Actuarial and nancial applications of simulation

Actuarial and nancial applications of simulation

Math 276 Actuarial Models Spring 2008 semester

EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions

EA Valdez University of Connecticut - Storrs Lecture Weeks 6 and 7

Simulation results

page 1

Modeling stock prices In nance, we are always interested in the return on stocks. The Normal distribution is a typical distribution model for return on the stock; indeed equivalent to modeling the value of the stock as a lognormal distribution. Assume that the return on the stock is normally distributed with annual mean and annual standard deviation . Denote by St the value of the asset at time t and St +t denoting the the value t periods later. Thus, the percentage change (or return) of the value of the stock between times t and t + t is approximated by log St log St +t = log St St +t = log (1 + rt ) rt ,

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

where rt = (St St +t )/St .

page 2

The lognormal distribution (and geometric diffusions)

Actuarial and nancial applications of simulation EA Valdez

Another way to write the stock price at time t + t is St +t = St exp t + Z t , where Z is standard normal N(0, 1). If you know diffusion processes, this is the discrete analogue of the geometric diffusion: dS = dt + dB , S where dB is a Brownian motion (or Weiner) process with dB = Z dt .

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 3

Illustrative example

Actuarial and nancial applications of simulation EA Valdez

Consider a stock paying no dividends with a volatility = 0.05 per annum and with an expected return of = 0.10 per annum with continuous compounding. The stock price process can be written as dS = 0.10dt + 0.20dB or (in the discrete sense) with S small interval of time S = 0.15t + 0.20Z t . S The gures in the following page demonstrate this price process (by simulation) for different time intervals: year (t = 1), month (t = 1/12), week (t = 1/52), and day (t = 1/365). Here we assume the initial stock price is 100.

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 4

R code to generate the stock price process The following is a routine in R to generate the stock price process. Function is called simstock.R.
# function to generate (discrete) stock price process simstock <- function(mu,sigma,dt,n.yr){ n.gen <- (1/dt)*n.yr S.vector <- rep(0,n.gen+1) S0 <- 100 S.vector[1] <- S0 for(i in 2:(n.gen+1)) { zrandom <- rnorm(1) ds <- S.vector[i-1]*(mu*dt + sigma*zrandom*sqrt(dt)) S.vector[i] <- S.vector[i-1] + ds } # output S.vector }

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Then run the following commands:

> > > > source("C:\\...\\Math276-Spring2008\\Rcodes-2008\\Week67\\simstock.R") st.yr <- simstock(0.1,0.2,1,20) year <- 0:20 plot(st.yr~year,type="l",ylab="stock price")

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions

> st.mo <- simstock(0.1,0.2,1/12,20) > month <- 0:(12*20) > plot(st.mo~month,type="l",ylab="stock price") > st.wk <- simstock(0.1,0.2,1/52,20) > week <- 0:(52*20) > plot(st.wk~week,type="l",ylab="stock price") > st.dy <- simstock(0.1,0.2,1/365,20) > day <- 0:(365*20) > plot(st.dy~day,type="l",ylab="stock price")

Simulation results

page 5

Pictorial illustration of the stock price process

Actuarial and nancial applications of simulation EA Valdez

700

600

400

Modeling stock prices

The lognormal distribution Illustrative example

500

stock price

stock price

300

400

R code to generate the stock price process Pictorial illustration Lognormal property

200

300

200

100

100

Illustration of generating distribution of stock price Distribution graph

10 year

15

20

50

100 month

150

200

Generating distribution of a portfolio of assets

Modeling aggregate claims

1000

Distribution assumptions Case illustration

400

800

Simulation

stock price

stock price

300

Modeling in life insurance

The Gompertz lifetime distribution

400

600

200

Simulating from Gompertz Simulating the loss

200

100

Parameter assumptions Simulation results

0 2000 4000 day 6000

200

400 week

600

800

1000

page 6

Lognormal property of stock prices In the geometric Brownian motion, the change in the log S between time 0 and T has a Normal distribution with

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

log ST log S0 N ( 2 /2)T , 2 T , where S0 is the initial stock price while ST is the stock price T periods later. This is equivalent to ST having a lognormal distribution: log ST N log S0 + ( 2 /2)T , 2 T . It is straightforward to show that the mean is given by E(ST ) = S0 eT , and the variance is
2 2T Var(ST ) = S0 e e
2

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

1 .
page 7

R code to illustrate simulating the stock price at time T The following is a routine in R to generate the stock price at time T . Function is called simstprice.R.
# function to generate stock price/value at time T simstprice <- function(n.gen,S0,mu,sigma,T){ zrandom <- rnorm(n.gen) l.mu <- log(S0) + (mu - 0.5*sigma^2)*T l.sigma <- sigma*sqrt(T) nrandom <- l.mu + l.sigma*zrandom S.vector <- exp(nrandom) # output S.vector }

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph

Then run the following commands:

> > > > source("C:\\...\\Math276-Spring2008\\Rcodes-2008\\Week67\\simstprice.R") out1 <- simstprice(1000,40,0.1,0.3,1/2) source("C:\\...\\Math276-Spring2008\\Rcodes-2008\\Week67\\Data.SummStats.R") Data.SummStats(out1) Value Number 1000.00 Mean 41.65 5th Q 28.51 25th Q 35.26 Median 40.82 75th Q 46.71 95th Q 58.28 Variance 80.92 StdDev 9.00 Minimum 21.00 Maximum 77.67 Skewness 0.61 Kurtosis 0.59

Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 8

Graphical representation of the distribution

> hist(out1,br=25,xlab="initial price = $40",ylab="stock price after half year",main="1,000 simulations from logNormal with mu=30%, sigma=10%",freq=FALSE) > lines(density(out1),xlim=c(20,80),col="blue")

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution

1,000 simulations from logNormal with mu=30%, sigma=10%

Illustrative example R code to generate the stock price process Pictorial illustration

0.05

Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

stock price after half year

0.04

0.03

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

0.02

0.01

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions

0.00
20

30

40

50 initial price = $40

60

70

80

Simulation results

page 9

R code to illustrate simulating a portfolio of assets The following is a routine in R to generate the value of a portfolio of stocks at time T . Function is called simportfolio.R.
# # # # function to generate a portfolio of securities at time T the variables required: n.gen (number to simulate), S0.vector (vector of initial values of the stocks), mu.vector/sigma.vector (self-explanatory), n.holdings (the number of holdings for each corresponding stock), T (valuation date)

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

simportfolio <- function(n.gen,S0.vector,mu.vector,sigma.vector,n.holdings,T){ n.port <- length(S0.vector) S.matrix <- matrix(0,nrow=n.gen,ncol=n.port) for (i in 1:n.port){ zrandom <- rnorm(n.gen) l.mu <- log(S0.vector[i]) + (mu.vector[i] - 0.5*sigma.vector[i]^2)*T l.sigma <- sigma.vector[i]*sqrt(T) nrandom <- l.mu + l.sigma*zrandom S.matrix[,i] <- exp(nrandom) } # multiply with the holdings S.matrix <- t(n.holdings*t(S.matrix)) # output # this actually sums the rows with each column referring to one particular stock rowSums(S.matrix) }

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Then run the following commands:

> source("C:\\...\\Math276-Spring2008\\Rcodes-2008\\Week67\\simportfolio.R") > S0.vector <- c(200,25,50,45,60,12,100,135,98,20) > mu.vector <- c(.05,.1,.12,.09,.1,.2,.15,.08,.11,.25) > sigma.vector <- c(.1,.2,.2,.1,.2,.3,.25,.1,.15,.3) > n.holdings <- 10*(5:14) > out1 <- simportfolio(1000,S0.vector,mu.vector,sigma.vector,n.holdings,5) > min(out1) [1] 75287.67 > max(out1) [1] 238245.3

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 10

Graphical representation of the resulting portfolio distribution

> hist(out1,br=25,xlab="",ylab="portfolio value",main="Distribution of Portfolio Value after 5 years",freq=FALSE,xlim=c(50000,250000)) > lines(density(out1),xlim=c(50000,250000),col="blue")

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example

Distribution of Portfolio Value after 5 years

0.0e+00 5.0e06 1.0e05 1.5e05 2.0e05 2.5e05

R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

portfolio value

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

50000

100000

150000

200000

250000

page 11

Some statistics on the resulting portfolio distribution

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example

> source("C:\\...\\Math276-Spring2008\\Rcodes-2008\\Week67\\Data.SummStats.R") > Data.SummStats(out1) Value Number 1.000e+03 Mean 1.181e+05 5th Q 9.157e+04 25th Q 1.042e+05 Median 1.167e+05 75th Q 1.279e+05 95th Q 1.530e+05 Variance 3.533e+08 StdDev 1.880e+04 Minimum 7.529e+04 Maximum 2.382e+05 Skewness 9.300e-01 Kurtosis 2.080e+00 >

R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 12

The distribution of the total claim amount In general (property/casualty) insurance, we often study the distribution of the aggregate claim amount dened by S = X1 + X2 + + XN , where Xi is the amount of the i -th claim and N refers to the number of claims. Here we are referring to the total claims only for a xed period, e.g. one year, though this can be a stochastic process (over time). The standard assumptions in the model are:
1

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

the claim amounts Xi are i.i.d. (independent and identically distributed) random variables; and the claim amounts X1 , X2 , . . . and the claim count N are all independent.

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions

The aggregate sum S has what we call a compound distribution, and in many instances, it is not possible to derive explicit form of its distribution.

Simulation results

page 13

Claim size and claim count distribution assumptions

Actuarial and nancial applications of simulation EA Valdez

For purposes of illustration, we shall assume the following:

Modeling stock prices

The claim size Xi has the Pareto(, ) distribution with CDF F (x ) = 1 x +

The lognormal distribution Illustrative example R code to generate the stock price process

, for x > 0.

Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph

The claim count N is assumed to have a Poisson() distribution.

Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Note that to simulate from the Pareto, it can be shown that, using the inverse transform method, the following generates a Pareto(, ) random variable: X = (1 U )1/ 1 .

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 14

Case illustration

Actuarial and nancial applications of simulation EA Valdez

Consider a portfolio of 10,000 automobile insurance policies with the following assumptions:
The period is exactly one year where each policy pays an annual premium of = $500. Expenses include an overhead (or xed) expense of $500,000 and a per policy expense of $2.50. The aggregate claims distribution assume that claim size has a Pareto with = 625 and = 1.5 and claim count has a Poisson with = 900.

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration

The Prot/Loss (P/L) for this insurance portfolio is clearly Premiums - (Claims + Expenses) where:
Premiums: 10, 000 = 10, 000(500) = 5, 000, 000 Claims: S = X1 + X2 + + XN Expenses: 500, 000 + 2.5 10, 000 = 525, 000

Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 15

R code to illustrate simulating the aggregate claims The following is a routine in R to generate the aggregate claim amount for a portfolio of auto insurance policies. Function is called simillustrate.R.
# function to simulate the aggregate claims using the illustration simillustrate <- function(n.gen,ptheta,palpha,nlambda){ n.claims <- rpois(n.gen,nlambda) S.vector <- rep(0,n.gen) for (i in 1:n.gen) { urandom <- runif(n.claims[i]) claims <- ptheta * ((1-urandom)^(-1/palpha)-1) S.vector[i] <- sum(claims) } # output S.vector }

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Then run the following commands:

> > > > source("C:\\...\\Math276-Spring2008\\Rcodes-2008\\Week67\\simillustrate.R") out1 <- simillustrate(10000,625,1.5,900) source("C:\\...\\Math276-Spring2008\\Rcodes-2008\\Week67\\Data.SummStats.R") Data.SummStats(out1) Value Number 1.000e+04 Mean 1.130e+06 5th Q 8.545e+05 25th Q 9.566e+05 Median 1.048e+06 75th Q 1.173e+06 95th Q 1.532e+06 Variance 3.618e+11 StdDev 6.015e+05 Minimum 6.710e+05 Maximum 2.487e+07 Skewness 2.076e+01 Kurtosis 6.096e+02

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 16

Histograms of the aggregate claims distribution

> hist(out1/1000000,br=50,xlab="in millions",ylab="frequency",main="Distribution of Aggregate Claims",freq=FALSE) > min(log(out1)) [1] 13.41646 > max(log(out1)) [1] 17.02933 > hist(log(out1),br=50,xlab="in logarithm",ylab="frequency",main="Distribution of Aggregate Claims",freq=FALSE,xlim=c(13,17)) >

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property

Distribution of Aggregate Claims

Distribution of Aggregate Claims

Illustration of generating distribution of stock price Distribution graph

1.0

2.5

Generating distribution of a portfolio of assets

0.8

2.0

Modeling aggregate claims

Distribution assumptions Case illustration

frequency

frequency

1.5

0.6

Simulation

0.4

1.0

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions

0.2

0.5

0.0

10

15 in millions

20

25

0.0

Simulation results

13

14

15 in logarithm

16

17

page 17

Prot/loss analysis
> pl <- 5000000 - (out1 + 525000) > Data.SummStats(pl) Value Number 1.000e+04 Mean 3.345e+06 5th Q 2.943e+06 25th Q 3.302e+06 Median 3.427e+06 75th Q 3.518e+06 95th Q 3.620e+06 Variance 3.618e+11 StdDev 6.015e+05 Minimum -2.040e+07 Maximum 3.804e+06 Skewness -2.076e+01 Kurtosis 6.096e+02

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph

Profit/Loss Distribution

Generating distribution of a portfolio of assets

1.2

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

1.0

frequency

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz

0.4

0.6

0.8

Simulating the loss Parameter assumptions

0.0

0.2

Simulation results

20

15

10

5
page 18

in millions

Introduction - life insurance models The actuarial equivalence principle has one main drawback: it simply looks at the mean/average of the loss-at-issue distribution. Alternative is to also examine the variability of this loss, but this does not give the complete picture of the loss distribution. A much better alternative is to examine the loss distribution itself. In many cases, it is impossible to derive explicit form of the loss distribution. Simulating the loss distribution is one method to do it main drawback is it may require computer-intensive calculations. We demonstrate this only for a whole life insurance policy issued to a single person - in practice, you would be doing this for a portfolio of insurance contracts.

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 19

The Gompertz lifetime distribution

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

Assume that mortality follows the Gompertz with force of mortality x = Bc x , where B and c are constants satisfying B > 0 and c > 1. It is easy to show that for an issue age x , its future lifetime Tx follows the survival pattern Bc x STx (t ) = P (Tx > t ) = exp ct 1 log(c ) for t 0. ,

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 20

Simulating from Gompertz

Actuarial and nancial applications of simulation EA Valdez

We can use the inverse transform method to simulate from Gompertz. Begin with a random number U , generate a Gompertz lifetime, say T , from the following equation: exp Bc cT 1 log(c )
x

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price

= U,

Distribution graph Generating distribution of a portfolio of assets

or equivalently, we have T = 1 log(c ) log(U ) log 1 . log(c ) Bc x

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

Running this procedure m (number of simulations) times, we can then have a simulated distribution of the Gompertz lifetime.

page 21

Simulating the loss-at-issue

Actuarial and nancial applications of simulation EA Valdez

With a simulated value of T , we can then simulate a value of the present value of the loss-at-issue. For example, in a (fully continuous) whole life insurance contract, we have , L0 = bT v T a T where v = 1/(1 + i ) = e is the discount factor, is the annual premium assumed to be payable continuously throughout the year, and bT is the amount of insurance payable at death. Again, run this procedure for m number of times to get a simulated distribution of the loss-at-issue.

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 22

Simulating the loss after k years When computing reserves, we need to evaluate the loss at that point. Suppose we are interested in the loss after k years, then it can be shown that the simulated lifetime for the person who is then aged x + k is T = log(c ) log(U ) 1 log 1 , log(c ) Bc x +k

Actuarial and nancial applications of simulation EA Valdez

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

where U is U (0, 1) generated value. For the same (fully continuous) whole life insurance contract, we would have the loss after k years evaluated as , Lk = bT v T a T where T is the future lifetime of the person x who is now aged x + k .

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 23

Parameter assumptions

Actuarial and nancial applications of simulation EA Valdez

To illustrate, we assume the following Gompertz parameter values: B = 0.0000429 and c = 1.1070839. In addition, benet amount is $100, premium is $0.0095 per $1 of insurance, and i = 5%. Number of simulations: 50,000. Apart from calculating the losses at issue, we also calculate reserves (or losses) at the end of 10 years. The R routine is called Gompertz.SimulationT.R - too long to print in these slides; but is available on the website.

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 24

Some summary statistics of the simulation results

Actuarial and nancial applications of simulation EA Valdez

> source("C:\\...\\Math276-Spring2008\\Rcodes-2008\\Week67\\Gompertz.SimulationT.R") Value Number 50000.00 Mean 41.19 5th Q 18.90 25th Q 34.41 Median 42.97 75th Q 49.71 95th Q 57.16 Variance 136.32 StdDev 11.68 Minimum 0.00 Maximum 71.08 Skewness -0.74 Kurtosis 0.40 Value Number 50000.00 Mean -0.18 5th Q -12.12 25th Q -8.90 Median -4.79 75th Q 2.82 95th Q 28.04 Variance 212.20 StdDev 14.57 Minimum -15.75 Maximum 99.99 Skewness 2.79 Kurtosis 10.19

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 25

- continued

Actuarial and nancial applications of simulation EA Valdez

Value Number 50000.00 Mean 31.74 5th Q 10.85 25th Q 24.90 Median 33.18 75th Q 39.83 95th Q 47.26 Variance 119.26 StdDev 10.92 Minimum 0.01 Maximum 59.04 Skewness -0.54 Kurtosis -0.14 Value Number 50000.00 Mean 10.16 5th Q -7.56 25th Q -2.36 Median 4.20 75th Q 15.98 95th Q 50.89 Variance 353.57 StdDev 18.80 Minimum -12.77 Maximum 99.95 Skewness 1.93 Kurtosis 4.21

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Modeling aggregate claims

Distribution assumptions Case illustration Simulation

Modeling in life insurance

The Gompertz lifetime distribution Simulating from Gompertz Simulating the loss Parameter assumptions Simulation results

page 26

Graphical displays of the simulation results

Distribution of T(30) Distribution of Lossatissue

Actuarial and nancial applications of simulation EA Valdez

0.00 0.02 0.04 0.06

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

frequency

0.00

10

30 t.30

50

70

frequency

0.02

20

20

40 loss.30

60

80

Distribution of T(40)

Distribution of Loss at 10 yrs

Modeling aggregate claims

Distribution assumptions Case illustration

0.04

0.04

Simulation

frequency

frequency

0.02

Modeling in life insurance

0.02

The Gompertz lifetime distribution Simulating from Gompertz

0.00

0.00

Simulating the loss Parameter assumptions Simulation results

10

20

30 t.40

40

50

60

20

40

60

80

100

loss.40

page 27

Graphical displays of simulating repeatedly

T(30) loss at issue

Actuarial and nancial applications of simulation EA Valdez

0.00 0.02 0.04 0.06

Modeling stock prices

The lognormal distribution Illustrative example R code to generate the stock price process Pictorial illustration Lognormal property Illustration of generating distribution of stock price Distribution graph Generating distribution of a portfolio of assets

Density

0.00

20

40

60

Density

0.02

20

20

40

60

80

N = 50000

Bandwidth = 1.182

N = 50000

Bandwidth = 0.9113

T(40)

loss after 10 yrs

Modeling aggregate claims

Distribution assumptions Case illustration

0.04

Simulation

Density

Density

0.02

Modeling in life insurance

0.02

The Gompertz lifetime distribution Simulating from Gompertz

0.00

0.00

Simulating the loss Parameter assumptions Simulation results

10

20

30

40

50

60

20

20

40

60

80

N = 50000

Bandwidth = 1.124

N = 50000

Bandwidth = 1.397

page 28