Information Ratio: Active Return, Tracking Error, and Reliability

Learn how to calculate and interpret information ratio with matched portfolio and benchmark returns, explicit tracking-error conventions, annualization, fees, and statistical limits.

For educational purposes only; not investment advice. Investing may result in loss.

Direct answer

The information ratio measures average benchmark-relative return per unit of variability in that relative return. In its common active-return form, information ratio = annualized average active return / annualized tracking error. A positive value means average outperformance for the measured sample; a negative value means average underperformance. It is not a probability of skill or a guarantee of future excess return.

The number is meaningful only after defining the portfolio, benchmark, return convention, frequency, dates, currency, fee and tax basis, standard-deviation convention, annualization rule, and whether the inputs are expected or realized. The same manager can have different information ratios against different legitimate benchmarks.

How it works

Analyze an information ratio in this order:

Define the decision and benchmark. Use the benchmark specified by the mandate or a defensible investable alternative matched to asset class, style, geography, currency, duration, credit, leverage, and return variant. A broad market index, policy blend, liability benchmark, and peer group answer different questions.
Build synchronized total-return series. Match valuation dates, time zones, frequency, currency conversion, distributions, withholding, hedging, external cash-flow treatment, and gross-of-fee or net-of-fee basis. Do not subtract a price index from a portfolio total return or mix stale monthly marks with live daily benchmark prices.
Choose the relative-return convention. A common arithmetic series is active return_t = portfolio return_t - benchmark return_t. Some standards or mandates use geometric relative return, geometric relative return_t = (1 + portfolio return_t) / (1 + benchmark return_t) - 1. State the convention and do not combine one numerator with the other convention’s risk series.
Estimate average active return and tracking error. For n observations, mean active return = sum of active returns / n. Ex-post tracking error is the standard deviation of the same relative-return observations. A sample estimate is sample tracking error = sqrt(sum((active return_t - mean active return)^2) / (n - 1)); a population divisor of n gives a different result and must be labeled.
Annualize consistently. With P non-overlapping observations per year and an iid-style approximation, annualized average active return = periodic mean active return × P and annualized tracking error = periodic tracking error × sqrt(P). Therefore, annualized information ratio = periodic mean active return / periodic tracking error × sqrt(P). Serial correlation, smoothing, overlapping periods, changing exposures, or irregular dates can make square-root annualization misleading.
Separate ex-post measurement from ex-ante forecasts. A realized ratio uses historical active returns and realized tracking error. An expected ratio uses forecast active return and a forward-looking covariance or risk model. Model horizon, factor definitions, holdings date, and constraints must match; an ex-ante ratio is not directly comparable with an ex-post ratio merely because both are annualized.
Test reliability and economic usefulness. Show observation count, start and end dates, rolling windows, subperiods, benchmark changes, outliers, drawdowns, skew, tail loss, turnover, capacity, fees, taxes, and exposures. If tracking error is 0, the ratio is undefined, not infinite. A high ratio from a short, smoothed, backfilled, or selected history is weak evidence.

Information ratio differs from Sharpe ratio: Sharpe compares excess return over a cash or risk-free reference with total portfolio volatility, while information ratio compares benchmark-relative return with benchmark-relative volatility. It also differs from Jensen’s alpha and a residual-return ratio derived from a regression. Confirm which definition a vendor uses before comparing numbers.

For a constant frequency, using periodic units in both numerator and denominator without annualizing produces a periodic ratio; multiplying it by sqrt(P) produces the common annualized ratio. Compounded portfolio outperformance over a horizon is a separate wealth result and need not equal arithmetic mean active return times the number of periods.

Example

Assume monthly arithmetic active returns, in percentage points, are 0.80%, -0.40%, 0.50%, 0.10%, -0.20%, 0.70%, -0.10%, 0.40%, 0.00%, 0.60%, -0.30%, 0.30%. Portfolio and benchmark returns are matched total returns, net of the same stated fee basis:

Mean: the 12 observations sum to 2.40%, so monthly mean active return is 2.40% / 12 = 0.2000% and the arithmetic annualized numerator is 0.2000% × 12 = 2.4000%.
Tracking error: squared deviations from the monthly mean sum to 1.8200 percentage-points squared. Using the sample divisor, monthly tracking error is sqrt(1.8200 / 11) = 0.4068%; annualized tracking error is 0.4068% × sqrt(12) = 1.4091%. Using divisor 12 instead would give 0.3894% monthly, demonstrating why the convention matters.
Information ratio: the annualized result is 2.4000% / 1.4091% = 1.7032, equivalently 0.2000% / 0.4068% × sqrt(12) = 1.7032. This is only one year of data, so reporting four decimal places does not imply that much statistical certainty.
Fees and comparison: if a gross active-return estimate is 1.8000% annually and a consistently charged fee reduces it by 0.8000 percentage points while tracking error remains illustratively 1.4091%, gross and net information ratios are 1.8000% / 1.4091% = 1.2774 and 1.0000% / 1.4091% = 0.7097. Another strategy with 3.0000% active return and 12.0000% tracking error has 3.0000% / 12.0000% = 0.2500, more raw active return but less per unit of active risk.

Risks

Identify the portfolio, composite, share class, account, and mandate being measured.
Record benchmark name, identifier, provider, methodology version, and any policy-blend weights.
Match asset class, style, geography, currency, duration, credit quality, leverage, and investability.
Use synchronized portfolio and benchmark valuation dates, times, and non-overlapping return intervals.
Match price, gross total-return, net total-return, hedged, and currency conventions.
State arithmetic-difference or geometric-relative-return methodology and use it consistently.
State gross-of-fee, net-of-fee, transaction-cost, withholding, and tax treatment.
Use time-weighted returns for manager comparison unless a different objective explicitly requires another measure.
Reconcile external cash flows, valuations, stale prices, estimated prices, and subsequent revisions.
State whether tracking error uses sample or population standard deviation and how missing data are handled.
State frequency, observations per year, annualization rule, and number of observations.
Do not use square-root annualization blindly with autocorrelation, smoothing, overlapping data, or irregular dates.
Separate ex-post realized ratios from ex-ante model forecasts and match their horizons.
Treat zero or near-zero tracking error as undefined or numerically unstable rather than exceptional skill.
Show rolling windows and subperiods; one favorable start or end date can dominate a full-period ratio.
Investigate outliers, benchmark changes, strategy drift, backfills, survivorship, and selected reporting histories.
Review drawdown, downside asymmetry, skew, tail loss, liquidity, leverage, derivatives, and concentration.
Attribute active return to security, sector, factor, currency, timing, and residual sources rather than calling the ratio skill.
Compare turnover, capacity, trading costs, fees, taxes, and implementation feasibility with the reported benefit.
Do not rank unlike strategies by information ratio without harmonizing all definitions and data.

Common misconceptions

“An information ratio of 1 means a 100% chance of beating the benchmark.” It is a ratio of estimated mean active return to active-return volatility, not a success probability.
“Tracking error measures how far the portfolio return was from the benchmark once.” It measures variability across a series of relative returns; one-period active return is not tracking error.
“Information ratio and Sharpe ratio are interchangeable.” Their reference returns and risk denominators differ.
“Annualizing always makes daily, monthly, and quarterly estimates comparable.” The usual scaling relies on frequency and dependence assumptions that may fail.
“The highest historical ratio identifies the most skilled manager.” Benchmark fit, sampling error, smoothing, exposures, costs, selection, and luck can produce the ranking.

Sources

CFA Institute: Investment Performance Measurement - The Sharpe Ratio and the Information Ratio.
CFA Institute: GIPS Standards Handbook for Fiduciary Management Providers.
CFA Institute: Analysis of Active Portfolio Management.
FINRA: Get Off the Bench - A Look at Benchmarks.
Investor.gov: How to Read a Mutual Fund or ETF Shareholder Report.
FINRA: Calculating Your Investment Returns.