Farm Bureau Actuarial Conference

Overview of Discussion
Value at Risk (VaR) • How did VaR come about?
• What is VaR?
• How does one calculate VaR? What are the
approaches to calculating VaR?
• An example
• Can VaR be applied to insurance?
Kevin C. Ahlgrim
University of Illinois at Urbana-Champaign

Value at Risk - An Introduction

Value at Risk - A Definition
• Origin of VaR stems from derivatives debacles
• Value at risk is a statistical measure of
• Suppose the CFO of your company just read possible portfolio losses
some of the horror stories about derivatives – A percentile of the distribution of outcomes
• She asks you how much the company could • Value at Risk (VaR) is the amount of loss
possibly lose by using derivatives that a portfolio will experience over a set
– In theory, the company could lose everything period of time with a specified probability
• How can you communicate the extent of risk – Under “normal market conditions”
without a lecture about derivative instruments? • Thus, VaR depends on some time horizon
• Answer: Value at Risk and a desired level of confidence

A Note about Value at Risk Numbers

Value at Risk - An Example
• Let’s use a 5% • Note that the VaR depends on the choice of
probability and a one- R et urn D ist ribution time period (t) and a probability (x)
day holding period – Without knowing t and x, a value at risk
P ro bability

• VaR is the one day loss number is meaningless

that will be exceeded • Longer t and lower x lead to higher VaR
V aR
only 5% of the time
• Implicit assumption is that the portfolio is
• It’s the tail of the return
constant over the time period t
00
$0
0)
00

,0

distribution
00
0,
10

$1

– t should be chosen to reflect portfolio turnover

($

• In the example, the P ortfolio Ga ins/L osses

VaR is about $60,000 • x should correspond with the risk tolerance
of the portfolio owner
 First - Identify the Market Factors Methods of Calculating VaR
• There are 3 methods to calculate VaR, but • Historical simulation
the first step is to identify the “market – Apply recent experience to current portfolio
factors” • Variance-covariance method
• Market factors are the variables that impact – Assume a normal distribution and use the
the value of the portfolio statistical properties to find VaR
– Stock prices, interest rates, exchange rates, etc. • Monte Carlo Simulation
• The different approaches to VaR are based – Generate scenarios to determine changes in
on how the market factors are modeled portfolio value

Historical Simulation Example of Historical Simulation

• Assume a one-day holding period and 5%
• Use the changes in the market factors over probability
the last 100 days (100 days is an example) • Suppose that a portfolio has two assets, a
• Translate the historical experience of the one-year T-bill and a 30-year T-bond
market factors into percentage changes • First, gather the 100 days of market info
• Apply the 100 percentage changes of the Date T-Bond Value % Change T-Bill Value % Change
12/10/xx 102 - 97 -
market factors to beginning portfolio values 12/9/xx 100 2.00% 98 -1.02%
12/8/xx 97 3.09% 98 0.00%
• Rank the 100 resulting values : : : : :
: : : : :
• VaR is the required percentile rank 9/2/xx 103 -2.91% 96 2.08%
9/1/xx 103 0.00% 97 -1.03%

Example of Historical Simulation (p.2) Example of Historical Simulation (p.3)

• Apply all changes to the current value of • Rank the resulting 100 portfolio values
assets in the portfolio • The 5th lowest portfolio value is the VaR
• T-bond value = 102 x % change Po rtfo li o
R an k D ate V al u e
• T-bill value = 97 x % change 1 11/ 12/ 97 195. 45
2 12/ 1/ 97 196. 24
T-Bond Modeled T-Bill Modeled Portfolio 3 10/ 17/ 97 197. 13
Date % Change Value % Change Value Value 4 10/ 13/ 97 197. 60
12/9/xx 2.00% 104.04 -1.02% 96.01 200.05 5 9/ 1/9 7 198. 00
12/8/xx 3.09% 105.15 0.00% 97.00 202.15 : : :
: : : : : : : : :
: : : : : : 99 12/ 8/ 97 202. 15
9/2/xx -2.91% 99.03 2.08% 99.02 198.05 100 9/ 25/ 97 203. 00
9/1/xx 0.00% 102.00 -1.03% 96.00 198.00
 Notes about Historical Simulation
Variance-Covariance Method
• Historical simulation is relatively easy to do
– Only requires knowing the market factors and • Assume all market factors follow a
having the historical information multivariate normal distribution
• Note that correlations between the market • Under this assumption, the distribution of
factors are implicit in this method because portfolio gains/losses can be determined
we are using historical information with statistical properties
– In our example, short bonds and long bonds
would typically move in the same direction
• From the distribution of gains and losses,
choose the required percentile to find VaR
• Be aware of market trends

Example of Variance-Covariance
Method Notes to Variance-Covariance
• Using the previous example, if both assets Method
are normally distributed, than the portfolio • This method is also known as the analytic
value is bivariate normal method
• We can find the VaR as follows: • It is conceptually more difficult given the
σ 2portfolio = X s2σ 2s + X l2σ 2l + 2 ρ sl X s X l σ sσ l need for multivariate analysis
VaR = 1.645 × σ portfolio • Explaining the method to management may
be as difficult as a lecture on derivatives
s = T - bill (s for short - term) , l = T - bond
X i = Current value of asset i
1.645 is 5% critical value for standard normal

Monte Carlo Simulation Notes to Monte Carlo Simulation

• In Monte Carlo simulation, we can specify • If the distribution of market factors is
the individual distributions of the future normal, Monte Carlo simulation and the
values of the market factors in the portfolio variance-covariance method will give
• Generate random samples from the assumed similar VaR
distribution • Initial setup is costly, but thereafter Monte
• Determine the final value of the portfolio Carlo simulation can be efficient
• Rank the portfolio values and find the
appropriate percentile to find VaR
Similarity to Actuarial Techniques Value at Risk in Insurance
• Risk Based Capital (RBC) requirements
– Apply factors to assets and liabilities that differ • Applying VaR to insurance is difficult
based on the risk of the instrument because of the time horizon in insurance
– Product represents required capital – It is much longer than is typical for most VaR
– C-1 factors are based on a VaR approach applications
• Dynamic Financial Analysis (DFA), • Historical simulation may hide risks
Dynamic Financial Condition Analysis • VaR is becoming a hot topic in insurance
models the distribution of profits/losses
research
– Most similar to Monte Carlo simulation
approach of VaR