Mean Median and Mode

Three summaries, three different questions

The arithmetic mean answers: what is the total divided evenly across the observations? The median answers: what is the middle observation after sorting the data? The mode answers: which value occurs most often? All three can describe the center of a data set, but they respond differently to extreme values and repeated observations. A reader should know which statistic was calculated before interpreting a reported average return, income, or fee.

For the observations 2, 4, 5, 6, and 8, the mean is (2 + 4 + 5 + 6 + 8) ÷ 5 = 5. The median is also 5, the third value after sorting. There is no mode because each value occurs once. When the data are symmetric and have no unusual concentration, these measures can be close, but they are not interchangeable definitions.

Now add an extreme value of 40. The observations are 2, 4, 5, 6, 8, and 40. The mean rises to 65 ÷ 6, about 10.83. The median is the average of the third and fourth observations, (5 + 6) ÷ 2 = 5.5. The median changes only slightly, while the mean is pulled upward. The outlier may be a valid event, a data error, or a different population; investigate before choosing a summary.

The median is resistant to extreme values because it depends on rank, not the size of every observation. That makes it useful for skewed data such as household income or home prices. But the median discards information about how far observations lie from the center. Two data sets can have the same median and very different dispersion. Pair a central-tendency measure with a description of spread and sample size.

Calculate the mean carefully

The arithmetic mean is the sum of observations divided by the number of observations. If a portfolio earned 4%, 6%, and 8% in three separate periods and each period is being treated as an equally weighted observation, the arithmetic mean is 6%. This describes the average of the period returns, not necessarily the compound annual growth rate of the investment.

A weighted mean assigns different importance to observations. Multiply each value by its weight, add the products, and divide by the sum of weights if the weights are not already proportions that total one. If an account has 60% in an asset with expected return 7% and 40% in another with expected return 3%, the weighted expected return is 0.60 × 7% + 0.40 × 3% = 5.4%, before fees and other assumptions.

Weights need a clear meaning. They may represent portfolio allocation, population size, transaction amount, or frequency. A weighted average of manager returns using assets under management answers a different question from an equal-weight average of managers. A data provider’s method can change the result. Read the denominator and weighting convention before comparing published statistics.

The mean is sensitive to outliers and data quality. One unusually large return can raise the average even if most investors experienced modest returns. A data set mixing different time periods, currencies, or account types may not be comparable. Check whether returns include dividends, fees, taxes, and flows. The formula can be correct while the population or inputs make the conclusion misleading.

Find the median in odd and even samples

Sort the observations from smallest to largest. With an odd number of observations, the median is the single middle value. For 3, 4, 8, 10, and 19, the median is 8. The positions matter, not the gaps. If the largest value changes from 19 to 1,000, the median remains 8 as long as the order of the middle observations does not change.

With an even number of observations, the median is typically the average of the two middle values. For 2, 5, 7, and 20, the median is (5 + 7) ÷ 2 = 6. The result does not need to be one of the observed values. Do not select the lower or upper middle value alone unless the question defines a different convention.

In grouped or large data, the median may be estimated from a frequency distribution or percentile. A reported median household income is not necessarily the income of a particular household; it marks the midpoint of a distribution under the source’s method. The population, geography, year, and inflation adjustment still matter. A statistic from one region or year cannot automatically describe an individual client.

Mode and frequency distributions

The mode is the most frequent observation or category. A data set can have one mode, more than one mode, or no meaningful mode. If returns are 3%, 3%, 4%, 6%, and 8%, the mode is 3%. If values are 2, 2, 5, 5, and 9, the data are bimodal, with modes 2 and 5. If all values occur once, there is no mode.

Mode can be useful for categorical data where arithmetic averaging makes no sense. The most common account type, transaction category, or survey response can be identified by frequency. It does not say which category has the highest dollar value or is most important. A category that appears often may represent a small share of total assets.

For continuous data, the exact values may all differ, so analysts often group observations into intervals. The modal class is the interval with the highest frequency. Grouping choices affect the result. Two researchers can choose different bin widths and identify different modal classes. Explain the method and avoid claiming a precise mode when the data are grouped.

Investment examples: averages can tell different stories

A fund reports annual returns of -20%, 10%, and 30%. The arithmetic mean is 6.67%. But $100 invested at the start becomes $80 after the first year, $88 after the second, and $114.40 after the third. The compound annual growth rate is (114.40/100)^(1/3) - 1, about 4.60%. The arithmetic average describes the average of the annual returns; it does not describe the compound wealth path.

A client asks whether 6.67% is the expected annual return. The answer depends on how the figures are used. Historical arithmetic mean may be relevant in some estimates of single-period expected return; geometric mean describes compounded growth over a multi-period history. Neither guarantees future results. Fees, taxes, cash flows, and the data period affect the client’s realized outcome.

A set of five adviser fees is 0.50%, 0.55%, 0.60%, 0.65%, and 2.00%. The mean fee is 0.86%, while the median is 0.60%. The 2.00% observation may represent a different service or account size. If a client wants to know what a typical comparable account pays, the median may be more informative, but only after confirming that the accounts offer comparable services and are measured on the same basis.

A portfolio expected-return estimate should use allocation weights. If weights change, the weighted mean changes even when asset-class assumptions are constant. A portfolio with 80% in a 5% expected-return asset and 20% in a 2% asset has an expected weighted return of 4.4%. If the client shifts to 40% and 60%, it becomes 3.8%. This is a planning estimate, not a promised return or a substitute for risk analysis.

When averages mislead

A mean can hide skew. One year of unusually high investment returns can make an average look attractive even though most years were lower. Report median and range or standard deviation when useful. A long-run mean also says nothing about the sequence of returns, which can matter to an investor making withdrawals.

Combining unlike observations can create a meaningless center. Averages across currencies, regions, account sizes, or investment mandates should use consistent definitions. A median across all clients may be dominated by a different client segment than the one under consideration. Define the population first, then choose a statistic.

The mode may be irrelevant when every value differs or when frequency does not reflect economic significance. A common transaction size is not necessarily the typical dollar exposure if a few large trades dominate total volume. Pair counts with amounts where the question concerns economic impact.

The mean and median do not measure volatility or risk by themselves. A data set with returns 4%, 5%, and 6% has mean 5% and median 5%; a set with -20%, 5%, and 30% has the same mean and median but much wider dispersion. Use a spread measure such as range or standard deviation when variability matters.

Worked case: a client survey

A firm surveys seven clients about monthly savings: $100, $150, $200, $250, $300, $400, and $2,000. The total is $3,400; the mean is about $485.71. The median is $250, the fourth observation. The mean is substantially higher because one client saves $2,000. If the firm wants to estimate typical monthly savings for its ordinary client, the median may describe the middle respondent better, but the sample size and selection method must also be considered.

If the survey sampled clients with high account balances more often, neither unweighted mean nor median may represent the full client population. The analyst may need sampling weights. If the question concerns total dollars saved, the sum matters. If it concerns the central client, median may be useful. If it concerns average resources per sampled account, the mean may be appropriate. The decision question determines the summary.

Choosing a statistic for an investment question

Suppose an adviser compares the average advisory fee paid by 100 households. If each household should count equally, the arithmetic mean of household fee rates answers one question. If the adviser wants the fee rate on the firm's total assets, the relevant calculation weights each household by assets. A $5 million account paying 0.50% represents more dollars and more assets than a $50,000 account paying the same rate. The unweighted average of rates and the ratio of total fees to total assets can differ.

For a portfolio with $400,000 in a fund charging 0.20% and $100,000 in a fund charging 1.00%, the weighted expense rate is (400,000 × 0.002 + 100,000 × 0.01) ÷ 500,000 = 0.36%. The simple average of the two stated rates is 0.60%, which overstates the portfolio's weighted rate because the more expensive fund has the smaller allocation. Keep rate units consistent: 0.36% is 36 basis points.

The same distinction matters when comparing investment returns. If Fund A managed $10 million and Fund B managed $1 million, an equal-weight average of their returns gives each fund the same influence. An asset-weighted result gives Fund A greater influence. Neither is universally correct; choose based on whether the question concerns a typical fund or the combined dollars invested. State the method so a reader can reproduce the figure.

Historical investment returns also require care with arithmetic versus geometric averages. The arithmetic average is useful for describing the average one-period observation or for certain expected-return estimates. The geometric average reflects compound growth across periods. If an investment rises 50% and then falls 50%, the arithmetic mean is zero, but $100 becomes $75. The compounded two-period return is -25%, and its annualized geometric return is about -13.4% per year.

Common mean, median, and mode mistakes

Do not say the mean is always the typical value. Outliers and skew can pull it away from most observations. Compare with the median and inspect the distribution.

Do not calculate the median before sorting. For an even sample, average the two middle observations under the usual convention.

Do not use an unweighted average when observations represent different portfolio weights or population sizes. Identify what the weights mean and make sure they total correctly.

Do not treat historical arithmetic mean as compound return. Compounding depends on the sequence and multiplicative growth factors.

Do not treat mode as the largest value or most valuable category. It is the most frequent observation or class.

How to solve central-tendency questions

First identify whether the question asks for mean, median, mode, or weighted mean. List values and units. For the mean, add and divide by the number of observations. For a weighted mean, multiply each value by its weight and sum. For the median, sort values and find the middle position. For mode, count frequency.

Then interpret the result in context. Ask whether there are outliers, whether observations are comparable, and what population is represented. If the question concerns portfolio return, distinguish period-average from compound return. If the data are skewed, explain why median may better describe the middle observation without claiming it is always superior.

For investment advice, connect the statistic to the decision. A historical average cannot assure a future return; a fee median cannot compare dissimilar services; and an average client profile cannot replace the actual client’s objectives. Statistics summarize data, while suitability requires individual facts.

Sources

OpenStax Introductory Statistics, measures of center; CFA Institute Research Foundation, measures of risk-adjusted return; NASAA Series 65 Test Specifications.