Moments for Linear Combinations of Independent Random Variables

An actuary needs $E[S]$ and $\text{Var}(S)$ for $S = c_1X_1 + \cdots + c_{20}X_{20}$ covering 20 independent risks. Building the full distribution is impractical, the moment formulas for linear combinations provide an exact shortcut.

For any random variables $X_1, \ldots, X_n$ (independent or not) and constants $a_1, \ldots, a_n, b$:

$$E\left[\sum_{i=1}^n a_i X_i + b\right] = \sum_{i=1}^n a_i E[X_i] + b$$

Linearity of expectation holds always, no independence required. This is one of the most powerful tools in probability.

For any random variables $X_1, \ldots, X_n$:

$$\text{Var}\left(\sum_{i=1}^n a_i X_i + b\right) = \sum_{i=1}^n a_i^2 \text{Var}(X_i) + 2\sum_{i < j} a_i a_j \text{Cov}(X_i, X_j)$$

HIGH-FREQUENCY: The constant $b$ does NOT affect variance. The coefficient $a_i$ is SQUARED in the variance terms.

When $X_1, \ldots, X_n$ are independent, all covariance terms vanish:

$$\text{Var}\left(\sum_{i=1}^n a_i X_i\right) = \sum_{i=1}^n a_i^2 \text{Var}(X_i)$$

Common mistakes

• Forgetting to square the coefficient in variance. For $2X_1$, the variance contribution is $4 \cdot \text{Var}(X_1) = 16$, not $2 \cdot 4 = 8$. Trap: 8 appears when you multiply the coefficient by the variance instead of squaring it.
• Including the constant in the variance. $\text{Var}(S + 10) = \text{Var}(S)$. Adding 10 shifts the distribution but does not spread it. Including $10^2 = 100$ in the variance gives $\text{Var}(S) = 98 + 100 = 198$. Trap: 198.
• Not squaring the negative coefficient for differences. $\text{Var}(-X_2) = (-1)^2 \text{Var}(X_2) = \text{Var}(X_2)$, not $-\text{Var}(X_2)$. Variance is always non-negative.

Bottom line

• Mean is linear (always): $E[\sum a_i X_i + b] = \sum a_i E[X_i] + b$, no independence needed
• Variance under independence: $\text{Var}(\sum a_i X_i) = \sum a_i^2 \text{Var}(X_i)$, coefficients SQUARED
• Constants vanish from variance: $\text{Var}(S + b) = \text{Var}(S)$
• Negative coefficients stay positive: $(-a)^2 = a^2$, so every variance contribution is non-negative

Exam shortcut

Set up a variance table with columns: Variable, Coefficient, Coeff Squared, Variance, Product. Fill each row, sum the Product column, then add covariance terms if needed. The variance of a sum must be positive, a negative result means a sign error. "Mean is Linear, Variance is Quadratic, Constants Vanish." Mean: coefficients enter linearly, constant enters additively. Variance: coefficients are squared, constant disappears. "The 1/n rule" for sample means: $\text{Var}(\bar{X}) = \sigma^2/n$. $\text{SD}(\bar{X}) = \sigma/\sqrt{n}$.

The full lesson (about 3,416 words, 23 min read) adds 4 worked examples, all 7 common mistakes, a self-check, free in the app.

Learning objectives

• 3h

Browse all free Exam P lessons or jump into free Exam P practice questions.