Quantitative Methods · Portfolio Mathematics · LO 1 of 3

When two risky assets walk into a portfolio together, does the portfolio become more dangerous or less?

The answer lives in how their returns move together, not just how risky each one is alone.

⏱ 8min-15min

6 questions

HIGH PRIORITYAPPLY🧮 Calculator

Why this LO matters

The answer lives in how their returns move together, not just how risky each one is alone.

INSIGHT

When returns move together in the same direction around their means, they add to portfolio risk. When they move oppositely, they cancel out and reduce it. Covariance tells you direction, not strength in absolute terms. A covariance of 0.0023 sounds tiny, but it is positive, meaning both assets tend to be above their own expected values at the same time, or both below at the same time. That co-movement still adds risk to the portfolio even though the number looks small. Correlation standardises this by dividing by the product of the two standard deviations, producing a number bounded between −1 and +1 that is far easier to interpret. The portfolio variance formula is just a mechanical way of adding up: each asset's own risk (its weighted variance), plus the co-movement risk between assets (weighted covariances). That co-movement term is where diversification does its work.

How does mixing two risky assets affect the portfolio's risk?

When you hold two investments together, your portfolio's risk is not simply the sum of their individual risks. The two assets may move together, opposite each other, or largely independently. That relationship, how one asset's returns behave when the other's return is above or below its average, is what determines whether combining them makes your portfolio safer or riskier.

The CFA curriculum calls this relationship covariance. The curriculum tests three things about it: calculating it, interpreting its sign, and then using it inside the portfolio variance formula. You also need to understand correlation coefficient, which is covariance standardized so it is easier to read and compare.

How do you find a portfolio's expected return?

Before we tackle risk, we need the expected return. This part is straightforward, it is just a weighted average.

Expected Portfolio Return

E(Rp) = w₁E(R₁) + w₂E(R₂) + ... + wₙE(Rₙ)

w = weight of each asset in the portfolio (must sum to 1)
E(Rₙ) = expected return of each asset

Use when: calculating the weighted average expected return of a portfolio.
Do not use: when assets have different expected returns and you need risk-adjusted comparison.
// The portfolio's expected return is just each asset's expected return multiplied by how much of the portfolio it makes up, added together.

Worked Example 1

Expected portfolio return from two fund weights

Yuna Park, a financial planner in Seoul, is building a balanced savings portfolio for a client. She allocates 40% of the portfolio to a domestic equity fund with an expected return of 4%, and 60% to a growth equity fund with an expected return of 8%. She wants to know the expected return on the combined portfolio.

🧠Thinking Flow — weighted average of expected returns

The question asks

What is the weighted average expected return of the two funds combined?

Key concept needed

Portfolio expected return, the weighted sum of each asset's expected return.

Step 1, Identify known values

w₁ = 0.40, E(R₁) = 4%, w₂ = 0.60, E(R₂) = 8%. The weights sum to 1.00. ✓

Step 2, Apply the expected return formula

E(Rp) = w₁ × E(R₁) + w₂ × E(R₂) E(Rp) = 0.40 × 4% + 0.60 × 8% E(Rp) = 1.60% + 4.80% = 6.40%

Step 3, Verify the result

The portfolio return must lie between the two individual returns (4% and 8%). 6.4% sits between them. ✓ ✓ Answer: 6.4%. This is simply the weighted average of the two funds' expected returns.

How does covariance describe whether two assets move together or apart?

Here is where most students go wrong. They assume that if two assets both carry risk, their combined risk must be the sum of those risks. That is only true if the assets move perfectly in sync, and most assets do not.

Covariance measures whether two assets tend to be above or below their expected values at the same time. That is the key to understanding diversification.

What covariance tells you about two assets

Positive covariance. When one asset's return is above its expected value, the other tends to also be above its expected value. The two assets generally move in the same direction. This does NOT automatically mean high risk, it means the returns are related.

Negative covariance. When one asset's return is above its expected value, the other tends to be below its expected value. The two assets tend to offset each other. This can reduce portfolio risk dramatically.

Zero covariance. No predictable relationship. Knowing one asset's return tells you nothing about the other's. Each behaves independently of the other.

Own covariance. Cov(R, R) = Var(R). An asset's covariance with itself is simply its variance.

Two businesses and the weather

Imagine you run two businesses in the same city: a restaurant on a pedestrian street and an ice cream cart in the same square. On sunny days, both do well. On rainy days, both do poorly. Their profits have positive covariance, when one is above average, the other tends also to be above average too. Now imagine you also own a company that sells umbrellas. On rainy days, the umbrella company does well and the restaurant does badly. Those two businesses have negative covariance. If you owned only the restaurant and the ice cream cart, a string of rainy days would hurt both at once. Adding the umbrella company helps, losses on the restaurant can be offset by gains on the umbrella sales. The wrong answer candidates give: "Positive covariance means the portfolio is riskier." The right framework: covariance does not measure risk alone, it measures how two assets move together. A portfolio with positively covarying assets can still be high risk or low risk depending on their individual variances. Covariance describes the relationship, not the absolute level.

How do you calculate covariance?

There are two ways to compute it. The exam will tell you which one to use.

Covariance from correlation and standard deviations

Cov(Rᵢ, Rⱼ) = ρᵢⱼ × σᵢ × σⱼ

ρᵢⱼ = correlation coefficient between assets i and j
σᵢ = standard deviation of asset i's returns
σⱼ = standard deviation of asset j's returns

Use when: you are given correlation and the two standard deviations directly.
Do not use: when you are given the covariance matrix directly, in that case, the covariance entries are already the result, not something to be recalculated.
// Covariance equals the correlation coefficient multiplied by both standard deviations. The correlation tells you the direction and strength of the linear relationship; the standard deviations scale it to the right units.

Covariance from a joint probability function

Cov(Rₐ, Rᵦ) = Σ [P(scenario) × (Rₐ − E(Rₐ)) × (Rᵦ − E(Rᵦ))]

P(scenario) = probability of that combination of returns occurring
Rₐ = actual return on asset A in that scenario
Rᵦ = actual return on asset B in that scenario
E(Rₐ), E(Rᵦ) = expected returns of each asset

Use when: you are given a joint probability table showing all possible return combinations for two assets.
Do not use: when you are given correlation directly, use the simpler ρσᵢσⱼ formula instead.
// For each possible return scenario, multiply the deviation of asset A from its mean by the deviation of asset B from its mean, weight it by how likely that scenario is, and sum across all scenarios.

Worked Example 2

Covariance from correlation and two standard deviations

Marcus Okonkwo holds two bonds in his fixed-income sleeve, a US corporate bond with a standard deviation of 12% and a Spanish government bond with a standard deviation of 25%. His research indicates a correlation of 0.20 between the two bond returns. He needs the covariance to assess the diversification benefit.

🧠Thinking Flow — covariance from correlation and standard deviations

The question asks

What is the covariance between the two bond returns?

Key concept needed

Covariance = correlation × σᵢ × σⱼ.

Step 1, Extract known values

σ₁ = 12%, σ₂ = 25%, ρ = 0.20.

Step 2, Apply the covariance formula

Many candidates first multiply the two standard deviations and ignore the correlation, that gives 12 × 25 = 300. The correct approach includes the correlation coefficient: Cov(R₁, R₂) = ρ × σ₁ × σ₂. Cov = 0.20 × 12 × 25 = 0.20 × 300 = 60.

Step 3, Sanity check

Covariance is always expressed in the same units as the product of the two standard deviations. Here: % × % = %². The result of 60 is in percent-squared, which is correct. ✓ ✓ Answer: Covariance = 60 (%²).

What does the covariance matrix tell you, and how do you read it?

A covariance matrix for n assets contains n² entries. The diagonal entries are variances. The off-diagonal entries are covariances. The matrix is symmetric, Cov(A,B) = Cov(B,A).

Reading a covariance matrix

Diagonal entries (top-left to bottom-right). Each asset's own variance. Var(A,A) = σ²(A). These are NOT covariances, they are variances.

Off-diagonal entries. Covariance between two different assets. Cov(A,B) appears in both positions: row A, column B AND row B, column A. Only one of these is unique.

Number of unique covariance terms. For n assets: n(n−1)/2. The covariance matrix has n² entries, subtract n diagonal variance terms = n²−n off-diagonal entries. Since the matrix is symmetric, divide by 2: (n²−n)/2 = n(n−1)/2.

Using the matrix for portfolio variance. Every entry in the matrix, variances and covariances alike, feeds into the calculation.

Worked Example 3

Reading a covariance matrix and extracting values

An analyst at Meridian Asset Management has the following covariance matrix for a hedge fund and a market index, expressed in percentage-squared:

	Hedge Fund	Market Index
Hedge Fund	256	110
Market Index	110	81

She needs to report both the correlation between the hedge fund and the market index, and how many unique covariance terms exist in a five-asset version of this matrix.

🧠Thinking Flow — extracting covariance from matrix, converting variance to standard deviation

The question asks

What is the correlation coefficient? And for a five-asset matrix, how many unique covariance terms exist?

Key concept needed

Correlation = covariance ÷ (σᵢ × σⱼ). Unique covariances = n(n−1)/2.

Step 1, Extract known values from the matrix

The off-diagonal entry is the covariance: Cov = 110. The diagonal entry for the hedge fund is its variance: Var = 256, so σ₁ = √256 = 16. The diagonal entry for the market index is its variance: Var = 81, so σ₂ = √81 = 9.

Step 2, Apply the correlation formula

ρ = Cov ÷ (σ₁ × σ₂) = 110 ÷ (16 × 9) = 110 ÷ 144 ≈ 0.76.

Step 3, Sanity check

Correlation must lie between −1 and +1. 0.76 is between 0 and +1, plausible. ✓

Step 4, Count unique covariance terms for a five-asset matrix

Total matrix entries: 5 × 5 = 25. Diagonal variance entries: 5. Off-diagonal entries: 25 − 5 = 20. The matrix is symmetric: 20 ÷ 2 = 10 unique covariance terms. ✓ Answer: Correlation ≈ 0.76. For five assets, 10 unique covariance terms exist.

Why do we need correlation if covariance already measures co-movement?

Covariance has a problem: its value depends on the units in which you measure returns. A covariance of 60 when returns are in percent looks very different from a covariance of 0.006 when returns are in decimals, even though they measure the same relationship.

Correlation coefficient fixes this by dividing covariance by the product of both standard deviations. The units cancel out. Correlation always falls between −1 and +1.

Correlation coefficient

ρᵢⱼ = Cov(Rᵢ, Rⱼ) / (σᵢ × σⱼ)

ρᵢⱼ = correlation, bounded between −1 and +1
Cov(Rᵢ, Rⱼ) = covariance between assets i and j
σᵢ, σⱼ = standard deviations of each asset

Use when: comparing the strength of the relationship between two assets regardless of their units or scale.
Do not use: for computing portfolio variance, the formula uses covariance, not correlation directly.
// Correlation is covariance divided by the product of the two standard deviations. The result is a pure number between −1 and +1. +1 means perfect positive linear relationship. −1 means perfect negative linear relationship. 0 means no linear relationship.

Reading the correlation coefficient

+1.0. Perfect positive co-movement. When asset A is above its mean, asset B is always above its mean by a proportional amount. No diversification benefit.

Between 0 and +1. Positive but imperfect relationship. Assets tend to move together. Some diversification benefit, the less correlated they are, the more benefit.

0. No linear relationship. Knowing one asset's return tells you nothing about the other. Maximum diversification potential.

Between −1 and 0. Negative relationship. Assets tend to offset each other. Strong diversification potential.

−1.0. Perfect negative co-movement. When one is above its mean, the other is always below. You can construct a risk-free portfolio in theory. Key rule for exam questions: Correlation measures strength by absolute value, not by sign. A correlation of −0.67 is stronger than a correlation of +0.33 because 0.67 is further from zero. Students often pick −0.67 as "weakest" because they focus on the negative sign instead of the magnitude.

FORWARD REFERENCE

Security Market Line (SML), the covariance between an asset's returns and market returns is called beta in the CAPM context. You will compute this same covariance, just in a different setting. For this LO, you only need to understand what covariance measures and how to calculate it. The SML and beta are fully covered in Equity.

→ Equity

How does covariance actually reduce portfolio risk?

Modern portfolio theory rests on a single insight: as you add assets, covariance terms increasingly dominate portfolio variance. With 20 assets, there are 20 variance terms and 380 covariance terms. The individual variances stop being the main story. The relationships between assets become everything.

This is the diversification benefit: combining assets whose returns do not move perfectly together reduces portfolio risk without reducing expected return.

If covariance = 0: covariance terms contribute nothing. Portfolio variance equals the sum of the weighted individual variances. Risk reduction is maximal.
If covariance is negative: the covariance terms subtract from portfolio variance. Even more risk reduction. A negatively correlated pair actively hedges each other.
If covariance is positive and large: the covariance terms add significantly to portfolio variance. Combining the assets makes the portfolio riskier than the weighted average of individual risks would suggest.

Key rule: Diversification benefit increases as covariance decreases. The lower the correlation, the more risk you can eliminate. Zero correlation gives maximum diversification. Correlation of +1 gives zero diversification benefit.

How do you compute portfolio variance when two assets are combined?

This is the calculation that most exam questions ask for. The portfolio's variance depends not just on each asset's own variance, it depends equally on how the assets covary with each other.

Portfolio variance for two assets

σ²p = w₁²σ₁² + w₂²σ₂² + 2w₁w₂Cov(R₁,R₂)

w₁, w₂ = portfolio weights (must sum to 1)
σ₁², σ₂² = variances of each asset's returns
Cov(R₁,R₂) = covariance between the two assets' returns

Use when: calculating the variance of a portfolio of exactly two assets.
Do not use: for more than two assets, use the n-asset matrix formula.
// Portfolio variance equals the weighted variance of each asset plus a term for how they move together. The 2 in front of the covariance term reflects that Cov(R₁,R₂) = Cov(R₂,R₁), both cross-products appear in the full expansion.

Worked Example 4

Portfolio standard deviation for two assets

Sofia Nakamura manages a balanced portfolio for her firm's pension fund. She holds two stocks with these characteristics:

	Stock 1	Stock 2
Expected return	7%	10%
Standard deviation	12%	25%
Portfolio weight	0.30	0.70
Correlation	0.20

She needs the portfolio's standard deviation of returns.

🧠Thinking Flow — two-asset portfolio variance with given weights, variances, and covariance

The question asks

What is the portfolio standard deviation when two stocks are combined?

Key concept needed

σ²p = w₁²σ₁² + w₂²σ₂² + 2w₁w₂Cov(R₁,R₂).

Step 1, Convert correlation to covariance

Cov(R₁, R₂) = ρ × σ₁ × σ₂ = 0.20 × 12 × 25 = 60.

Step 2, Compute each component of the portfolio variance formula

Many candidates first confuse the formula by inserting the correlation directly instead of the covariance, that gives the wrong cross-term. The formula requires Cov(R₁,R₂), not ρ. Component 1, weighted variance of Stock 1: w₁² × σ₁² = (0.30)² × (12)² = 0.09 × 144 = 12.96 Component 2, weighted variance of Stock 2: w₂² × σ₂² = (0.70)² × (25)² = 0.49 × 625 = 306.25 Component 3, weighted covariance term: 2 × w₁ × w₂ × Cov = 2 × 0.30 × 0.70 × 60 = 2 × 0.21 × 60 = 25.20

Step 3, Sum and take the square root

σ²p = 12.96 + 306.25 + 25.20 = 344.41 σp = √344.41

Step 4, Sanity check

The portfolio standard deviation must be below the weighted average of individual standard deviations (0.30 × 12% + 0.70 × 25% = 21.1%) because partial diversification is occurring, correlation of 0.20 is below +1. 18.56% is below 21.1%. ✓ ✓ Answer: Portfolio standard deviation = 18.56%. Exam answer: B.

🧮 BA II Plus Keystrokes

[2ND][CLRWORK]

Clear any previous calculations → 0

[0][.][3][0][x²][×][1][2][x²][+]

Begin component 1: 0.30² × 12² → 12.96

[0][.][4][9][×][6][2][5][+]

Add component 2: 0.49 × 625 → 319.21

[2][×][0][.][2][1][×][6][0]

Add component 3: 2 × 0.21 × 60 → 344.41

[=]

Confirm total portfolio variance → 344.41

[2ND][√x̄]

Begin square root of 344.41 → 344.41

[=]

Portfolio standard deviation → 18.56

⚠️ Calculator mode check first: Press [2ND][+][8][ENTER] to confirm your calculator is in AOS (Algebraic Operating System) mode. Chain mode produces wrong answers on multi-step arithmetic. The A/CNV setting must show "A" for algebraic. The most common error is skipping the covariance conversion step and inserting the correlation coefficient (0.20) directly into the formula as the third term. This gives 2 × 0.30 × 0.70 × 0.20 = 0.084, producing a portfolio standard deviation far below the correct 18.56, candidates who see 14.91 as an answer choice should recognise this error.

How do you compute portfolio variance for three assets using the covariance matrix?

For a three-asset portfolio, the expanded formula includes three variance terms and three covariance terms, each covariance multiplied by 2 because Cov(A,B) = Cov(B,A).

Portfolio variance for three assets

σ²p = w₁²Var₁ + w₂²Var₂ + w₃²Var₃ + 2w₁w₂Cov₁₂ + 2w₁w₃Cov₁₃ + 2w₂w₃Cov₂₃

All six terms come directly from the covariance matrix.
Diagonal entries (Var₁, Var₂, Var₃) are the three variances.
Off-diagonal entries (Cov₁₂, Cov₁₃, Cov₂₃) are the three distinct covariances.

Portfolio variance for n assets, covariance matrix form

σ²p = Σᵢ Σⱼ wᵢwⱼCov(Rᵢ,Rⱼ)

For i = 1 to n, j = 1 to n. All n² entries of the covariance matrix are used.
Diagonal terms: i=j, so wᵢ²σᵢ² = weight squared times own variance.
Off-diagonal terms: i≠j, each covariance appears twice (i,j and j,i) unless handled explicitly.

Worked Example 5

Portfolio standard deviation for three assets using the covariance matrix

Lena Osei has a three-asset portfolio with weights of 20%, 30%, and 50% for Assets 1, 2, and 3 respectively. She estimates the following covariance matrix (entries in %²):

	Asset 1	Asset 2	Asset 3
Asset 1	196	105	140
Asset 2	105	225	150
Asset 3	140	150	400

What is the portfolio's standard deviation of returns?

🧠Thinking Flow — three-asset portfolio variance from covariance matrix

The question asks

How do you compute portfolio variance when three assets are combined using a covariance matrix?

Key concept needed

σ²p = Σᵢ Σⱼ wᵢwⱼCov(Rᵢ,Rⱼ), every cell of the matrix, weighted by both assets' portfolio weights.

Step 1, Read the matrix correctly

Diagonal entries (variances): Var₁ = 196, Var₂ = 225, Var₃ = 400. Off-diagonal entries (covariances): Cov₁₂ = 105, Cov₁₃ = 140, Cov₂₃ = 150. The matrix is symmetric, Cov(1,2) = Cov(2,1), but the formula handles this automatically.

Step 2, Apply the three-asset portfolio variance formula

Many candidates forget the factor of 2 on the covariance terms. The compact double-sum formula includes both (i,j) and (j,i) pairs, which is why a "2" appears explicitly in the expanded form. σ²p = w₁²Var₁ + w₂²Var₂ + w₃²Var₃ + 2w₁w₂Cov₁₂ + 2w₁w₃Cov₁₃ + 2w₂w₃Cov₂₃ = (0.20)² × 196 + (0.30)² × 225 + (0.50)² × 400 + 2 × 0.20 × 0.30 × 105 + 2 × 0.20 × 0.50 × 140 + 2 × 0.30 × 0.50 × 150 = 0.04 × 196 + 0.09 × 225 + 0.25 × 400 + 2 × 0.06 × 105 + 2 × 0.10 × 140 + 2 × 0.15 × 150 = 7.84 + 20.25 + 100.00 + 12.60 + 28.00 + 45.00 = 213.69

Step 3, Take the square root

σp = √213.69 = 14.62%

Step 4, Sanity check

The answer 14.62% is above the smallest individual standard deviation (√196 = 14.00%) and below the largest (√400 = 20.00%). This is expected for a diversified three-asset portfolio with positive covariances. ✓ ✓ Answer: Portfolio standard deviation = 14.62%. Exam answer: C.

🧮 BA II Plus Keystrokes

[2ND][CLRWORK]

Clear previous calculations → 0

[0][.][0][4][×][1][9][6][=]

Compute w₁²Var₁: 0.04 × 196 → 7.84

[M+]

Store component 1 to memory → 7.84

[0][.][0][9][×][2][2][5][=]

Compute w₂²Var₂: 0.09 × 225 → 20.25

[M+]

Add to memory → 27.09

[0][.][2][5][×][4][0][0][=]

Compute w₃²Var₃: 0.25 × 400 → 100.00

[M+]

Add to memory → 127.09

[2][×][0][.][0][6][×][1][0][5][=]

Compute 2w₁w₂Cov₁₂: 2 × 0.06 × 105 → 12.60

[M+]

Add to memory → 139.69

[2][×][0][.][1][×][1][4][0][=]

Compute 2w₁w₃Cov₁₃: 2 × 0.10 × 140 → 28.00

[M+]

Add to memory → 167.69

[2][×][0][.][1][5][×][1][5][0][=]

Compute 2w₂w₃Cov₂₃: 2 × 0.15 × 150 → 45.00

[M+]

Add to memory → 212.69

[MR]

Recall total variance (add last 1.00) → 213.69

[2ND][√x̄]

Begin square root → 213.69

[=]

Portfolio standard deviation → 14.62

⚠️ A common mistake is omitting the factor of 2 from the three covariance cross-terms, which gives 13.07% (answer choice A). The covariance matrix is symmetric, Cov(1,2) and Cov(2,1) are the same entry, so each distinct covariance term must be counted twice in the expanded formula.

⚠️

Watch out for this

The variance-squared trap in the covariance cross-term Many candidates read the two-asset portfolio variance formula and see σ₁² and σ₂² in the first two terms. They then plug 12² and 25² correctly into those terms. But then they also plug 12² and 25² into the 2w₁w₂Cov(R₁,R₂) term, instead of using the computed covariance of 60. The wrong approach uses: 2 × 0.30 × 0.70 × (12² = 144) = 60.48 for the cross-term alone, producing σ²p ≈ 379.69 and σp ≈ 19.49%. The wrong answer 19.49% does not appear in the option set for this question. More dangerously: using the correlation coefficient (0.20) directly as the covariance value gives 2 × 0.30 × 0.70 × 0.20 = 0.084, producing σ²p ≈ 321.25 and σp ≈ 17.92%. This is close to the correct answer 18.56%, close enough that candidates who made this error often pick the wrong answer without realising. The correct approach: always compute Cov = ρσ₁σ₂ first. The covariance term requires the actual covariance number, not the variances, and not the correlation coefficient.

🧠

Memory Aid

FORMULA HOOK

Square the weights times the individual variances. Then double the weights times the covariance.

Practice Questions · LO1

6 Questions LO1

Score: — / 6

Q 1 of 6 — REMEMBER

An analyst is comparing four asset pairs. Which pair has the strongest linear relationship?

CORRECT: B

CORRECT: B, Correlation measures strength by absolute value, not by sign. +0.80 is further from zero (|0.80| = 0.80) than either −0.67 (|−0.67| = 0.67) or +0.33 (|0.33| = 0.33). A correlation of +0.80 means the two assets move together in a near-perfect positive line. Correlation of −0.67 is strong too, but weaker than +0.80.

Why not A? Students often think a negative correlation must be weak because the word "negative" sounds like "less." This is wrong. −0.67 means the assets move in opposite directions but still quite tightly, closer to a straight line than +0.33 does. Sign tells you direction. Magnitude tells you strength. You need both to compare.

Why not C? +0.33 is a weak positive relationship. The assets move in the same direction, but loosely, the points on a scatter plot are scattered far from the regression line. An investor relying on diversification would find +0.33 much less useful than +0.80 for risk reduction.

---

Q 2 of 6 — UNDERSTAND

Marcus holds two technology stocks. When Stock A's return is above its expected value, Stock B's return is also usually above its expected value. When Stock A falls below its mean, Stock B tends to also fall below its mean. The covariance between them is positive at 0.0023. What does this tell you?

CORRECT: B

CORRECT: B, Covariance measures whether two assets deviate from their own means on the same side or on opposite sides. When both stocks tend to be above their expected values simultaneously (or below simultaneously), each product of deviations is positive, and the covariance is positive. This describes co-movement around their respective means, not whether those means are the same.

Why not A? Students confuse covariance with having identical expected values. Covariance says nothing about whether two assets have the same mean. A stock with a 15% expected return can have positive covariance with a stock that has a 6% expected return, they both happen to be above or below their own targets together. Knowing the covariance tells you nothing about what those targets are.

Why not C? A covariance of 0.0023 is not large in absolute terms, but "small covariance" does not mean "low risk." Covariance describes the relationship between two assets' returns. Risk is measured by each asset's own variance and by how those variances combine. A positively covarying pair can still be low risk if both assets have small individual variances. Covariance and risk are different concepts.

---

Q 3 of 6 — UNDERSTAND

An analyst is given a covariance matrix for a five-asset portfolio. How many unique covariance terms, excluding variances, are needed to fully describe the co-movement between all five assets?

CORRECT: C

CORRECT: C, For five assets, the covariance matrix has 5 × 5 = 25 total entries. The five diagonal entries are variances, not covariances. That leaves 20 off-diagonal entries. But Cov(A,B) = Cov(B,A), so each unique pair of assets appears twice in the matrix (above and below the diagonal). The matrix is symmetric. Counting only one side: 20 ÷ 2 = 10 unique covariance terms.

Why not A? Five is the count of variances on the diagonal, each asset's own variance. These are not covariance terms at all. Confusing variances with covariances is a common error on covariance matrix questions.

Why not B? Students who count 20 are counting all off-diagonal entries without recognizing that the matrix is symmetric. Cov(Asset 1, Asset 2) and Cov(Asset 2, Asset 1) are the same number. The matrix writes it in two places. You only need it once. The correct count is always n(n−1)/2, which for n = 5 gives 5(4)/2 = 10.

---

Q 4 of 6 — APPLY

Priya builds a two-asset portfolio. She allocates 45% to Asset 1 (standard deviation 16%) and 55% to Asset 2 (standard deviation 10%). The correlation between the two assets is 0.50. What is the portfolio's standard deviation of returns?

CORRECT: C

CORRECT: C, Step 1: Convert correlation to covariance. Cov = 0.50 × 16 × 10 = 80. Step 2: Apply the two-asset portfolio variance formula: σ²p = w₁²σ₁² + w₂²σ₂² + 2w₁w₂Cov = (0.45)²(256) + (0.55)²(100) + 2(0.45)(0.55)(80) = 51.84 + 30.25 + 39.60 = 121.69. Step 3: √121.69 = 13.20%.

Why not A? 10.43% is what you get if you forget the covariance cross-term entirely and only sum the two weighted variance components (51.84 + 30.25 = 82.09, √82.09 = 9.06%, not 10.43). This error omits how the two assets move together. The covariance term always contributes to portfolio variance unless correlation equals zero.

Why not B? 21.50% is the result of a specific trap: inserting the variance values (σ² = 256 and σ² = 100) directly into the covariance cross-term instead of inserting the actual computed covariance of 80. This gives 2 × 0.45 × 0.55 × 256 ≈ 126.72 for the cross component alone, producing a final standard deviation far above the correct 13.20%. Students who make this error confuse which numbers go in which slot of the portfolio variance formula.

---

Q 5 of 6 — ANALYZE

An analyst claims: "A lower correlation between two assets means higher portfolio risk, because the assets are less predictable when they move independently." Is this statement correct or incorrect, and why?

CORRECT: C

CORRECT: C, The statement is backwards. Lower correlation means the two assets tend to deviate in opposite directions. When one is above its mean, the other tends to be below its mean. These deviations partially cancel each other out, reducing the portfolio's overall variance. The lower the correlation, the more the two assets hedge each other, and the lower the portfolio's risk relative to the weighted average of their individual risks.

Why not A? "Independent" assets (correlation close to zero) give the maximum diversification benefit. Independent means knowing one asset's return tells you nothing about the other, but it also means their deviations do not reinforce each other. Portfolio risk falls, not rises. The analyst has confused independence with unpredictability in the wrong direction.

Why not B? This answer reverses the logic. Lower correlation does not mean "fewer reliable return sources." It means the two return streams are not locked together, which is exactly what allows them to offset each other. The portfolio's expected return is unchanged (it depends only on weights and individual expected returns). What changes with lower correlation is the risk, it goes down.

---

Q 6 of 6 — TRAP

A portfolio consists of two assets with weights of 0.40 for Asset 1 and 0.60 for Asset 2. Asset 1 has a standard deviation of 20%, Asset 2 has a standard deviation of 15%. Their correlation is 0.20. What is the portfolio's standard deviation?

CORRECT: A

CORRECT: A, Step 1: Covariance = 0.20 × 20 × 15 = 60. Step 2: Portfolio variance = w₁²σ₁² + w₂²σ₂² + 2w₁w₂Cov = (0.40)²(400) + (0.60)²(225) + 2(0.40)(0.60)(60) = 64 + 81 + 28.80 = 173.80. Step 3: √173.80 = 13.31%. The portfolio standard deviation of 13.31% is well below the weighted average of individual standard deviations (0.40 × 20 + 0.60 × 15 = 17%) because the assets partially offset each other.

Why not B? 13.96% results from using correlation (0.20) directly inside the covariance cross-term instead of computing Cov = ρσ₁σ₂ first. This replaces the correct covariance of 60 with 0.20, producing σ² ≈ 194.78 and σ ≈ 13.96%. The correlation coefficient cannot replace the covariance in the formula, the formula requires the actual covariance value, not the correlation.

Why not C? 21.50% is the exact wrong number produced by the trap: using the variance values (20² = 400 and 15² = 225) in the 2w₁w₂Cov(R₁,R₂) term instead of using the actual covariance of 60. This gives 2(0.40)(0.60)(400) = 192.00 for the cross-term alone, inflating portfolio variance to 462.20 and producing σ ≈ 21.50%. Students who make this error confuse the variance terms (which go in w₁²σ₁² and w₂²σ₂²) with the covariance term (which uses Cov, not σ²). The covariance term uses the computed covariance, not the variances of the individual assets.

---

LO 1 Done ✓

Ready for the next learning objective.

🔒 PRO Feature

How analysts use this at work

Real-world applications and interview questions from top firms.

Quantitative Methods · Portfolio Mathematics · LO 2 of 3

Why does one risky asset seem safer alongside another , even though you own both?

Calculate and interpret covariance and correlation to understand how two assets move together, and why standardizing covariance reveals the true strength of that relationship.

⏱ 8min-15min

6 questions

HIGH PRIORITYANALYZE🧮 Calculator

Why this LO matters

Calculate and interpret covariance and correlation to understand how two assets move together, and why standardizing covariance reveals the true strength of that relationship.

INSIGHT

Two assets can both be risky yet make a portfolio safer. This works only if they do not move together. A bank stock that rises when bonds fall is more valuable in a portfolio than a bank stock that always rises at the same time as bonds. Covariance measures the raw extent to which two assets move together , both positive and negative , in percentage-squared units that are almost impossible to interpret on their own. Correlation standardizes that measurement into a bounded number between −1 and +1, where +1 means perfectly together, −1 means perfectly opposite, and 0 means no linear relationship. That single standardized number is what portfolio managers actually use to make decisions.

How to measure co-movement between two assets

Think about two friends who commute to work. Some days both are late. Some days both are early. Some days one is late and the other is on time. If you tracked this over a year, you could ask: how often do their delays move together? That is covariance. And if you wanted to compare this friendship pair to two other pairs of friends, you would need to adjust for the fact that one pair lives farther from the office and has more variable commute times overall. That adjustment is correlation.

The wrong approach is to look at how much each person varies independently and assume that tells you how they vary together. Two highly variable people may have nothing to do with each other. Two quiet people may move in perfect lockstep. Individual volatility and co-movement are separate measurements.

Both covariance and correlation start from the same raw material: a joint probability table that lists all possible return pairs and their likelihoods.

Building covariance and correlation from a joint probability table

Joint probability function. A table showing all possible combinations of returns for two assets and the probability that each pair occurs simultaneously. Use it to identify which return pairs can happen together and how likely each pair is. This determines the weight each scenario carries in the calculation.

Expected return for each asset separately. The weighted average of that asset's returns across all scenarios. Calculate this once per asset and hold it in memory. It is the reference point against which all returns are measured when computing deviations.

Deviation from expected return. Each return in a scenario minus that asset's expected return. Compute one deviation for each asset in each scenario. Calculate all deviations before multiplying anything. This is where many errors occur, because candidates skip one scenario or make a sign error in the subtraction.

Cross-product of deviations. Multiply Asset A's deviation by Asset B's deviation at each scenario. A positive cross-product means both assets deviated in the same direction in that scenario. A negative cross-product means they deviated in opposite directions.

Covariance. The probability-weighted sum of all cross-products. Multiply each cross-product by its joint probability, then add across all scenarios. The result is in percentage-squared units, which makes direct interpretation difficult. Do not stop here if the question asks for correlation.

Correlation. Covariance divided by the product of the two standard deviation|standard deviations. This removes the percentage-squared units and bounds the result between −1 and +1. Use this number to interpret the strength of the linear relationship: +1 means perfectly positive, −1 means perfectly negative, 0 means no linear relationship.

FORWARD REFERENCE

Standard deviation , what you need for this LO only

Standard deviation measures how much an asset's returns typically vary around their average. It is the square root of variance. For this LO, use the standard deviation values already calculated or provided in the problem , you do not need to derive them from scratch. You will study standard deviation fully in Quantitative Methods Learning Modules 1,2.

→ Quantitative Methods

FORWARD REFERENCE

Portfolio variance and portfolio standard deviation , what you need for this LO only

Portfolio variance measures the total risk of holding multiple assets together. It depends directly on covariance and correlation between the assets held. For this LO, you only need to calculate and interpret covariance and correlation given a joint probability table. The full portfolio variance formula is in Portfolio Management Module 7. That module shows how lower correlation directly reduces portfolio risk , even when the individual assets are each quite volatile.

→ Portfolio Management

Worked examples: from raw table to interpreted correlation

The four examples below follow one continuous thread. Examples 1 and 2 use Priya Menon's data at Meridian Capital to build covariance step by step. Example 3 standardizes that covariance into a correlation coefficient. Example 4 shifts to Tobias Ferreira at Centaur Asset Management to show why covariance magnitude cannot be compared across different asset pairs.

Worked Example 1

Calculating expected returns from a joint probability table

Priya Menon is a junior analyst at Meridian Capital, a Mumbai-based fund. She is evaluating two equity positions, Stock A and Stock B, that her portfolio manager holds simultaneously. The joint probability table below shows all three scenarios that her economic model predicts, along with the probability of each scenario occurring. Before Priya can calculate covariance, she needs the expected return for each stock.

Scenario	Probability	Return on Stock A	Return on Stock B
Expansion	0.20	25%	20%
Stable	0.50	12%	16%
Contraction	0.30	10%	10%

🧠Thinking Flow — Expected returns as the reference point for covariance

The question asks

What is the average return each stock earns, weighted by the probability of each scenario?

Key concept needed

Expected return, the probability-weighted average of all possible outcomes.

Step 1 , Identify the wrong approach

Many candidates add the three returns and divide by three. For Stock A, that gives (25 + 12 + 10) ÷ 3 = 15.67%. This ignores the probabilities entirely. The stable scenario has a 50% chance of occurring. It must carry more weight than the expansion scenario at 20%.

Step 2 , Apply probability weighting for Stock A

E(R_A) = 0.20 × 25% + 0.50 × 12% + 0.30 × 10% E(R_A) = 5.00% + 6.00% + 3.00% E(R_A) = 14.00%

Step 3 , Apply probability weighting for Stock B

E(R_B) = 0.20 × 20% + 0.50 × 16% + 0.30 × 10% E(R_B) = 4.00% + 8.00% + 3.00% E(R_B) = 15.00%

Step 4 , Sanity check

Both expected returns should fall inside the range of possible returns for each stock. Stock A's returns range from 10% to 25%, and 14% sits inside that range. Stock B's returns range from 10% to 20%, and 15% sits inside that range. Neither answer is an outlier. ✓ ✓ Answer: E(R_A) = 14%, E(R_B) = 15%. Hold both values. They are the reference points for every deviation in the covariance calculation that follows.

🧮 Stock A:

`2ND``FV`

Clear all registers → 0

`.20``×``25``=`

Expansion term for Stock A → 5

`+``.50``×``12``=`

Add stable term for Stock A → 11

`+``.30``×``10``=`

Add contraction term for Stock A → 14

Keysequence

What it does → Display

---

--- → ---

`.20``×``20``=`

Expansion term for Stock B → 4

`+``.50``×``16``=`

Add stable term for Stock B → 12

`+``.30``×``10``=`

Add contraction term for Stock B → 15

⚠️ Entering (25 + 12 + 10) ÷ 3 gives 15.67 for Stock A and (20 + 16 + 10) ÷ 3 gives 15.33 for Stock B. These are unweighted averages. They ignore the probability column and produce wrong expected returns. Every subsequent calculation , covariance, correlation , will be wrong if you use these numbers.

Record: E(R_A) = 14 Stock B: Record: E(R_B) = 15

Worked Example 2

Computing covariance from deviations and joint probabilities

Continuing from Example 1, Priya now has E(R_A) = 14% and E(R_B) = 15%. Her manager asks for the covariance of the two stocks' returns , a measure of how the two stocks move together across the three economic scenarios.

🧠Thinking Flow — Covariance as probability-weighted cross-products of deviations

The question asks

How much do Stock A and Stock B move together, measured as the probability-weighted sum of their return deviations multiplied together?

Key concept needed

Covariance, computed as the sum across all scenarios of (probability) × (Stock A deviation) × (Stock B deviation).

Step 1 , Identify the wrong approach

Many candidates multiply the raw returns together rather than the deviations. In the expansion scenario, that gives 0.20 × 25 × 20 = 100. This omits the expected-return subtraction entirely and produces a large positive number with no relationship to how the stocks move together around their averages.

Step 2 , Calculate deviations for each stock in each scenario

Expansion: Stock A deviation = 25 − 14 = +11. Stock B deviation = 20 − 15 = +5. Stable: Stock A deviation = 12 − 14 = −2. Stock B deviation = 16 − 15 = +1. Contraction: Stock A deviation = 10 − 14 = −4. Stock B deviation = 10 − 15 = −5.

Step 3 , Multiply each pair of deviations

Expansion cross-product: (+11) × (+5) = +55. Stable cross-product: (−2) × (+1) = −2. Contraction cross-product: (−4) × (−5) = +20. Both deviations on the same side of their means give a positive cross-product. Deviations on opposite sides give a negative cross-product.

Step 4 , Weight each cross-product by its joint probability and sum

Expansion: 0.20 × 55 = 11. Stable: 0.50 × (−2) = −1. Contraction: 0.30 × 20 = 6. Cov(R_A, R_B) = 11 + (−1) + 6 = 16

Step 5 , Sanity check

The covariance is positive. In both expansion (both above average) and contraction (both below average), the two stocks deviate in the same direction. The only scenario where they deviate in opposite directions is stable, contributing only −1. A positive covariance signals that these two stocks tend to rise and fall together. ✓ ✓ Answer: Cov(R_A, R_B) = 16 %². This value cannot be interpreted for magnitude. Only its sign and its use in the correlation formula are meaningful at this stage.

🧮 BA II Plus Keystrokes

`2ND``FV`

Clear all registers → 0

`.20``×``(``25``−``14``)``×``(``20``−``15``)``=`

Probability-weighted cross-product, expansion → 11

`+``.50``×``(``12``−``14``)``×``(``16``−``15``)``=`

Add probability-weighted cross-product, stable → 10

`+``.30``×``(``10``−``14``)``×``(``10``−``15``)``=`

Add probability-weighted cross-product, contraction → 16

⚠️ Omitting one of the three scenarios is the single most common error. Omitting the contraction scenario gives 10 instead of 16. Omitting the stable scenario gives 17 instead of 16. Both results appear plausible and may appear as wrong answer choices. Always confirm that you have entered exactly as many terms as there are rows in the joint probability table.

Final display: 16

Worked Example 3

Standardizing covariance into a correlation coefficient

Priya now has all the pieces: Cov(R_A, R_B) = 16, the standard deviation of Stock A is σ_A = 5.69%, and the standard deviation of Stock B is σ_B = 3.74%. Her manager wants the correlation coefficient between the two stocks so she can compare this pair's relationship to other asset pairs in the portfolio , something she cannot do with covariance alone.

🧠Thinking Flow — Correlation as standardized covariance

The question asks

What is the correlation coefficient between Stock A and Stock B, and what does it tell us about the strength and direction of their relationship?

Key concept needed

Correlation , covariance divided by the product of the two standard deviations.

Step 1 , Identify the wrong approach

Many candidates attempt to interpret covariance directly. They see Cov = 16 and try to decide whether 16 is "large" or "small." This is not possible. Covariance is expressed in %², and 16 %² has no natural benchmark. You cannot compare it across different asset pairs because it scales with the magnitude of the returns. Standardize first. Interpret after.

Step 2 , Apply the correlation formula

ρ(R_A, R_B) = Cov(R_A, R_B) ÷ (σ_A × σ_B) ρ(R_A, R_B) = 16 ÷ (5.69 × 3.74) Denominator: 5.69 × 3.74 = 21.28 Result: 16 ÷ 21.28 = 0.752

Step 3 , Verify the unit cancellation

Covariance is in %². Standard deviation is in %. The denominator is % × % = %². So %² ÷ %² cancels to a pure number with no units. Correlation must be unitless. ✓

Step 4 , Sanity check

Correlation must lie between −1 and +1. Our result is 0.752, which sits inside the valid range. The positive sign confirms the finding from Example 2: the two stocks tend to move in the same direction. The magnitude of 0.75 indicates a strong, though not perfect, positive linear relationship. ✓ ✓ Answer: ρ(R_A, R_B) ≈ 0.75. When Stock A's return is above its average, Stock B's return tends to be above its average as well. Because the correlation is not +1.0, the relationship is not perfectly rigid , there is still some diversification benefit from holding both.

🧮 BA II Plus Keystrokes

`2ND``FV`

Clear all registers → 0

`16``÷``(``5.69``×``3.74``)``=`

Divide covariance by product of standard deviations → 0.7519...

⚠️ Dividing covariance by only one standard deviation gives 16 ÷ 5.69 = 2.81. A result greater than 1 is impossible for correlation. If your answer exceeds +1 or falls below −1, you have used only one standard deviation in the denominator. Check that both σ_A and σ_B appear in your calculation before proceeding.

Round to two decimal places: 0.75

Worked Example 4

Interpreting covariance sign and correlation magnitude across asset pairs

Tobias Ferreira is a risk analyst at Centaur Asset Management in São Paulo. He is reviewing a four-asset portfolio and has extracted covariance and correlation values for two asset pairs from his risk system. He must interpret what these numbers mean for portfolio construction without recalculating them from scratch.

Pair	Covariance	Std Dev Asset 1	Std Dev Asset 2	Correlation
Pair X: Equity A vs Equity B	+0.00135	4.17%	3.86%	+0.84
Pair Y: Bond C vs Bond D	+0.000318	1.32%	2.64%	+0.91

Tobias notes that Pair X has a much larger covariance than Pair Y, yet his manager claims Pair Y's assets are more tightly linked. Who is correct, and why?

🧠Thinking Flow — Covariance magnitude is not comparable across pairs

The question asks

Can we compare the size of covariance across different asset pairs to judge which pair moves more tightly together?

Key concept needed

Covariance cannot be compared across pairs. Its magnitude depends on the scale of the underlying returns. Correlation, which standardizes covariance, is the correct tool for comparison.

Step 1 , Identify the wrong approach

Many candidates rank pairs by covariance magnitude. They see Pair X covariance = 0.00135 and Pair Y covariance = 0.000318, and conclude that Pair X has the stronger relationship. This is incorrect. The two covariances reflect the scale of each asset's returns, not just the tightness of the relationship.

Step 2 , Compare using correlation

Pair X correlation = +0.84. Pair Y correlation = +0.91. Pair Y has the higher correlation. Its assets move more tightly together in a linear sense, even though the raw covariance number is smaller.

Step 3 , Explain the scale effect

Pair X involves equity assets with standard deviations near 4%. Equities are volatile. Their covariance is large in absolute terms not because the relationship is tighter, but because large return swings multiply together to produce large cross-products. Pair Y involves bond assets with standard deviations near 1,2.5%. Bonds are less volatile. Their covariance is small because the underlying return swings are small , but the two bonds still move together almost perfectly (ρ = 0.91) within their quieter range.

Step 4 , Sanity check

Both correlations are between −1 and +1. ✓ Both are positive, confirming that within each pair, returns move in the same direction. Tobias's manager is correct. ✓ ✓ Answer: Tobias's manager is correct. Pair Y's assets (ρ = 0.91) are more tightly linked than Pair X's assets (ρ = 0.84), even though Pair X's covariance is numerically larger. Covariance is not a valid comparison tool across different asset pairs because it scales with each asset's own volatility.

With the four examples complete, one specific trap accounts for more wrong answers on this LO than any other. Read the next section before attempting the quiz.

⚠️

Watch out for this

The Raw-Returns Covariance Trap. A candidate who multiplies raw returns instead of deviations from expected returns gets: 0.20 × 25 × 20 + 0.50 × 12 × 16 + 0.30 × 10 × 10 = 100 + 96 + 30 = 226. A candidate who omits one scenario entirely gets either 10 or 17 instead of 16. The correct covariance, computed by subtracting expected returns first and then probability-weighting the cross-products of deviations, is 16. Candidates make this error because they treat covariance like a weighted average of return products rather than a weighted average of deviation products. The formula measures co-movement around the mean, not the level of returns themselves. Before submitting, count your terms: the number of probability-weighted cross-products must equal the number of rows in the joint probability table, and each term must subtract the expected return before multiplying.

🧠

Memory Aid

CONTRAST ANCHOR

Covariance measures direction and scale. Correlation measures direction and strength.

Practice Questions · LO2

6 Questions LO2

Score: — / 6

Q 1 of 6 — REMEMBER

The covariance between two asset returns is best described as:

CORRECT: B

CORRECT: B , Covariance measures co-movement around the mean. For each scenario, you subtract each asset's expected return from its actual return (the deviation), multiply the two deviations together (the cross-product), and then weight that cross-product by the probability of the scenario occurring. Summing across all scenarios gives the covariance.

Why not A? This describes multiplying raw returns rather than deviations from expected returns. Using raw returns ignores whether returns are above or below their average , the very information covariance is designed to capture. This approach produces a large positive number (like 226 in the Priya Menon example) that has no relationship to how the two assets move together around their respective means.

Why not C? This describes the correlation formula, not covariance. Dividing by the product of standard deviations is the standardization step that converts covariance into correlation. Covariance itself has no such division , it retains the percentage-squared units that come from multiplying two return deviations together. Applying this formula to raw returns would produce an even more meaningless result.

---

Q 2 of 6 — UNDERSTAND

An analyst computes a covariance of +43 %² between two assets. Her colleague computes a covariance of +12 %² between a different pair of assets. Which of the following conclusions is most accurate?

CORRECT: C

CORRECT: C , Both covariances are positive, so both pairs tend to move in the same direction. When one asset in a pair is above its average, the other tends to be above its average too. The sign of covariance is always informative. However, the magnitude of covariance scales with the volatility of the underlying assets, not just with the tightness of the relationship. A pair of volatile equities will produce a large covariance even if the relationship is moderate. A pair of quiet bonds will produce a small covariance even if the relationship is nearly perfect. Correlation, which divides by the product of standard deviations, is required to compare strength across pairs.

Why not A? Ranking pairs by covariance magnitude mistakes scale for strength. The asset pair with covariance 43 may involve assets with very high individual volatility, inflating the cross-products. The pair with covariance 12 may involve low-volatility assets that move together almost perfectly. Without knowing the standard deviations, no conclusion about relative strength is valid. Tobias Ferreira's example at Centaur Asset Management shows exactly this pattern.

Why not B? The direction of the relationship can always be read directly from the sign of covariance. A positive covariance means the two assets tend to deviate from their means in the same direction. A negative covariance means they tend to deviate in opposite directions. Correlation is needed for strength comparison, not for direction.

---

Q 3 of 6 — APPLY

Nadia Osei is an analyst at Crestline Investments in Accra. She is evaluating two assets, Fund P and Fund Q, using the joint probability table below. E(R_P) = 9.4% and E(R_Q) = 9.8%. What is the covariance between Fund P and Fund Q?

Scenario	Probability	Return on Fund P	Return on Fund Q
Boom	0.30	18%	14%
Normal	0.50	8%	10%
Bust	0.20	2%	4%

CORRECT: B

CORRECT: B , Compute the deviation and cross-product for each scenario, weight by probability, and sum.

Boom: (18 − 9.4) × (14 − 9.8) = 8.6 × 4.2 = 36.12. Weighted: 0.30 × 36.12 = 10.836.

Normal: (8 − 9.4) × (10 − 9.8) = (−1.4) × 0.2 = −0.28. Weighted: 0.50 × (−0.28) = −0.140.

Bust: (2 − 9.4) × (4 − 9.8) = (−7.4) × (−5.8) = 42.92. Weighted: 0.20 × 42.92 = 8.584.

Cov = 10.836 + (−0.140) + 8.584 = 14.56 %².

Why not A? This result comes from omitting the bust scenario entirely. Without the third row, the sum is 10.836 + (−0.140) = 10.696, which rounds to approximately 11.20. Omitting any scenario is a counting error. The number of probability-weighted terms must equal the number of rows in the joint probability table.

Why not C? This result comes from treating the normal scenario cross-product as positive rather than negative. Fund P's deviation in the normal scenario is −1.4 (below average), and Fund Q's deviation is +0.2 (above average). These are on opposite sides of their means, so their product is −0.28, not +0.28. Using the absolute value instead of the signed deviation overstates the total covariance and produces a result near 18.

---

Q 4 of 6 — APPLY+

Continuing from Q3, the standard deviation of Fund P is σ_P = 5.48% and the standard deviation of Fund Q is σ_Q = 3.49%. Using a covariance of 14.56 %², what is the correlation between Fund P and Fund Q, and what does it indicate about portfolio diversification?

CORRECT: C

CORRECT: C , Divide covariance by the product of the two standard deviations.

ρ = 14.56 ÷ (5.48 × 3.49) = 14.56 ÷ 19.12 = 0.762, rounded to 0.76.

A correlation of 0.76 is strong and positive. Both assets tend to rise above their means together and fall below their means together. Holding both in a portfolio does reduce risk compared to holding either alone , but the reduction is modest because the assets are so strongly linked. Maximum diversification benefit requires low or negative correlation.

Why not A? A correlation of 2.66 is mathematically impossible. Correlation is bounded between −1 and +1 by construction. A result above 1 signals a specific calculation error: dividing covariance by only one standard deviation. Here, 14.56 ÷ 5.48 = 2.66. If your correlation exceeds 1 in absolute value, check that you used the product of both standard deviations in the denominator.

Why not B? ρ = 0.48 results from an incorrect denominator. A candidate who uses variance rather than standard deviation , for example, squaring σ_P to get 30.03 and using that alone in the denominator , arrives at a result near 0.48. The denominator of the correlation formula is always the product of the two standard deviations (not variances), which provides the percentage-squared units needed to cancel with the percentage-squared units of covariance.

---

Q 5 of 6 — ANALYZE

Malik Johansson is a portfolio analyst at Nordic Bridge Fund in Stockholm. He is comparing two pairs of assets: His manager says: "Pair 1 is a better diversifier than Pair 2." Malik disagrees. Which of the following best supports Malik's position?

Pair	Covariance	Correlation
Pair 1: two commodity futures	+38.4 %²	+0.62
Pair 2: two investment-grade bonds	+1.7 %²	+0.88

CORRECT: B

CORRECT: B , Correlation is the correct tool for comparing linear co-movement across asset pairs. Pair 2 has ρ = 0.88, meaning its two bonds move together very tightly. Pair 1 has ρ = 0.62, a meaningful but weaker positive relationship. The lower the correlation, the greater the diversification benefit. Pair 1's two commodity futures move less in sync, so combining them reduces portfolio risk more effectively than combining Pair 2's bonds. Malik is right to disagree with his manager.

Why not A? This option makes exactly the error the LO warns against: ranking asset pairs by covariance magnitude and treating a larger covariance as evidence of a stronger relationship. Pair 1's large covariance reflects the high individual volatility of commodity futures , the underlying return swings are large, so the cross-products are large. The covariance says nothing about whether the relationship is tighter than Pair 2's. Only correlation, which adjusts for each asset's own volatility, makes the comparison valid.

Why not C? Portfolio weights are needed to calculate the portfolio's overall variance, but the question asks specifically about the relative co-movement between each pair of assets. Correlation answers this question directly and completely without needing to know weights. Weights determine how much of the diversification benefit is captured in a specific portfolio. They do not change the underlying correlation between the asset pair.

---

Q 6 of 6 — TRAP

Yemi Adeyinka is an analyst at Capeview Capital in Lagos. She is calculating the covariance between Stock X and Stock Y using the following joint probability table. E(R_X) = 11% and E(R_Y) = 13%. Yemi multiplies each pair of raw returns by their joint probability and sums the products. What value does she get, and is it the correct covariance?

Scenario	Probability	Return on Stock X	Return on Stock Y
Strong	0.25	22%	20%
Moderate	0.45	10%	14%
Weak	0.30	5%	8%

CORRECT: A

CORRECT: A , Yemi's calculation: 0.25 × 22 × 20 + 0.45 × 10 × 14 + 0.30 × 5 × 8 = 110 + 63 + 12 = 147.6 (approximately). This number is large and positive, but it is not covariance. It is the probability-weighted average of the product of raw return levels, which has no useful economic meaning. Covariance measures co-movement around the mean, so deviations must be computed first.

The correct covariance uses deviations: Strong (22−11)×(20−13) = 11×7 = 77. Moderate (10−11)×(14−13) = (−1)×1 = −1. Weak (5−11)×(8−13) = (−6)×(−5) = 30. Weighted sum: 0.25×77 + 0.45×(−1) + 0.30×30 = 19.25 − 0.45 + 9.00 = 27.80 %².

Why not B? The value 27.8 is actually the correct covariance , but this option states it is what Yemi computed using raw returns. Yemi did not compute 27.8. She computed 147.6 using the wrong approach. This option is designed to test whether candidates can distinguish between what Yemi calculated and what the correct answer is. The number 27.8 only appears when you apply the proper deviation-based formula.

Why not C? Option C presents the correct number from Yemi's raw-return calculation but the wrong conclusion , claiming 147.6 is valid covariance. Covariance is explicitly defined as the probability-weighted sum of cross-products of deviations from expected returns, not cross-products of raw return levels. The raw-returns approach inflates the result by including the effect of return levels themselves, which has nothing to do with co-movement around the mean. Any result as large as 147.6 for typical equity returns should immediately signal an error.

---

Glossary

joint probability function

A table listing all possible combinations of outcomes for two variables, together with the probability that each combination occurs simultaneously. Like a weather forecast that specifies the probability of "hot and rainy," "hot and dry," "cool and rainy," and "cool and dry" , covering every possible pair rather than each outcome separately.

covariance

A measure of how two variables move together around their averages. Positive covariance means when one is above its average, the other tends to be above its average too. Negative covariance means the opposite. Think of tracking whether two friends' exam scores move together , positive covariance means when one scores above their usual level, the other tends to as well.

correlation

Covariance standardized by dividing by the product of two standard deviations, producing a number always between −1 and +1. A correlation of +1 means perfect positive movement together, −1 means perfect opposite movement, and 0 means no linear relationship. Like covariance, but converted to a universal scale so you can compare any two pairs fairly, regardless of how volatile each asset is on its own.

expected return

The probability-weighted average of all possible returns for an asset across all scenarios. If there is a 60% chance of earning 10% and a 40% chance of earning 20%, the expected return is 14%. It is the return you would earn on average if the same investment were repeated many times.

deviation from expected return

The difference between an actual return in a specific scenario and that asset's expected return. If Stock A's expected return is 12% and it earns 18% in the boom scenario, the deviation is +6%. Deviations show how far above or below average a particular outcome falls, and they can be positive or negative.

standard deviation

A measure of how widely an asset's returns typically spread around their average. High standard deviation means returns vary a lot from period to period. Low standard deviation means they cluster close to the average. Like the difference between a bus that arrives at exactly 8:00 every day versus one that arrives anywhere between 7:45 and 8:20.

portfolio variance

The total risk of holding a combination of assets together, measured as the weighted sum of individual asset variances plus terms that capture co-movement between pairs. It depends on both how risky each asset is individually and how tightly they move together. Lower correlation between assets reduces portfolio variance even when each asset is individually risky.

portfolio standard deviation

The square root of portfolio variance, expressed in the same percentage units as individual asset returns. This makes the total portfolio risk easier to interpret alongside the expected return of the portfolio.

LO 2 Done ✓

Ready for the next learning objective.

🔒 PRO Feature

How analysts use this at work

Real-world applications and interview questions from top firms.

Quantitative Methods · Portfolio Mathematics · LO 3 of 3

Why does a portfolio that never crashes still lose to one that does,according to safety-first investing?

Measure shortfall risk using Roy's safety-first ratio, then identify the optimal portfolio by selecting the highest ratio, not the highest return.

⏱ 8min-15min

6 questions

HIGH PRIORITYEVALUATE🧮 Calculator

Why this LO matters

Measure shortfall risk using Roy's safety-first ratio, then identify the optimal portfolio by selecting the highest ratio, not the highest return.

INSIGHT

Your goal is not the highest return. It is the widest cushion per unit of shakiness. The safety-first ratio measures how many standard deviations of buffer sit between where you expect the portfolio to land and where you absolutely cannot go below. A portfolio with a 10% expected return and 8% volatility is safer at a 3% threshold than a portfolio with 16% expected return and 24% volatility. The first sits further from disaster relative to how dispersed its outcomes are. Roy's criterion picks the portfolio with the biggest distance-in-standard-deviations, not the biggest distance in raw percentage points.

Understanding Roy's Safety-First Framework

Imagine you are planning a road trip and your car has just enough fuel to reach the next town, with a small reserve. Two routes are available. Route A is faster on average but passes through terrain where fuel consumption is wildly unpredictable. Route B is slower on average but fuel consumption is steady and predictable. Your goal is not to arrive earliest , it is to avoid running out of fuel entirely.

That is Roy's safety-first problem. The fuel threshold is shortfall risk. The safety-first ratio measures how much cushion you have, adjusted for how unpredictable the journey is.

The wrong move: picking Route A because its average arrival time is faster. The right move: computing which route gives you the most buffer-in-standard-deviations above the fuel minimum.

Roy's Safety-First Framework

Shortfall risk. The probability that a portfolio's return will fall below a minimum acceptable level, called the threshold return. Use it to identify which portfolio has the lowest chance of failing to meet the investor's goal.

Threshold return (R_L). The minimum return the investor needs to avoid invading principal or missing a financial goal. Express it as a percentage. If the client gives you an absolute cash amount, divide it by the current portfolio value first.

Safety-first ratio (SFRatio). The distance from the portfolio's expected return to the threshold, measured in standard deviations. Calculate it as [E(R_P) − R_L] / σ_P. The portfolio with the highest ratio is optimal under Roy's criterion.

Roy's selection rule. Choose the portfolio that maximises the safety-first ratio. This automatically minimises the probability of falling short, given the normality assumption below.

Probability of shortfall. Once you have the safety-first ratio, convert it using the standard normal distribution: P(return < R_L) = Normal(−SFRatio). This step requires the negative sign. Omitting it gives the probability of exceeding the threshold, not falling below it.

The normality assumption. All calculations in this LO assume portfolio returns follow a normal distribution, a symmetrical, bell-shaped curve. Without this assumption, the ratio cannot be translated into a probability.

Safety-First Ratio

SFRatio = [E(R_P) − R_L] / σ_P

E(R_P) = Expected return of the portfolio (%)
R_L = Threshold return , minimum acceptable return (%)
σ_P = Standard deviation of portfolio returns (%)

Condition: R_L must be in the same units as E(R_P).
If R_L is given as a cash amount, divide by portfolio value first.
Condition: Returns must be approximately normally distributed for
the ratio to convert to a probability.

FORWARD REFERENCE

Normal distribution and Z-scores , what you need for this LO only

A normal distribution is a symmetrical, bell-shaped probability curve. The mean sits at the centre. Standard deviations mark equal-width bands on both sides. A Z-score measures how many standard deviations a specific value sits away from the mean. The safety-first ratio is a Z-score. To convert it to a probability, feed it into the standard normal cumulative distribution function (CDF) with a negative sign: Normal(−SFRatio). You will study the normal distribution and Z-score calculations fully in Quantitative Methods Module 2. For this LO, you only need to apply Normal(−SFRatio) using a provided Z-table or the complement rule: Normal(−z) = 1 − Normal(+z).

→ Quantitative Methods

How to Apply Roy's Criterion: Step by Step

Three steps, in order. Skipping any step produces the wrong answer.

[PROCESS: Applying Roy's Safety-First Criterion]

Convert the threshold to a percentage. If the investor specifies a cash withdrawal amount, divide it by the current portfolio value. If the threshold is already stated as a percentage return, use it directly.

Calculate the safety-first ratio for every portfolio. Apply SFRatio = [E(R_P) − R_L] / σ_P to each candidate allocation. Do not stop at one.

Select the portfolio with the highest ratio. The highest ratio corresponds to the lowest shortfall probability. This is the safety-first optimal portfolio.

Optional step 4, when asked: Convert the ratio to a shortfall probability. Apply Normal(−SFRatio) using the complement rule: 1 − Normal(+SFRatio).

Worked Example 1

Converting an absolute withdrawal goal into a threshold return

Priya Nambiar manages a GBP 1,350,000 diversified portfolio for a client in Edinburgh. The client plans to withdraw GBP 50,000 at year-end to fund a business partnership. The firm condition: the withdrawal must not reduce the portfolio below its starting value of GBP 1,350,000. Priya needs to establish the minimum return threshold before she can compare any asset allocations.

🧠Thinking Flow — Converting an absolute goal to a percentage threshold

The question asks

What annual return, expressed as a percentage, must the portfolio earn so that a GBP 50,000 withdrawal leaves the principal intact?

Key concept needed

Threshold return (R_L) expressed as a percentage. Many candidates use the absolute withdrawal amount directly as if it were already a percentage. GBP 50,000 is a currency amount. The formula requires a return rate. These cannot be mixed.

Step 1 , Name the wrong approach

Many candidates write R_L = GBP 50,000 and stop. That is a pound amount, not a return rate. Entering a currency figure into SFRatio = [E(R_P) − R_L] / σ_P gives a dimensionally incorrect result. The numerator would combine a percentage and a currency amount.

Step 2 , The correct approach: convert to a percentage return

Divide the required cash amount by the current portfolio value. R_L = GBP 50,000 / GBP 1,350,000 = 0.03704 = 3.70% (rounded to two decimal places). Interpretation: if the portfolio returns exactly 3.70%, the client can withdraw GBP 50,000 and the residual portfolio value will equal the original GBP 1,350,000. Any return below 3.70% forces the client to invade principal.

Step 3 , Sanity check

The result must be a small positive percentage, not a large number and not a currency amount. 3.70% is plausible for a conservative minimum return on a diversified portfolio. If the result had been negative, the client would need no return at all to fund the withdrawal, which contradicts the problem setup. ✓ ✓ Answer: R_L = 3.70%.

Worked Example 2

Calculating the safety-first ratio for four portfolios and selecting the optimal one

Continuing with Priya's client in Edinburgh, whose threshold return is now established as R_L = 3.70%. Priya has constructed four candidate asset allocations. She must apply Roy's safety-first criterion to identify which single allocation minimises the probability that the portfolio return falls below 3.70%.

Allocation	Expected annual return	Standard deviation
A	16%	24%
B	12%	17%
C	10%	12%
D	9%	11%

🧠Thinking Flow — Calculating four safety-first ratios and identifying the maximum

The question asks

Which allocation has the highest safety-first ratio, and therefore the lowest probability of falling below R_L = 3.70%?

Key concept needed

Safety-first ratio for every allocation. The winner is not determined by expected return alone or by standard deviation alone.

Step 1 , Name the two wrong selection rules

Wrong rule 1: choose the highest expected return. That points to Allocation A (16%). Ignoring dispersion is exactly what Roy's criterion corrects for. Wrong rule 2: choose the lowest standard deviation. That points to Allocation D (11%). But if the expected return barely clears the threshold, even a small downward move breaches it. The correct rule: compute the ratio for every allocation, then rank.

Step 2 , Calculate the safety-first ratio for each allocation

Allocation A: SFRatio = (16 − 3.70) / 24 = 12.30 / 24 = 0.5125 Allocation B: SFRatio = (12 − 3.70) / 17 = 8.30 / 17 = 0.4882 Allocation C: SFRatio = (10 − 3.70) / 12 = 6.30 / 12 = 0.5250 Allocation D: SFRatio = (9 − 3.70) / 11 = 5.30 / 11 = 0.4818

Step 3 , Identify the maximum

Ranking highest to lowest: C (0.5250) > A (0.5125) > B (0.4882) > D (0.4818). Allocation C has the highest safety-first ratio.

Step 4 , Sanity check

Allocation A has a higher expected return than C (16% vs 10%) but a lower safety-first ratio (0.5125 vs 0.5250). This is correct: A's extra expected return is more than offset by its extra volatility. The ratio discounts the return cushion by dispersion, exactly as it should. ✓ ✓ Answer: Allocation C, with SFRatio = 0.5250, is the safety-first optimal portfolio.

🧮 BA II Plus Keystrokes

`16−3.70=`

Numerator for Allocation A → 12.3

`÷24=`

SFRatio for Allocation A → 0.5125

`12−3.70=`

Numerator for Allocation B → 8.3

`÷17=`

SFRatio for Allocation B → 0.4882...

`10−3.70=`

Numerator for Allocation C → 6.3

`÷12=`

SFRatio for Allocation C → 0.525

`9−3.70=`

Numerator for Allocation D → 5.3

`÷11=`

SFRatio for Allocation D → 0.4818...

⚠️ Using the absolute GBP amount (50,000) instead of R_L = 3.70% in the numerator. For Allocation A, this produces (16 − 0.037) / 24 = 0.665, a completely wrong ratio. Always convert the withdrawal amount to a percentage return before entering any figure into the formula.

Worked Example 3

Calculating the probability of falling below the threshold return

Priya's client in Edinburgh now asks a follow-up question: "Even with the best allocation, what is the actual probability that I will earn less than 3.70% this year?" Priya must convert the safety-first ratio for Allocation C into a shortfall probability using the standard normal distribution.

🧠Thinking Flow — Converting the safety-first ratio to a shortfall probability

The question asks

Given SFRatio = 0.525 for Allocation C and R_L = 3.70%, what is P(return < 3.70%)?

Key concept needed

The standard normal CDF with a negative argument. The safety-first ratio is a Z-score, not a probability. It requires one more step to become a probability.

Step 1 , Name the wrong approach

Many candidates report the SFRatio (0.525) directly as if it were already a probability. It is not. Reporting 0.525 or 52.5% as the shortfall probability is wrong. The ratio must be entered into the standard normal CDF with a negative sign.

Step 2 , Set up the correct formula

P(R_C < 3.70%) = Normal(−0.525)

Step 3 , Evaluate Normal(−0.525) using the complement rule

Step 3a: Round −0.525 to −0.53 for use with a printed Z-table. Step 3b: Look up Normal(+0.53) in the standard normal CDF table. Normal(0.53) = 0.7019. Step 3c: Apply the complement: Normal(−0.53) = 1 − 0.7019 = 0.2981. Expressed as a percentage: approximately 29.8%, or about 30%. (A spreadsheet function NORM.S.DIST(−0.525) without rounding gives 0.2998, approximately 30%.)

Step 4 , Sanity check

Allocation C's expected return of 10% sits well above the threshold of 3.70%. The probability of falling below the threshold should be meaningfully less than 50%. A result of approximately 30% is plausible. If the result had exceeded 50%, the setup would be wrong: a portfolio with an expected return above the threshold cannot have a majority of outcomes below it under a symmetric distribution. ✓ ✓ Answer: P(return < 3.70%) ≈ 30%. There is approximately a 30% probability that Allocation C returns less than 3.70% in a given year.

🧮 BA II Plus Keystrokes

Lookupz=0.53intheZ-table

Find cumulative probability for +0.53 → 0.7019

`1−0.7019=`

Complement rule: Normal(−0.53) = 1 − Normal(0.53) → 0.2981

⚠️ Looking up Normal(+0.525) instead of Normal(−0.525). This gives 1 − 0.2998 = 0.7002, approximately 70%. That is the probability of exceeding the threshold, not falling short of it. The exam may include 70% as an answer choice. The negative sign in Normal(−SFRatio) is not optional.

Worked Example 4

Full end-to-end application when R_L must be derived from an absolute amount

Tomás Ferreira, a portfolio analyst at Meridian Wealth Partners in Lisbon, is advising a client who holds a EUR 2,000,000 portfolio. The client intends to withdraw EUR 90,000 in one year and wants to avoid reducing the portfolio below its current value. Tomás must identify which of three candidate allocations is optimal under Roy's safety-first criterion.

Allocation	Expected annual return	Standard deviation
X	6.5%	8.35%
Y	7.5%	10.21%
Z	8.5%	14.34%

🧠Thinking Flow — Full end-to-end Roy's criterion application

The question asks

Which allocation minimises the probability that the return falls below the threshold needed to fund the EUR 90,000 withdrawal without invading principal?

Key concept needed

Safety-first ratio applied after converting the absolute withdrawal goal to a percentage R_L. Both the conversion step and the ratio comparison step are required.

Step 1 , Convert the withdrawal goal to a percentage threshold

R_L = EUR 90,000 / EUR 2,000,000 = 0.045 = 4.50% Any return below 4.50% means the EUR 90,000 withdrawal partially invades the EUR 2,000,000 principal.

Step 2 , Calculate the safety-first ratio for each allocation

Allocation X: SFRatio = (6.5 − 4.5) / 8.35 = 2.0 / 8.35 = 0.2395 Allocation Y: SFRatio = (7.5 − 4.5) / 10.21 = 3.0 / 10.21 = 0.2938 Allocation Z: SFRatio = (8.5 − 4.5) / 14.34 = 4.0 / 14.34 = 0.2789

Step 3 , Select the allocation with the highest ratio

Ranking: Y (0.2938) > Z (0.2789) > X (0.2395). Allocation Y is the safety-first optimal choice.

Step 4 , Sanity check

Allocation Z has the highest expected return (8.5%) but does not win. Its standard deviation of 14.34% is almost double Y's 10.21%, erasing the return advantage when measured in standard-deviation units. Allocation X has the lowest volatility but also the smallest return cushion above R_L: only 2.0 percentage points. The middle allocation Y balances cushion and dispersion most efficiently. ✓ ✓ Answer: Allocation Y, with SFRatio = 0.2938, is the safety-first optimal portfolio.

🧮 BA II Plus Keystrokes

`90000÷2000000=`

Convert withdrawal to R_L as a decimal → 0.045

`×100=`

Convert to percentage → 4.5

`(6.5−4.5)÷8.35=`

SFRatio for Allocation X → 0.23952...

`(7.5−4.5)÷10.21=`

SFRatio for Allocation Y → 0.29384...

`(8.5−4.5)÷14.34=`

SFRatio for Allocation Z → 0.27895...

⚠️ Selecting Allocation Z because it has the highest expected return (8.5%). Allocation Z produces a safety-first ratio of only 0.279, which is lower than Allocation Y's 0.294. The exam places the highest-return portfolio as an answer choice precisely to catch candidates who stop at return maximisation without ranking all ratios.

Worked Example 5

Connecting the safety-first ratio to the Sharpe ratio

During a team meeting at Meridian Wealth Partners, a junior analyst named Keiko Tanaka asks Tomás why the safety-first ratio looks so similar to the Sharpe ratio she studied last month. Tomás explains that the two ratios share an identical formula structure, and that the Sharpe ratio is actually a special case of Roy's safety-first ratio.

🧠Thinking Flow — Recognising the Sharpe ratio as a special case of the safety-first ratio

The question asks

Under what condition does the safety-first ratio become the Sharpe ratio, and what does this imply about portfolio selection using the Sharpe ratio?

Key concept needed

The safety-first ratio formula with R_L replaced by R_F (the risk-free rate) equals the Sharpe ratio. This is a direct consequence of the formula structure, not a coincidence.

Step 1 , Compare the two formulas side by side

Safety-first ratio: SFRatio = [E(R_P) − R_L] / σ_P Sharpe ratio: Sharpe = [E(R_P) − R_F] / σ_P The only difference is in the numerator: R_L (investor-specific threshold) versus R_F (market-observable risk-free rate).

Step 2 , State the condition for equivalence

When R_L = R_F, the two ratios are identical. If an investor's personal minimum acceptable return happens to equal the current risk-free rate, maximising the safety-first ratio and maximising the Sharpe ratio produce the same portfolio ranking.

Step 3 , Derive the implication

When portfolios are ranked by the Sharpe ratio, the portfolio with the highest Sharpe ratio also minimises the probability that the portfolio return falls below the risk-free rate. The Sharpe ratio is therefore a shortfall minimisation criterion relative to a specific benchmark: the risk-free rate.

Step 4 , Sanity check

This relationship must be directional: a higher Sharpe ratio means a lower probability of underperforming the risk-free rate. A portfolio that earns far above R_F with low volatility has both a high Sharpe ratio and a low shortfall probability relative to R_F. The logic is consistent. ✓ ✓ Answer: The safety-first ratio equals the Sharpe ratio when R_L = R_F. Ranking portfolios by the Sharpe ratio is equivalent to choosing the portfolio that minimises the probability of earning less than the risk-free rate, assuming normally distributed returns.

⚠️

Watch out for this

The Return-Maximiser Trap A candidate who selects the portfolio with the highest expected return without computing the safety-first ratio chooses Allocation A (16% return) and reports a ratio of 0.5125, missing the correct answer entirely. The safety-first optimal portfolio is Allocation C, with a ratio of 0.5250, despite having a lower expected return of 10%. The correct approach is to compute [E(R_P) − R_L] / σ_P for every portfolio in the table and rank the ratios, not the expected returns. Candidates make this error because they assume that maximising expected return and minimising shortfall risk point to the same portfolio. Roy's criterion explicitly adjusts the return cushion by dispersion, so a high return paired with very high volatility can be less safe than a moderate return with moderate volatility. Before selecting an answer, verify that you have computed and ranked the ratio for every allocation shown.

🧠

Memory Aid

FORMULA HOOK

Cushion over shakiness: the portfolio with the widest cushion per unit of shake wins, not the one that reaches highest.

Practice Questions · LO3

6 Questions LO3

Score: — / 6

Q 1 of 6 — REMEMBER

Roy's safety-first criterion states that an investor should select the portfolio that:

CORRECT: B

CORRECT: B , Roy's safety-first criterion defines the optimal portfolio as the one that minimises P(R_P < R_L), where R_L is the investor's minimum acceptable threshold return. The criterion is explicitly about shortfall probability, not about return maximisation or variance minimisation in isolation.

Why not A? This describes mean-variance optimisation, not Roy's criterion. Mean-variance optimisation trades off expected return against variance along the efficient frontier. Roy's criterion has a different objective: it fixes a threshold R_L and minimises the chance of falling below it. The two approaches can point to different portfolios when the investor's R_L does not equal the risk-free rate.

Why not C? Selecting the lowest-standard-deviation portfolio ignores the return cushion above the threshold entirely. A portfolio with very low volatility but an expected return only slightly above R_L can still have a high shortfall probability. If the cushion is tiny, even a small downward move breaches the threshold. Roy's criterion requires evaluating both the cushion and the dispersion together, which is exactly what the safety-first ratio does.

---

Q 2 of 6 — UNDERSTAND

An investor's threshold return is R_L = 5%. Portfolio P has an expected return of 8% and a standard deviation of 12%. Which of the following best describes what the safety-first ratio measures in this context?

CORRECT: C

CORRECT: C , The safety-first ratio SFRatio = [E(R_P) − R_L] / σ_P = (8 − 5) / 12 = 0.25. This number is dimensionless because both the numerator and denominator are in percentage units. It represents how many standard deviations of cushion separate the expected return from the threshold. A larger ratio means the threshold sits farther below the centre of the return distribution, implying a lower shortfall probability.

Why not A? The safety-first ratio is not itself a probability. It is a Z-score. To convert it to a probability, you must apply the standard normal CDF with a negative sign: P(R_P < R_L) = Normal(−0.25) ≈ 40.1%. Reporting the ratio of 0.25 as a probability confuses the intermediate result with the final answer and skips the mandatory lookup step.

Why not B? That description defines the Sharpe ratio, where the benchmark is the risk-free rate R_F rather than the investor-specific threshold R_L. The Sharpe ratio and the safety-first ratio share the same formula structure, but they differ in what fills the numerator's benchmark. When R_L = R_F, the two ratios are identical. When R_L differs from R_F, they diverge and can rank portfolios differently.

---

Q 3 of 6 — APPLY

Priya Nambiar manages a GBP 800,000 portfolio. Her client requires an annual withdrawal of GBP 32,000 and insists the portfolio value must not fall below its current level after the withdrawal. Calculate the threshold return R_L that Priya should use when applying Roy's safety-first criterion.

CORRECT: C

CORRECT: C , R_L = GBP 32,000 / GBP 800,000 = 0.04 = 4.00%. This percentage represents the minimum annual return the portfolio must earn so that the GBP 32,000 withdrawal leaves the GBP 800,000 principal intact. The safety-first ratio formula requires R_L in the same units as E(R_P), which is a percentage, so this conversion step is mandatory before any ratio calculation can begin.

Why not A? GBP 32,000 is the absolute currency amount of the withdrawal, not a return rate. Entering a pound amount directly into the numerator of SFRatio = [E(R_P) − R_L] / σ_P creates a dimensionally inconsistent expression: the numerator would mix a percentage with a currency figure, producing a meaningless result. Always divide the required cash amount by the current portfolio value before using it in the formula.

Why not B? This result comes from moving the decimal incorrectly after computing 32,000 / 800,000 = 0.04. Writing 0.40% instead of 4.00% is a factor-of-ten error. A quick reasonableness check catches it: a 0.40% return on GBP 800,000 generates only GBP 3,200, far less than the GBP 32,000 needed for the withdrawal. The threshold must be high enough to actually fund the withdrawal.

---

Q 4 of 6 — APPLY+

Fatima Al-Rashid is evaluating three portfolios for a client whose threshold return is R_L = 2.5%. Portfolio data are shown below. The client states that the maximum tolerable shortfall probability is 25%. Which portfolio is safety-first optimal? Assume normally distributed returns and that Normal(0.25) ≈ 0.599 and Normal(0.33) ≈ 0.629.

Portfolio	Expected Return	Standard Deviation
Alpha	7.0%	18.0%
Beta	5.5%	9.0%
Gamma	4.0%	6.0%

CORRECT: B

CORRECT: B , Safety-first ratios: Alpha = (7.0 − 2.5) / 18.0 = 0.250; Beta = (5.5 − 2.5) / 9.0 = 0.333; Gamma = (4.0 − 2.5) / 6.0 = 0.250. Beta has the highest ratio at 0.333. Shortfall probability for Beta: Normal(−0.333) = 1 − Normal(+0.333) ≈ 1 − 0.629 = 37.1%. For Alpha and Gamma: Normal(−0.250) = 1 − 0.599 = 40.1%. Beta has the lowest shortfall probability of the three and is the safety-first optimal portfolio, even though none of the three achieves a shortfall probability below 25% with these parameters.

Why not A? Alpha has an expected return of 7.0% but a safety-first ratio of only 0.250, identical to Gamma and below Beta's 0.333. Selecting Alpha on the basis of its highest expected return commits the return-maximiser error. Alpha's standard deviation of 18.0% is double Beta's 9.0%, erasing the return advantage entirely when measured in standard-deviation units. The ratio is 0.250 for Alpha versus 0.333 for Beta.

Why not C? Gamma has the lowest standard deviation (6.0%) but also the smallest return cushion above R_L: only 4.0 − 2.5 = 1.5 percentage points. Its safety-first ratio is 1.5 / 6.0 = 0.250, the same as Alpha and below Beta. Low volatility alone does not guarantee safety-first optimality. The ratio must be computed for every portfolio. The lowest-volatility choice is not automatically the winner.

---

Q 5 of 6 — ANALYZE

An analyst claims: "Ranking portfolios by the Sharpe ratio and ranking them by Roy's safety-first ratio will always produce identical orderings, regardless of the investor's threshold return." Evaluate this claim.

CORRECT: C

CORRECT: C , The Sharpe ratio is [E(R_P) − R_F] / σ_P and the safety-first ratio is [E(R_P) − R_L] / σ_P. The formulas are structurally identical, but the benchmarks differ: R_F is market-observable and common to all investors, while R_L is investor-specific. When R_L = R_F, subtracting the same constant from every portfolio's expected return leaves the ranking unchanged. When R_L differs from R_F, the numerators shift differently and portfolios that rank similarly near R_F can diverge near a different R_L.

Why not A? The shared denominator σ_P is not what drives the difference. Both ratios divide by the same σ_P for any given portfolio. The ranking difference arises entirely from the numerator: R_F versus R_L. If R_L = 0 but R_F = 3%, every portfolio's Sharpe numerator is smaller by 3 percentage points, which affects rankings differently depending on each portfolio's return level. The denominators being identical does not prevent the numerator difference from changing rankings.

Why not B? Normality is required for the safety-first ratio to be converted into a shortfall probability via the standard normal CDF. It has nothing to do with whether the Sharpe and safety-first ratio rankings agree. The condition for ranking equivalence is R_L = R_F, not any assumption about the shape of the return distribution. A candidate who confuses the normality requirement with the equivalence condition will select this option.

---

Q 6 of 6 — TRAP

Tomás Ferreira at Meridian Wealth Partners is comparing four asset allocations for a client with a threshold return of R_L = 3.7%. Portfolio data are shown below. Applying Roy's safety-first criterion, which allocation is optimal?

Allocation	Expected Return	Standard Deviation
W	16%	24%
X	12%	17%
Y	10%	12%
Z	9%	11%

CORRECT: B

CORRECT: B , The safety-first ratios are: W = (16 − 3.7) / 24 = 0.5125; X = (12 − 3.7) / 17 = 0.4882; Y = (10 − 3.7) / 12 = 0.5250; Z = (9 − 3.7) / 11 = 0.4818. Allocation Y has the highest ratio at 0.5250 and is the safety-first optimal portfolio. Under Roy's criterion, the winner is whichever portfolio places the threshold return farthest below the expected return in standard-deviation units, not whichever has the highest expected return.

Why not A? Allocation W has the highest expected return (16%) and a safety-first ratio of 0.5125. Selecting W is the classic return-maximiser trap. A candidate who stops at "highest expected return" without computing all four ratios will choose W. But W's standard deviation of 24% is double Y's 12%, and that extra volatility more than offsets the higher return when measured as standard-deviation cushion. Y's ratio of 0.5250 strictly exceeds W's 0.5125. The exam places W's ratio as an answer choice precisely to catch candidates who pick the highest return without ranking all ratios.

Why not C? Allocation Z has the lowest standard deviation (11%) but also the smallest return cushion: only 9 − 3.7 = 5.3 percentage points above the threshold. Its ratio of 5.3 / 11 = 0.4818 is the lowest of the four. Selecting Z reflects the opposite error: treating low volatility as a proxy for safety without computing how far the expected return actually sits above the threshold. Roy's criterion requires both dimensions, cushion and shakiness, to be evaluated together.

---

Glossary

Shortfall risk

The probability that an investment's return will fall below a minimum acceptable level, forcing you to use savings or miss a financial goal. Example: if you need to withdraw money from your portfolio but the market falls short, you might have to dip into your emergency fund instead. This LO measures shortfall risk using the safety-first ratio.

threshold return

The minimum annual percentage return an investor requires to meet a financial goal without reducing their principal. Example: if you have GBP 1,000,000 and need to withdraw GBP 50,000, your threshold return is 5%. Earn less than that and you invade capital. Also called the minimum acceptable return.

Threshold return

The minimum percentage return required to avoid falling short of a financial goal. Used interchangeably with threshold return throughout this LO.

Safety-first ratio

A number that measures how many standard deviations of buffer exist between a portfolio's expected return and the minimum return the investor needs. Calculated as [E(R_P) − R_L] / σ_P. Example: if a portfolio's expected return is 10%, the threshold is 4%, and the standard deviation is 6%, the ratio is (10 − 4) / 6 = 1.0, meaning the expected return sits exactly one standard deviation above the minimum. The higher the ratio, the lower the shortfall risk.

normal distribution

A symmetrical, bell-shaped probability curve where most outcomes cluster near the centre (the mean) and fewer occur at the extremes. Example: student test scores in a large class often follow a normal distribution, with many students near the average and fewer scoring very high or very low. In portfolio management, this distribution is assumed to describe how investment returns spread around their expected value.

standard normal distribution

A specific normal distribution with a mean of zero and a standard deviation of one. Example: converting a raw score to a Z-score and looking it up in a standard normal table is the same process used in this LO when computing P(return < R_L) = Normal(−SFRatio). All Z-score probability lookups use this distribution.

Z-score

A number expressing how many standard deviations a specific value sits away from the mean of a distribution. Example: a Z-score of 1.5 means the value is 1.5 standard deviations above the mean. In this LO, the safety-first ratio is the Z-score measuring how far the expected return sits above the threshold.

Sharpe ratio

A measure of portfolio performance calculated as [E(R_P) − R_F] / σ_P, where R_F is the risk-free rate. It is structurally identical to the safety-first ratio, but uses the risk-free rate as its benchmark instead of the investor's personal threshold return. When R_L = R_F, the Sharpe ratio and the safety-first ratio produce identical portfolio rankings.

risk-free rate

The return an investor earns on an investment with no default risk, typically a short-term government bill. It serves as the baseline benchmark in the Sharpe ratio calculation. In the safety-first framework, the Sharpe ratio is the special case where R_L equals the risk-free rate.

LO 3 Done ✓

You have completed all learning objectives for this module.

🔒 PRO Feature

How analysts use this at work

Real-world applications and interview questions from top firms.

Quantitative Methods · Portfolio Mathematics · Job Ready

From exam to career

Portfolio risk measurement, covariance analysis, and safety-first portfolio construction

LO 5

Covariance and correlation: building the relationship between two assets

How analysts use this at work

Credit analysts at investment banks like JPMorgan and HSBC use covariance and correlation to assess whether the yield premium on a corporate bond actually compensates for the credit risk the investor is taking on. They compare the correlation between the bond's credit spread and the broader equity market's returns. If correlation is high, the bond behaves like equity in a downturn, and the yield premium may not be sufficient. The output is a relative value trade recommendation that a portfolio manager reviews before committing capital.

Performance attribution analysts at institutional asset managers like State Street and Northern Trust use covariance calculations when they decompose a portfolio's realized risk into contributions from individual factors. They calculate covariance terms between the portfolio's factor exposures and report them to clients as part of GIPS-compliant performance presentations. The covariance term tells clients whether their portfolio's risk came from assets moving together or from assets moving against each other.

Interview questions

Fidelity Investments Quantitative Analyst "You are given a joint probability table for two stocks. A junior analyst calculates the covariance by multiplying raw returns rather than deviations from expected returns. She gets 147.6. What did she do wrong, and what is the correct covariance?"

T. Rowe Price Portfolio Analytics Analyst "You observe two asset pairs: one with covariance 0.00135 and correlation 0.84, another with covariance 0.000318 and correlation 0.91. Your manager says the first pair is more tightly linked. Is she right, and why?"

Mercer Senior Research Analyst "When would you use covariance rather than correlation to make a professional decision? Give a specific example where covariance gives you information correlation cannot."

One-line to use in your interview

Interviewers listen for industry-specific language. It signals you understand the concept, not just the definition. Use the plain English version to adapt it in your own words.

I use covariance to measure how two assets move together and correlation to compare the strength of that relationship across different asset pairs, because covariance scales with how volatile each asset is, which makes it impossible to compare across pairs without standardizing first.

In plain English

Covariance tells you whether two assets move together or apart. But a high covariance can mean the assets are just more volatile, not that they are more connected. Correlation puts the relationship on a universal scale so you can compare any two asset pairs directly.

LO 5

Shortfall risk and safety-first portfolio construction

How analysts use this at work

Investment consultants at firms like Mercer and Willis Towers Watson advise pension funds and endowments on asset allocation. They use the safety-first ratio when an institutional client has a specific minimum return requirement, such as an annuity insurer needing to meet a 4.5% guaranteed crediting rate. The consultant calculates the safety-first ratio for each candidate portfolio and presents the ranking to the investment committee. The ratio translates statistical cushion into a shortfall probability that non-technical board members can actually interpret and act on.

Portfolio managers at hedge funds like Bridgewater and Oaktree use shortfall risk analysis when deciding whether a new strategy is worth launching. They backtest the strategy's safety-first ratio against the investor's minimum return threshold over historical periods. If the ratio is low, the strategy's historical volatility relative to its return cushion means it would have failed the investor's goal too often. That output directly affects the go/no-go decision on a fund launch.

Interview questions

Bridgewater Associates Portfolio Analyst "A portfolio has an expected return of 10%, a standard deviation of 12%, and a client threshold return of 3.7%. Your manager asks you to calculate the shortfall probability and explain what the number actually tells us about the portfolio's risk profile."

Oaktree Capital Fixed Income Analyst "Client A needs a threshold return of 3.7% and Client B needs 7.5%. You have a portfolio with 10% expected return and 12% standard deviation. Would you recommend the same portfolio to both clients? Walk through your reasoning."

Morgan Stanley Wealth Management Investment Strategist "Your manager asks you to explain why the Sharpe ratio and the safety-first ratio use the same formula but can rank portfolios differently. What is the single condition under which they produce identical rankings, and why does it matter?"

One-line to use in your interview

Interviewers listen for industry-specific language. It signals you understand the concept, not just the definition. Use the plain English version to adapt it in your own words.

I use the safety-first ratio to rank portfolios by how far their expected return sits above the client minimum relative to how much the portfolio swings, because the highest-return portfolio is not always the safest when its volatility is disproportionately high.

In plain English

Do not just pick the portfolio with the best average return. Pick the one that sits farthest above the minimum acceptable level relative to how much it swings. A portfolio earning 10% with a 4% floor beats one earning 15% with a 12% floor when measured this way.