharpe Ratio = (Return - Risk-free rate) / Standard Deviation. Measures return per unit of total risk.

Advantages: simplicity, universality.

Disadvantages: assumes a normal distribution, does not distinguish upside and downside volatility.

Sortino Ratio uses downside deviation instead of standard deviation, penalizing only for negative volatility. More relevant for investors concerned with losses.

Information Ratio = Active Return / Tracking Error. Measures active return relative to the benchmark per unit of active risk. Shows how efficiently the manager uses deviation from the benchmark.

Alpha and Beta

CAPM alpha — return unexplained by market risk: Alpha = Portfolio Return - [Rf + Beta * (Market Return - Rf)]. A positive alpha indicates added value.

Multi-factor alpha uses models with several factors (Fama-French, Carhart). This helps to understand if alpha is explained by exposure to known factors (value, size, momentum) or reflects genuine skill.

Persistence of alpha is a key issue: does past alpha predict future alpha? Research shows weak persistence for most strategies. This questions the usefulness of historical track record for forecasting.

Attribution Analysis

Brinson attribution splits the result into allocation effect (choice of sector weights) and selection effect (choice of securities within sectors). This allows one to understand where alpha came from — from strategic decisions or stock picking.

Fixed income attribution is more complex: duration contribution, yield curve positioning, spread effect, security selection. Multiple sources of returns require detailed breakdown.

Risk-based attribution explains returns through exposures to risk factors. Factor mimicking portfolios allow replication of the part of return not explained by Brinson.

Practical Considerations

Survivorship bias — databases include only surviving managers. Underperformers close and disappear from the statistics, inflating industry average results.

Look-ahead bias arises when using data unavailable at the moment of decision-making. Particularly dangerous in backtesting.

Sample size and statistical significance: a short track record does not allow one to separate skill from luck. Even several years of good results may be chance. Years of data are needed for statistically significant conclusions.