Chapter 4: Market Risk
Content
• Overview
• Variance-Covariance
• Historic method
• Monte Carlo
• Expected shortfall
Market risk
• Uncertainty of earnings on the trading portfolio due to
adverse market conditions
• This topic: to measure the potential loss to trading
portfolio
Trading portfolio
Methods to measure market risk
• Variance-covariance (RiskMetrics)
• Historic
• Monte Carlo simulation
• Expected shortfall
Measure market risk
• Use Value-at-Risk (VaR)
• The idea: the bank will suffer loss beyond this amount if
an extreme market condition occurs (corresponding to a
small probability).
Value at Risk (VaR)
• VaR at a probability P(Z ≤ z) = 5%
VaR
• If VaR = $20 million in a day at P(Z ≤ z) = 5%
• Interpretation:
• A 5% chance that the bank will lose ≥ $20million in a day
• A 95% chance that the bank will lose at most $20 million
Variance-Covariance
• Measure risk for one day
• Assume daily return is normally distributed
• One-day VaR measurement: also called Daily Earnings At
Risk (DEAR)
One-day VaR = Market Value × One-day Return Volatility
One-day VaR for shares
• A stock’s risk has two component
• Systematic risk (e.g., β in CAPM)
• Unsystematic risk
• If the bank holds the stock market portfolio (e.g., stock index)
• Systematic risk β = 1
• Unsystematic risk = 0
One-day VaR for shares
• Volatility:
• −2.33σ: with probability of 1%, daily return may be smaller
than -2.33σ
• −1.65σ: with probability of 5%, daily return may be smaller
than -1.65σ
Example
• A bank holds $100 million in market portfolio
• One-day expected market return =0, standard
deviation σ = 0.1
• Volatility = −2.33σ at probability 1%
• Calculate One-day VaR
Extension: If β ≠ 1
• If the bank holds a well-diversified portfolio but not the
market index: β ≠ 1, at probability 1%:
VaR = Market Value × β × (−2.33σ)
Extension: If E(r) ≠ 0
• If the portfolio expected return E(r) ≠ 0, at probability 1%:
VaR = Market Value × β × (E(r) − 2.33σ)
Example:
If E(r) ≠ 0
• E(r) = 0.135, σ = 0.244, at
probability 5%
DEAR for bonds
• DEAR = Market Value × (Price sensitivity × Adverse yield move)
missing averse movement in yield
Example
• Seven-year zero-coupon bond, yield = 8%, market value of
position $100 million
• Daily average return =0, standard deviation σ = 10 basis
points
• Calculate DEAR if yield increases at unlikely probability
1%, or P(Z ≥ z) = 1%
Example
DEAR for foreign exchange - FX
DEAR = Dollar value of position × FX volatility
• Dollar value of position = FX position × $ per unit of FX
• Daily exchange rate has standard deviation σ
• FX volatility: −2.33σ with probability 1%
Example
• A bank holds 10 million euros
• Current exchange rate: $1.1/€1, σ = 0.5%
• Calculate DEAR at probability 1%
N-day VaR
• Convert DEAR to VaR for N-day
• VaRN =DEAR×N0.5
Portfolio risk
• Portfolio with two assets: Stock (S) and Bond (B)
DEAR2P =DEAR2S +DEAR2B +2ρS,BDEARSDEARB
Portfolio risk
• Portfolio with three assets: Stock (S), Bond (B), and
Foreign Currency (F)
DEAR2P = DEAR2S +DEAR2B + DEAR2F
+ 2ρS,BDEARSDEARB
+ 2ρS,FDEARSDEARF
+ 2ρB,FDEARBDEARF
Pros and cons of Variance-Covariance method
• Pros: easy to calculate
• Cons: normal distribution assumption?
Historic method
• Calculate VaR using historical distribution of return
• With stock’s daily return:
• Calculate daily return
• Sort daily return from smallest to largest → return distribution
• Calculate VaR at the desired probability
Example
• Collect the most recent 500 daily returns of a stock
(approximately 2 years)
• Sort the return
• 5 observations (1% of 500) have return ≤ the first percentile return
• VaR = Market value × first percentile return
Pros and cons of Historic method
• Pros: No assumption of distribution is required
• Cons:
• Past distributions may not repeat in the future
(especially sharp downturn)
• Must go back to more distant past to have more
observations
→ less relevant to current market condition
Monte Carlo simulation
• A simple example
• Use historical data to obtain the parameters of security
return distribution (𝑅, σ)
• Randomly create a large number of daily returns using
a type of distribution (e.g., normal distribution) and the
parameters
• Sort the newly created returns from smallest to largest
• Calculate VaR at the desired probability
Expected Shortfall
Problem with VaR
• VaR is derived from the specific point on the probability
distribution → the minimum loss
• Does not provide information about the size of the loss
beyond that minimum point
Expected shortfall
• Expected shortfall is the average of VaRs, with each VaR
corresponding to a probability within the extreme market
condition
• Discrete distribution: ES = −E(∆V|∆V < −VaR)
• Continuous distribution: approximation with scaling
factor from t-distribution
Use 2.665σ instead of 2.33σ at probability 1%
Use 2.063σ instead of 1.65σ at probability 5%
Example: discrete
Example: continuous
• A bank holds $100 million in market portfolio
• One-day expected market return =0, standard
deviation σ = 0.1
• Calculate ES at probability 1%
Problem sets
• Chapter 15: 4, 6, 7, 9, 10, 11, 15, 16, 17, 19, 22, 23, 25, 26, 29