Home › CFA® Level III › Performance Measurement
CFA® Level III · Performance Measurement · free lesson
Portfolio Performance Evaluation
Measuring a return is easy; explaining it and judging whether it reflects skill is the hard part. This module separates measurement, attribution and appraisal, decomposes active returns with Brinson, factor and fixed-income models, assigns results to the asset owner or the manager, tests whether a benchmark is fit for purpose, and applies the risk-adjusted measures (Sharpe, Treynor, M-squared, information and appraisal ratios, Sortino, capture ratios and drawdown) used to judge a manager.
What you will learn
1. Distinguish performance measurement, attribution and appraisal, and describe the features of a sound attribution process.
- Measurement reports return and risk; attribution explains them relative to a benchmark; appraisal judges skill versus luck.
- Return attribution explains excess return; risk attribution explains active or total risk.
- Macro attribution works at the asset-owner level; micro attribution explains one manager's portfolio.
- A good model mirrors how decisions are actually made, so each effect corresponds to a real choice.
- Effects should sum to the total excess return; a large residual signals a mismatch between model and process.
- Geometric excess returns compound across periods; arithmetic effects need a linking algorithm.
2. Interpret returns-based, holdings-based and transactions-based attribution, including Brinson, factor-based and fixed-income attribution, and risk attribution.
- Returns-based attribution needs the least data and is least accurate; holdings-based misses intra-period trades; transactions-based is the most accurate.
- BHB and BF give the same total allocation effect; BF credits an overweight only if the segment beat the total benchmark.
- Selection uses benchmark weights; interaction is the active weight times the segment's selection return.
- Factor attribution: active return = active factor exposures times factor returns, plus a residual credited to selection.
- Bond attribution separates income, curve (duration) and spread effects from selection.
- Risk attribution decomposes tracking error for relative mandates and total volatility for absolute mandates.
3. Separate the results produced by the asset owner's decisions from those produced by the investment manager's decisions.
- Owners control strategic allocation, benchmark and style choices, manager weights and their own tactical tilts.
- Managers control only decisions within their mandate, measured against the benchmark they were given.
- Macro attribution layers: risk-free, asset category (policy), benchmark or style, manager, allocation.
- Manager effect = actual weight times (manager return minus its own benchmark return), summed over managers.
- A manager can add value while the fund loses money or lags a broad index because of the owner's choices.
4. Evaluate asset-based and liability-based benchmarks, apply tests of benchmark quality, and explain why alternative investments are hard to benchmark.
- Valid benchmarks are specified in advance, appropriate, measurable, unambiguous, reflective of the manager's opinions, accepted by the manager and investable.
- Peer-group medians fail the specified-in-advance and investable tests and carry survivorship bias.
- For liability-driven investors the benchmark is the return on liabilities; the funded ratio improves when assets beat it.
- Quality tests: beta near 1, active returns uncorrelated with the benchmark's style return, low tracking error, high coverage, low turnover.
- Hedge fund, private equity, real estate and commodity benchmarks suffer from bias, smoothing, lags or roll effects.
5. Calculate and interpret risk-adjusted performance measures, including the Sortino ratio, appraisal ratio, capture ratios and drawdown, and judge whether a manager has skill.
- Sharpe and M-squared use total risk; Treynor and Jensen's alpha use beta; information and appraisal ratios use active or residual risk.
- M-squared minus the market return ranks portfolios exactly as the Sharpe ratio does, but in percentage points.
- Sortino uses downside deviation below a minimum acceptable return.
- Capture ratio = upside capture / downside capture; above 1 is a favourable asymmetry.
- Maximum drawdown and its duration capture path-dependent losses that volatility hides.
- The t-statistic of skill is roughly IR x square root of years, so proving skill takes a long record.
Module at a glance
- Measurement says what happened, attribution says why, appraisal says whether it was skill.
- Brinson-Fachler allocation credits an overweight only when the segment beat the total benchmark; selection uses benchmark weights; interaction is the cross term.
- Returns-based, holdings-based and transactions-based attribution trade data needs against accuracy; bond attribution separates duration, curve and spread effects from selection.
- Owners own the strategic allocation, benchmark and style choices; managers own only their results against the assigned benchmark.
- Valid benchmarks are specified in advance and investable; peer medians fail, and liability-driven investors are judged against the liability return.
- Sharpe and M-squared for total risk, Treynor and alpha for beta, information and appraisal ratios for active risk, Sortino for downside risk.
- Capture ratios and maximum drawdown reveal asymmetry and path-dependent losses; proving skill takes years because t is about IR times the square root of T.
This free module has 5 lessons, 18 practice questions with full explanations and 1 essay case with a model answer.