Mean Return

Mean return is an average of a defined set of investment returns. The result depends on which average is used: the arithmetic mean summarizes typical one-period observations, while the geometric mean measures the constant compound rate that links beginning and ending wealth across periods.

Key Takeaways

• “Average return” is incomplete unless the arithmetic, geometric, weighted, or another mean is identified.
• The arithmetic mean adds periodic returns and divides by the number of observations.
• The geometric mean compounds periodic wealth relatives and is generally the appropriate summary of realized multi-period growth.
• A probability-weighted mean is an expected return only when the weights are valid probabilities for future outcomes.
• Sample period, return frequency, missing data, outliers, fees, income, and currency can materially change the mean.
• Mean return does not show volatility, drawdown, skewness, liquidity, or the sequence of returns.
• A historical mean is not a guaranteed or necessarily unbiased forecast of future return.

Arithmetic Mean Return

For (n) periodic returns:

Suppose an investment has returns of 10%, -5%, and 20% over three years. Its arithmetic mean is:

The arithmetic mean answers: “What was the average one-year observation in this sample?” It does not state the compound annual growth of money invested through all three years.

Worked Example: Geometric Mean Return

The geometric mean return is:

Using the same returns:

The three-period cumulative return is:

An initial value of $100 grows to $125.40. A constant return of approximately 7.84% per year would reach the same ending value over three years.

Arithmetic Versus Geometric Mean

The geometric mean is less than or equal to the arithmetic mean for a valid nonconstant return sequence. They are equal when every periodic return is identical.

Why Volatility Creates a Gap

Assume a two-period investment gains 50% and then loses 50%.

The arithmetic mean is zero:

But wealth falls from $100 to $75, so the cumulative return is -25% and the geometric mean is approximately -13.40% per period:

The loss requires a gain larger than 50% to recover because the second percentage applies to a different capital base. This is why arithmetic average return should not be used as a substitute for compound performance.

Weighted Mean Return

When observations or assets have different weights:

If normalized portfolio weights sum to one, expected portfolio return is:

For 70% in an asset with expected return of 6% and 30% in an asset with expected return of 2%, the weighted expected return is 4.8%.

This is a cross-sectional portfolio weighting, not a time-series geometric return. Portfolio risk also depends on variances and covariances, not just weighted mean returns.

Mean Return Versus Expected Return

Expected return is a probability-weighted mean of a future return distribution:

A historical arithmetic mean gives each observed period equal weight unless the method states otherwise. It becomes an expected-return estimate only after an analyst makes the additional assumption that the sample is informative about the future.

Mean return and expected return can have the same numerical formula in a specific model, but the labels are not universally interchangeable.

Mean Versus Median Return

The median is the middle observation after returns are sorted. The mean uses every observation and can be pulled toward an extreme result.

For returns of -10%, 2%, 3%, 4%, and 51%:

• arithmetic mean: 10%
• median: 3%

The mean is mathematically correct, but 10% does not describe a typical observation in this skewed sample. Reporting the median, range, percentiles, and distribution can provide necessary context.

Return Frequency and Annualization

Monthly, quarterly, and annual means are not directly comparable. The return interval must be stated.

• Multiplying an arithmetic monthly mean by 12 produces an arithmetic annual expectation only under simplifying assumptions; it is not the realized compound annual return.
• Compounding twelve monthly returns and taking the twelfth root produces a geometric monthly summary.
• Annualized return should use the compound wealth relationship when summarizing historical multi-period performance.

Changing from daily to monthly observations can also change the sample mean because of missing values, nontrading days, stale prices, and the way distributions are assigned.

Total, Gross, Net, Nominal, and Real Inputs

The averaging method cannot repair inconsistent source returns.

Before calculating a mean, decide whether each observation is:

• price return or total return
• gross or net of investment fees
• nominal or adjusted for inflation
• pre-tax or after-tax
• hedged or unhedged for currency exposure
• measured in the same currency and over the same interval

A mean of price-only returns should not be compared with a benchmark mean that assumes reinvested distributions.

Sample Design and Data Quality

Historical mean return depends on the chosen evidence:

1. Start and end dates: A different market cycle can materially change the result.
2. Observation frequency: Daily, monthly, and annual samples have different noise and compounding properties.
3. Survivorship: A data set containing only surviving securities can overstate historical experience.
4. Corporate actions: Splits, dividends, spin-offs, and mergers require consistent adjustment.
5. Stale or estimated prices: Illiquid holdings may understate observed volatility and distort returns.
6. Outliers: Extreme results may be genuine; deleting them requires a documented reason.
7. Changing exposure: A long sample may combine different leverage, strategy, sector, or duration profiles.
8. Fees and implementation: Index returns and investable portfolio returns can differ.

Longer samples contain more observations but can span obsolete regimes. Shorter samples may be more current but less stable. There is no universally correct window.

What Mean Return Does Not Show

• dispersion around the mean
• maximum loss or drawdown
• skewness and tail severity
• sequence of gains and losses
• liquidity and transaction capacity
• credit or counterparty risk
• leverage and path-dependent liquidation risk
• whether the result came from broad market exposure or security selection

Review mean return with standard deviation, drawdown, downside scenarios, and an appropriate benchmark.

Sources and Calculation Checks

• NIST defines the arithmetic mean and its calculation for a sample.
• NIST documents the geometric mean, which underlies compound return calculations.
• OpenStax demonstrates average portfolio returns and standard deviation.
• Investor.gov’s Performance Claims bulletin explains why methodology, periods, dividends, taxes, market conditions, and backtesting labels matter.

Common Mistakes and Limitations

• Saying “average return” without naming the averaging method.
• Using the arithmetic mean as a compound growth rate.
• Averaging percentages calculated from incompatible denominators.
• Mixing daily, monthly, and annual observations.
• Mixing price returns with total returns.
• Ignoring fees, inflation, taxes, and currency.
• Removing unfavorable outliers without a rule established before reviewing results.
• Treating a historical mean as a guaranteed future return.
• Comparing means from different sample periods or benchmark methodologies.
• Reporting mean without dispersion, downside, or observation count.

Related Terms

• Expected Return: Probability-weighted average of modeled future returns.
• Annualized Return: Equivalent compound yearly rate derived from a cumulative return.
• Total Return: Price change plus income over a defined measurement period.
• Standard Deviation: Dispersion measure that adds context to a mean return.
• Portfolio Variance: Portfolio dispersion measure incorporating weights and covariances.
• Compound Annual Growth Rate: Endpoint compound growth measure related to geometric mean return.

FAQs

Educational Use

This article provides general financial education. Historical and modeled averages do not guarantee future performance or constitute personalized investment, tax, legal, statistical, or portfolio advice.