Every risk-adjusted measure compares a portfolio to a benchmark. If the benchmark is invalid, all of them are decoration. Bailey, Richards and Tierney's seven properties are the standard test, and a valid benchmark is:
- Specified in advance: chosen and disclosed before the period, not selected afterwards from the ones that lost.
- Appropriate: consistent with the manager's actual style and universe.
- Measurable: its return can be computed with reasonable frequency.
- Unambiguous: the constituents and their weights are clearly identified.
- Reflective of current investment opinion: the manager has knowledge of and a view on the securities in it.
- Accountable: the manager accepts it as the standard of judgement; "I am accountable for this" is a commitment, not a label.
- Investable: you could passively hold it instead. A benchmark you cannot buy is not an alternative, so beating it proves nothing.
Manager peer-group medians fail several outright: not specified in advance, not unambiguous, not investable, and survivorship-biased because funds that closed are absent from this year's median. A universe median is useful context and an invalid benchmark; both are true.
Beyond the checklist, run the quantitative quality tests: coverage (what fraction of portfolio holdings appear in the benchmark), turnover (a benchmark should be cheap to hold), and the systematic-bias test: regress the manager's active return (RP − RB) on the benchmark's own active return against a broad market index (RB − RM). A valid benchmark shows no significant relationship; a significant positive coefficient means the "alpha" is a style tilt already embedded in the benchmark choice.
Advertisement
What are misfit return and the normal portfolio?
Where a benchmark is imperfect, decompose. Insert the manager's normal portfolio (the mix she would hold with no active views) between the portfolio and the investor's benchmark:
- True active return = portfolio − normal portfolio (what she actually did)
- Misfit return = normal portfolio − investor's benchmark (a style mismatch the investor chose, not the manager)
- and the two sum to total active return.
With a portfolio at 12.4%, a normal portfolio at 11.0% and an investor benchmark at 9.6%, the total active +2.8 points splits exactly in half: +1.4 true, +1.4 misfit. Half her apparent outperformance is her; half is the investor having hired a value manager and measured her against a blend. Risks combine in quadrature: with true active risk 3.2% and misfit risk 4.1%, total active risk is √(3.2² + 4.1²) = 5.20%, so the true-active IR is 1.4/3.2 = 0.44 while the headline IR is 2.8/5.20 = 0.54. Attributing misfit to the manager flatters her by ten points of information ratio.
Why does a price return benchmark flatter a fund?
And then the defect that costs Indian investors real money. A price return index (PRI) counts only price change; a total return index (TRI) reinvests dividends. Benchmarking a fund's NAV return (which includes the dividends the fund received) against a PRI compares a total return to a price return, and hands the manager the entire dividend yield as free alpha. Take a large-cap fund up 14.6% against a Nifty-50-style PRI at 13.1% and its TRI at 14.4% (illustrative values; pull the live ones from niftyindices.com): against PRI the fund shows +1.50 points of alpha; against TRI, +0.20. Eighty-seven percent of the apparent skill was dividends. Compounded, a 1.3-point annual wedge manufactures 6.67% of phantom cumulative alpha over five years and 13.79% over ten. SEBI closed this in India with its January 2018 circular requiring mutual-fund schemes to benchmark against Total Return Indices, effective 1 February 2018. Read the circular itself on sebi.gov.in rather than a summary of it. The lesson generalises far beyond India: before you compute a single basis point of alpha, confirm that both sides of the subtraction are the same kind of return.