Quantitative Risk Management in R: Value-At-Risk and Expected Shortfall

QUANTITATIVE RISK MANAGEMENT IN R
Value-at-risk and
expected shortfall
Quantitative Risk Management in R
Value-at-risk (VaR)
● Consider the distribution of losses over a fixed time
period (day, week, etc.)
● 𝛼-VaR is the 𝛼-quantile of the loss distribution
● 𝛼 known as confidence level (e.g. 95%, 99%)
● Should lose no more than 𝛼-VaR with probability 𝛼

95% VaR illustrated

Loss Distribution
Mean loss = -2.4
0.25
95% VaR = 1.6
95% ES = 3.3
0.20
probability density
0.10 0.15
0.05
5% probability
0.0
-10 -5 0 5 10
Expected shortfall (ES)

● Increasingly important in banking regulation
● Tail VaR (TVaR), conditional VaR (CVaR) or expected
shortfall (ES)
● 𝛼-ES is expected loss given that loss exceeds 𝛼-VaR
● Expectation of tail of distribution

95% ES illustrated
Loss Distribution
Mean loss = -2.4
0.25
95% VaR = 1.6
95% ES = 3.3
0.20
probability density
0.10 0.15
0.05
5% probability
0.0
-10 -5 0 5 10
Let’s practice!
International equity
portfolio example
International equity portfolio

● Imagine a UK investor who has invested her wealth:
● 30% FTSE, 40% S&P 500, 30% SMI
● 5 risk factors: FTSE, S&P 500 and SMI indexes, GBP/USD and
GBP/CHF exchange rate
> riskfactors <- merge(FTSE, SP500, SMI, USD_GBP, CHF_GBP, all = FALSE)
["/2012-12-31", ]
Displaying the risk factors

> plot.zoo(riskfactors)
Historical simulation
● Simple method that is widely used in financial industry
● Resample historical risk-factor returns and examine their
effect on current portfolio
● Loss operator shows effect of different risk-factor returns
on the portfolio
● Loss operator functions will be provided in the exercises
Empirical estimates of VaR and ES

> losses <- rnorm(100)
> losses_o <- sort(losses, decreasing = TRUE)
> head(losses_o, n = 8)
[1] 1.836163 1.775163 1.745427 1.614479 1.602120 1.590034 1.483691
1.408354 
> quantile(losses, 0.95)
95%
1.590638
> qnorm(0.95)
[1] 1.644854
> mean(losses[losses > quantile(losses, 0.95)])

[1] 1.714671
> ESnorm(0.95)
[1] 2.062713
Let’s practice!
Option portfolio and

Black-Scholes
European options and Black-Scholes

● European call option: gives right but not obligation to buy
stock for price K at time T
● European put option: gives right but not obligation to sell
stock for price K at T
● Value at time t < T depends on:
● Stock price S, time to maturity T-t, interest rate r,
annualized volatility 𝜎 or sigma
● Pricing by Black-Scholes formula

Pricing a first call option

> K <- 50
> T <- 2
> t <- 0
> S <- 40
> r <- 0.005
> sigma <- 0.25
> Black_Scholes(t, S, r, sigma, K, T, "call")
[1] 2.619183
> Black_Scholes(t, S, r, sigma*1.2, K, T, "call")
[1] 3.677901
● Price increases with volatility

● Option above is in-the-money
Implied volatility X needs change

● Volatility not directly observable
● Market participants use implied volatility, the value of
volatility implied by quoted option price
The VIX index

> plot(VIX)
Let’s practice!
for the option example
● Portfolio: single European call option on equity index
● Consider losses and profits over one day
● Changes to index value S, implied volatility 𝜎 and interest

rate r affect value of portfolio
● We consider S and 𝜎 (and assume r stays constant)
● Create loss operator taking S and 𝜎 as input and giving

the loss or profit as output
Estimating VaR and ES

● Apply loss operator lossop() to historical log-returns of
S&P 500 and VIX to get simulated losses
● Estimate VaR by sample quantile as before
● Estimate ES by average of losses exceeding VaR estimate
Let’s practice!
Wrap up
Not the end of the story…

Consider two things:
1. Can we improve risk sensitivity of VaR and ES estimates?
● Filtered historical simulation, GARCH models, EWMA
volatility filters
2. Can we improve simple empirical estimates of VaR and ES?
● Parametric tail models, heavy-tailed distributions,
extreme value theory
Thanks for taking the course!

Quantitative Risk Management in R: Value-At-Risk and Expected Shortfall

Uploaded by

Copyright:

Available Formats

Quantitative Risk Management in R: Value-At-Risk and Expected Shortfall

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Quantitative Risk Management in R: Value-At-Risk and Expected Shortfall

Uploaded by

Copyright:

Available Formats

QUANTITATIVE RISK MANAGEMENT IN R

● 𝛼-VaR is the 𝛼-quantile of the loss distribution

● 𝛼 known as confidence level (e.g. 95%, 99%)

● Should lose no more than 𝛼-VaR with probability 𝛼

95% VaR illustrated

95% VaR = 1.6

Expected shortfall (ES)

● 𝛼-ES is expected loss given that loss exceeds 𝛼-VaR

● Expectation of tail of distribution

95% VaR = 1.6

International equity portfolio

Displaying the risk factors

Empirical estimates of VaR and ES

> mean(losses[losses > quantile(losses, 0.95)])

Option portfolio and

European options and Black-Scholes

● Pricing by Black-Scholes formula

Pricing a first call option

● Price increases with volatility

Implied volatility X needs change

The VIX index

● Changes to index value S, implied volatility 𝜎 and interest

● We consider S and 𝜎 (and assume r stays constant)

● Create loss operator taking S and 𝜎 as input and giving

Estimating VaR and ES

Not the end of the story…

Thanks for taking the course!

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.