Standard Deviation as a Measure of Return Risk

Why Standard Deviation Is Used for Return Risk

A return variance measures how widely possible returns are dispersed around their expected return. Because the calculation squares deviations, its units are squared: if returns are expressed in percentage points, variance is in squared percentage points. That can make variance less intuitive to interpret directly. Standard deviation is the square root of variance, so it returns the measure to the same units as the returns themselves.

This tutorial treats a one-period return as a random variable with a discrete distribution. The scenarios are assumed to be mutually exclusive and exhaustive, their probabilities sum to 1, and all returns refer to the same investment period and measurement convention. No normal-distribution assumption is needed. These definitions and calculations are aligned with the SOA VEE Accounting and Finance syllabus.

Definition: For a return random variable $R$ with expected return $E[R]$, its standard deviation is the nonnegative square root of its variance:

\sigma_R=\sqrt{\operatorname{Var}(R)}

For discrete outcomes $r_j$ with probabilities $p_j$, where $\sum_j p_j=1$, the variance and standard deviation are

\operatorname{Var}(R)=\sum_j p_j\bigl(r_j-E[R]\bigr)^2, \qquad \sigma_R=\sqrt{\sum_j p_j\bigl(r_j-E[R]\bigr)^2}.

When returns are entered as decimals, such as $0.08$ for 8%, the variance is in squared decimal-return units and the standard deviation is a decimal return. When returns are entered as percentage points, such as 8 for 8%, the variance is in squared percentage points and the standard deviation is in percentage points. Either convention works, but the inputs and reported units must be consistent.

Why Taking the Square Root Gives a Useful Measure

The variance calculation weights each squared deviation by its probability. Squaring makes positive and negative deviations both contribute positively, and gives greater weight to deviations farther from the mean. The square root does not undo that weighting; it changes the final unit back to the original return unit. This is why standard deviation is easier to compare with an expected return stated as a percentage.

The unit conversion can be seen directly. If a return deviation is measured in percentage points, its square is measured in percentage-points squared. Taking the square root of the variance therefore gives a number in percentage points. More generally, multiplying a return variable by a constant $c$ multiplies its variance by $c^2$, and its standard deviation by $|c|$:

\operatorname{Var}(cR)=c^2\operatorname{Var}(R), \qquad \sigma_{cR}=|c|\sigma_R.

The absolute value is necessary because a standard deviation is nonnegative, even if the scale factor is negative. For a fixed initial investment $V_0$, a dollar gain or loss is $V_0R$. Since $V_0$ is positive, the standard deviation of the dollar gain or loss is $V_0\sigma_R$. The final worked example applies this scaling.

Calculating and Interpreting Standard Deviation

A reliable approach is to establish the expected return first, calculate the probability-weighted squared deviations from that mean, and only then take the square root. Keep the variance unrounded until the final step. The following examples use hypothetical investments and discrete return scenarios.

Worked Example: Converting Variance to Percentage-Point Standard Deviation

A hypothetical investment has a one-year return of 2% with probability 0.25, 8% with probability 0.50, and 14% with probability 0.25. Calculate its expected return, variance, and standard deviation. Interpret the units. The scenario probabilities are the stated assumptions; no distribution beyond these three outcomes is assumed.

Method and probability check. The probabilities sum to $0.25+0.50+0.25=1.00$, so they describe an exhaustive distribution. Use percentage points throughout the variance calculation.

Expected return. The probability-weighted mean is $0.25(2)+0.50(8)+0.25(14)=0.50+4.00+3.50=8.00\%$.

Variance. The deviations from 8% are $-6$, $0$, and $6$ percentage points. Their squares are $36$, $0$, and $36$. Therefore, $0.25(36)+0.50(0)+0.25(36)=9+0+9=18$ percentage-points squared.

Standard deviation. Take the nonnegative square root: $\sigma_R=\sqrt{18}=4.242640\ldots$ percentage points. Rounded only at the final step to four decimal places, the standard deviation is 4.2426 percentage points, or about 4.24%.

Check and interpretation. The deviations sum to zero when probability-weighted: $0.25(-6)+0.50(0)+0.25(6)=0$, as expected for deviations from the mean. The standard deviation has return units, while the variance has squared return units. It describes dispersion around the 8% expected return; it does not assert that every possible return is within 4.2426 percentage points of 8%.

Common mistake avoided. Reporting 18 as the standard deviation confuses variance with standard deviation. The correction is to take $\sqrt{18}$; the result is about 4.2426 percentage points, not 18 percentage points.

Worked Example: Keeping Decimal and Percentage Units Consistent

A hypothetical fund has a one-period return of $-10\%$ with probability 0.20, $5\%$ with probability 0.50, and $15\%$ with probability 0.30. Find its expected return and standard deviation, first using percentage points and then checking the result using decimal returns. Assume the three scenarios are the full distribution for the period.

Method and probability check. The probabilities sum to $0.20+0.50+0.30=1.00$. Use one unit convention at a time so that the variance and standard deviation retain clear units.

Expected return in percentage points. The mean is $0.20(-10)+0.50(5)+0.30(15)=-2+2.5+4.5=5\%$.

Variance in percentage-point units. Relative to the 5% mean, the deviations are $-15$, $0$, and $10$ percentage points. Thus, $0.20(225)+0.50(0)+0.30(100)=45+0+30=75$ percentage-points squared.

Standard deviation. The square root is $\sqrt{75}=8.660254\ldots$ percentage points, or 8.6603 percentage points when rounded to four decimal places.

Decimal-unit check. The returns are $-0.10$, $0.05$, and $0.15$, with mean $0.05$. The deviations are $-0.15$, $0$, and $0.10$. The variance is $0.20(0.15^2)+0.50(0^2)+0.30(0.10^2)=0.0045+0+0.0030=0.0075$. Its square root is $\sqrt{0.0075}=0.08660254\ldots$, which is a decimal return equal to 8.660254…%.

Check and interpretation. The decimal-unit standard deviation agrees with the percentage-point result after converting units. It is larger than the 5% expected return, but that comparison alone does not determine whether the distribution is acceptable for a particular investor. Standard deviation summarizes dispersion; it does not describe every feature of the distribution.

Common mistake avoided. The decimal variance $0.0075$ is not $0.0075\%$. Taking its square root gives $0.0866$ as a decimal return, which must be expressed as about 8.6603%, not 0.0866%.

Standard Deviation Is Not a Guaranteed Range

A frequent misinterpretation is to treat the mean plus or minus one standard deviation as a range that must contain every possible return. That conclusion does not follow from the definition. Standard deviation is a summary of dispersion around the mean, not a lower or upper bound on outcomes.

In the first worked example, the mean is 8% and the standard deviation is about 4.2426 percentage points. The interval from the mean minus one standard deviation to the mean plus one standard deviation is approximately 3.7574% to 12.2426%. Yet the stated distribution includes returns of 2% and 14%, both outside that interval. The calculation is internally consistent: a standard deviation does not guarantee that all outcomes lie within one standard deviation of the mean.

Nor does standard deviation say whether unusually low returns are more likely than unusually high returns. Squared deviations count distance from the mean without keeping the direction. Two distributions can have the same standard deviation but different probabilities of losses, different extremes, or different shapes. Use the scenario probabilities and return outcomes when the question asks about a particular event, such as the probability of a negative return.

Key takeaway: Standard deviation is the square root of return variance and is expressed in the same units as returns. It measures dispersion around the expected return, but it is neither a guaranteed outcome range nor a probability statement by itself.

Scaling Return Risk into Dollars

For a fixed positive investment amount, dollar gains and losses scale directly with the return. The same factor therefore scales standard deviation. This translation can help make units concrete, but it does not change the underlying return distribution or guarantee a particular dollar outcome.

Worked Example: Converting Return Standard Deviation to Dollars

An investor places $40,000 in a hypothetical one-period investment. Its return is 0% with probability 0.25, 4% with probability 0.50, and 10% with probability 0.25. Calculate the expected return, return standard deviation, and standard deviation of the dollar gain or loss. Assume the initial amount is fixed and the listed outcomes exhaust the distribution.

Method and probability check. The probabilities sum to $0.25+0.50+0.25=1.00$. First calculate the return statistics in percentage points, then multiply the return standard deviation by the initial investment.

Expected return. The probability-weighted mean is $0.25(0)+0.50(4)+0.25(10)=0+2+2.5=4.5\%$.

Return variance. The deviations from 4.5% are $-4.5$, $-0.5$, and $5.5$ percentage points. The weighted squared deviations are $0.25(20.25)=5.0625$, $0.50(0.25)=0.125$, and $0.25(30.25)=7.5625$. Their sum is $5.0625+0.125+7.5625=12.75$ percentage-points squared.

Return standard deviation. Taking the square root gives $\sqrt{12.75}=3.570714\ldots$ percentage points, or 3.5707 percentage points to four decimal places. As a decimal return, this is $0.03570714\ldots$.

Dollar standard deviation. The dollar gain or loss equals $40,000 times the return, so its standard deviation is $$40{,}000(0.03570714\ldots)=$1,428.2856\ldots$. Rounded to the nearest cent, the dollar standard deviation is $1,428.29.

Check using the dollar outcomes. The possible dollar gains are $0, $1,600, and $4,000. Their expected value is $0.25(0)+0.50(1{,}600)+0.25(4{,}000)=800+1{,}000=$1{,}800$. Deviations from $1,800 are $-$1{,}800$, $-$200$, and $2,200. The probability-weighted squared deviations are $0.25(3{,}240{,}000)+0.50(40{,}000)+0.25(4{,}840{,}000)=810{,}000+20{,}000+1{,}210{,}000=2{,}040{,}000$ dollars squared. Their square root is $1,428.2856…, matching the scaled result and rounding to $1,428.29.

Common mistake avoided. Multiplying the return variance by $40,000 would give the wrong scale and units. Variance scales by the square of the investment amount, while standard deviation scales by the investment amount itself. Here the standard deviation is $0.03570714\ldots\times$40,000$, not $0.1275\times$40,000$ or $12.75\times$40,000$.

Exam Traps to Check Before Reporting an Answer

Stopping at variance. If asked for standard deviation, take the square root of the variance. Keep the variance and standard deviation labels distinct.
Mixing decimal and percentage inputs. If the returns used in the variance calculation are decimals, report the square root as a decimal return or convert it to a percentage. Do not attach a percent sign to the decimal variance.
Dropping the unit. A variance in percentage-point units is in squared percentage points; its square root is in percentage points. A dollar standard deviation is in dollars.
Rounding too early. Carry the variance and square root at full available precision, then round the final standard deviation as instructed.
Calling the result a guaranteed range. Standard deviation does not set a maximum possible deviation and does not, by itself, specify the probability of a return being in any interval.

A useful final check is to ask whether the reported standard deviation has the same type of unit as the return. If a return is stated in percent, its standard deviation should also be stated in percent or percentage points, not squared percent. If the result is in dollars, the calculation should have scaled the return standard deviation by a dollar amount.

Check Your Understanding

For each hypothetical discrete return distribution, calculate the expected return and standard deviation. Use percentage points for the return calculations unless a question asks for a dollar result. Round standard deviations to four decimal places where needed.

A one-period return is 3% or 11%, each with probability 0.50. Find the expected return and standard deviation. (Answer: Expected return 7%; standard deviation 4 percentage points.)
A fund has returns of $-5\%$, 5%, and 15% with probabilities 0.20, 0.50, and 0.30. Find the expected return and standard deviation. (Answer: Expected return 6%; variance 49 percentage-points squared; standard deviation 7 percentage points.)
A hypothetical investment has returns of 0%, 8%, and 20% with probabilities 0.25, 0.50, and 0.25. Find the mean and standard deviation, and identify whether all outcomes lie within one standard deviation of the mean. (Answer: Mean 9%; standard deviation $\sqrt{51}=7.1414$ percentage points; no, the 0% and 20% outcomes lie outside the interval.)
A fixed $60,000 investment has returns of $-2\%$, 4%, and 10% with probabilities 0.25, 0.50, and 0.25. Find the expected return, return standard deviation, and standard deviation of dollar gains or losses. (Answer: Expected return 4%; return standard deviation $\sqrt{18}=4.2426$ percentage points; dollar standard deviation $2,545.58.)
Investment A has returns of 2% and 10%, each with probability 0.50. Investment B has returns of $-2\%$, 6%, and 14% with probabilities 0.25, 0.50, and 0.25. Calculate both expected returns and standard deviations, then state which has the greater standard deviation. (Answer: Both expected returns are 6%; A has standard deviation 4 percentage points and B has standard deviation $\sqrt{32}=5.6569$ percentage points, so B has the greater standard deviation.)

What You'll Learn