Law of Total Expectation

E[X] = E[E[X|Y]]: decompose an expectation by conditioning on a variable that splits the problem into tractable cases.

1. The idea

The law of total expectation states $E [X] = E [E [X ∣ Y]]$ : compute $X$ 's conditional mean within each value (or scenario) of $Y$ , then average those means weighted by $Y$ 's distribution.

This is often the fastest way to compute an unconditional mean when $X$ 's behavior differs across natural regimes — market states, customer segments, or successive stages of a process.

It generalizes the ordinary weighted-average intuition: $E [X] = \sum_{y} P (Y = y) E [X ∣ Y = y]$ for discrete $Y$ , or an integral for continuous $Y$ .

2. Why it matters

This is the standard decomposition behind mixture models, regime-based pricing, and any 'expected value given a scenario' interview question.

3. Core mathematics

Law of total expectation

E [X] = E [E [X ∣ Y]] = y \sum P (Y = y) E [X ∣ Y = y]

4. How to recognize it

Think of this when…

The population or process naturally splits into distinct groups or regimes
Each group has an easy-to-state conditional mean
You are given conditional means and group probabilities and asked for the overall mean

5. Worked example

Worked example

A stock is in a bull regime 60% of the time (expected daily return +0.5%) and a bear regime 40% of the time (expected daily return -0.8%). Find the unconditional expected daily return.

1
Let $R$ be daily return and $Y$ be the regime.
$E [R] = P (bull) E [R ∣ bull] + P (bear) E [R ∣ bear]$
2
Substitute the given values.
$E [R] = 0.6 (0.005) + 0.4 (- 0.008)$
3
Compute.
$E [R] = 0.003 - 0.0032 = - 0.0002$
Answer: $E [R] = - 0.02%$ per day

6. Common mistakes

Averaging the conditional means without weighting by the probability of each condition
Confusing $E [X ∣ Y]$ (a random variable) with $E [X ∣ Y = y]$ (a fixed number for a specific $y$ )
Applying the law when the conditioning variable doesn't actually simplify the calculation

7. Quick check

60% of customers are 'high value' with expected spend $200, 40% are 'low value' with expected spend $50. What is the overall expected spend?

8. Practice

1 QuantSolve questions use this concept.

Practice this concept

9. Related methods

Conditional Expectation

Create a free account to save lesson progress.