Every fund manager claims to add value. The fee they charge is justified, they say, by the extra return they deliver. But financial markets are noisy. A manager can outperform for three years in a row purely by chance. And a genuinely skilled manager can underperform for two years in a row and still be doing their job well. So how do you tell the difference? That is what performance evaluation is for.
Measuring Returns the Right Way
Total return: The full gain or loss from an investment including price changes, dividends, and interest payments
Active return: The difference between a portfolio's return and its benchmark's return. Positive means the manager beat the benchmark
Benchmark: A standard index used to judge a portfolio's performance, such as the Nifty 50 or BSE Sensex for Indian equity portfolios
Measuring returns sounds simple. It is not. When clients add money to or withdraw money from a portfolio during the year, the timing of those flows affects the apparent return. Standard best practices, set out in the Global Investment Performance Standards, define exactly how to handle this so that comparisons between managers are fair.
A second subtlety: arithmetic returns are misleading over multiple periods. If a portfolio gains 10% in January and loses 10% in February, the arithmetic sum is 0%. But the investor has actually lost money. 100 rupees grew to 110 rupees and then fell to 99 rupees. The real return is negative 1%.
Why logarithmic returns matter: Logarithmic returns (also called log returns) correctly account for compounding. They can simply be added across time periods and always give the right answer. Arithmetic returns cannot be added this way. For measuring and comparing performance over multiple periods, log returns are more accurate.
Portfolio Q earns +10%, -10%, and +15% in three successive months. The arithmetic sum of returns is +15%. But the actual cumulative return is (1.10) times (0.90) times (1.15), minus 1, which equals 13.85%. Using log returns gives 12.97%, which correctly reflects the compounding penalty from the -10% month. A portfolio with steadier returns will outperform a volatile one with the same arithmetic average.
Risk-Adjusted Performance: Three Key Ratios
Why adjust for risk at all? Any manager can increase returns by simply taking more risk. Borrowing money to double up on equities will double the return in a good year and double the loss in a bad year. We need measures that reward managers for the return they earn per unit of risk they take.
Sharpe Ratio: Return above the risk-free rate, divided by the portfolio's volatility. Higher is better. It measures return earned per unit of total risk
Information Ratio: Return above the benchmark, divided by tracking error (how much the portfolio deviates from the benchmark). Measures skill at beating the benchmark consistently
Sortino Ratio: Similar to the Sharpe Ratio but only counts downside volatility as risk. Useful when a client has a specific minimum return they cannot fall below
The Sharpe Ratio and Information Ratio measure different things. Raising or lowering the risk-free interest rate does not change a manager's Information Ratio at all, because the benchmark and the portfolio move together when rates shift. But raising interest rates does reduce the Sharpe Ratio, because it raises the return that cash earns, making it harder for the portfolio to look good by comparison.
The Sortino Ratio is most useful for clients in retirement who have a clear minimum they need. A fund that never falls below 4% real return but occasionally only earns 5% looks the same as a fund with average return of 7% but occasional -5% years under the Sharpe Ratio. The Sortino separates these correctly.
Mr R's fund has an average active return of 2% per year above its benchmark, with a tracking error of 4%. His Information Ratio is 2 divided by 4 = 0.5. Ms S manages a similar fund with an active return of 1% and tracking error of 1%. Her Information Ratio is 1 divided by 1 = 1.0. Ms S is the more consistent alpha generator, even though Mr R earned more in absolute terms last year.
Value at Risk and Expected Shortfall: Understanding Tail Risk
Value at Risk (VaR): The maximum loss you would expect to suffer on 95% of days (or any chosen probability). It does not tell you how bad the worst 5% of days can get
Expected Shortfall (ES): The average loss on the worst 5% of days. It fills the gap that VaR leaves by describing how bad the tail actually is
VaR has a blind spot. Two portfolios can have identical VaR values but very different risk profiles. One portfolio holds plain equities. Another holds equities plus a large short position in put options. In normal markets, both look similar. In a market crash, the options come into the money and the second portfolio suffers catastrophic losses far beyond what the VaR suggested.
Expected Shortfall captures this. It is the average of all the losses worse than the VaR threshold. Regulators, including the Basel Committee for global banks, shifted from VaR to Expected Shortfall in 2016 for exactly this reason. For any portfolio with options or other non-linear instruments, Expected Shortfall is the more honest risk measure.
Attribution: What Actually Drove the Return?
What is attribution? Attribution breaks a portfolio's performance into parts, each traceable to a specific decision. Did the manager beat the benchmark because they chose the right sectors? Or the right individual stocks within those sectors? Attribution answers this.
The most widely used attribution framework was developed in 1986 and splits active return into three parts.
Allocation effect: The return earned by overweighting sectors that performed well and underweighting those that performed poorly. This reflects macro-level sector calls
Selection effect: The return earned by picking stocks within each sector that outperformed other stocks in that sector. This reflects stock-picking skill
Interaction effect: The extra return from correctly overweighting sectors where the stock picks were also good. This is the combination of both skills working together
A portfolio holds 70% equities and 30% bonds. The benchmark holds 60% equities and 40% bonds. Equities returned 10% and bonds returned 5%. The allocation effect is positive because the portfolio overweighted the better-performing asset class. Now within equities, if the portfolio's chosen stocks returned 9% while the equity benchmark returned 10%, the selection effect is negative. The attribution splits the total active return cleanly into these two sources.
A second type of attribution, called factor-based attribution, is used when detailed holdings data is not available, or when the portfolio's returns are better explained by systematic exposures such as value versus growth, small versus large companies, or credit quality. The portfolio's active returns are statistically regressed against these factors. The result reveals whether the manager's alpha comes from genuine stock selection or simply from loading up on known risk factors.
How Long Does It Take to Know If a Manager Is Skilled?
The uncomfortable truth about track records: It takes far longer than most people assume to statistically confirm that a manager is genuinely skilled rather than just lucky. The noise in markets is that large relative to the signal.
Using standard statistical tests, a portfolio earning 10% per year with 20% annual volatility needs at least 16 years of data before you can be 95% confident the true expected return is positive. This is longer than most investment mandates last. It is longer than many clients stay with one manager.
The picture is better for measuring active returns. A manager with a consistent Information Ratio of 1 (meaning active return equals tracking error) needs only 4 years of data to reach the same 95% confidence. This is why a sustained high Information Ratio is so valuable: it proves skill faster.
Ms T manages a fund with 2% active return and 2% tracking error (an Information Ratio of 1). Her wealth manager calculates: t-statistic equals (2% divided by 2%) multiplied by the square root of N years. Setting this equal to 2 and squaring both sides gives N = 4. Four years of consistently high Information Ratio is enough to be confident the performance is real, not luck.
Monitoring should be continuous, not occasional. A statistical tool called CUSUM (Cumulative Sum control chart) detects accumulated underperformance faster than any other method for a given rate of false alarms. Research shows it can identify a genuinely underperforming manager in about 3.5 years rather than 16 years. When underperformance is flagged, the investigation needs to answer one key question: was the underperformance caused by a flaw in the investment process, or by a deviation from a sound process? Each answer implies a different course of action.