p = read.csv('stocks.csv')
## Calculate returns. Note that first column is dates
y = apply(log(p),2,diff)
## Specify portfolio value and VaR probability
portfolio_value = 1000
p = 0.01
using CSV, DataFrames;
p = CSV.read("stocks.csv", DataFrame);
## Convert prices of first two stocks to returns
y1 = diff(log.(p[:,1]));
y2 = diff(log.(p[:,2]));
y = hcat(y1,y2);
T = size(y,1);
value = 1000; # portfolio value
p = 0.01; # probability
y1 = y[,1] # select one asset
ys = sort(y1) # sort returns
op = ceiling(length(y1)*p) # p percent smallest, rounded up
VaR1 = -ys[op]*portfolio_value
print(VaR1)
ys = sort(y1) # sort returns
op = ceil(Int, T*p) # p percent smallest, rounding up
VaR1 = -ys[op] * value
println("Univariate HS VaR ", Int(p*100), "%: ", round(VaR1, digits = 3), " USD")
w = matrix(c(0.3,0.2,0.5)) # vector of portfolio weights
## The number of columns of the left matrix must be the same as the number of rows of the right matrix
yp = y %*% w # obtain portfolio returns
yps = sort(yp)
VaR2 = -yps[op]*portfolio_value
print(VaR2)
w = [0.3; 0.7] # vector of portfolio weights
yp = y * w # portfolio returns
yps = sort(yp)
VaR2 = -yps[op] * value
println("Multivariate HS VaR ", Int(p*100), "%: ", round(VaR2, digits = 3), " USD")
ES1 = -mean(ys[1:op])*portfolio_value
print(ES1)
using Statistics;
ES1 = -mean(ys[1:op]) * value
println("ES: ", round(ES1, digits = 3), " USD")
sigma = sd(y1) # estimate volatility
VaR3 = -sigma * qnorm(p) * portfolio_value
print(VaR3)
using Distributions;
sigma = std(y1); # estimate volatility
VaR3 = -sigma * quantile(Normal(0,1),p) * value
println("Normal VaR", Int(p*100), "%: ", round(VaR3, digits = 3), " USD")
sigma = sqrt(t(w) %*% cov(y) %*% w)[1] # portfolio volatility
## Note: the trailing [1] is to convert a single element matrix to float
VaR4 = -sigma * qnorm(p)*portfolio_value
print(VaR4)
sigma = sqrt(w'*cov(y)*w) # portfolio volatility
VaR4 = -sigma * quantile(Normal(0,1), p) * value
println("Portfolio normal VaR", Int(p*100), "%: ", round(VaR4, digits = 3), " USD")
library(QRM)
scy1 = (y1)*100 # scale the returns
res = fit.st(scy1)
sigma1 = unname(res$par.ests['sigma']/100) # rescale the volatility
nu = unname(res$par.ests['nu'])
## Note: We are removing the names of the fit.st output using unname()
VaR5 = - sigma1 * qt(df=nu,p=p) * portfolio_value
print(VaR5)
## Julia does not have a function for fitting Student-t data yet
## Currently: there exists Distributions.jl with fit_mle
## usage: Distributions.fit_mle(Dist_name, data[, weights])
##
## using Distributions;
## res = fit_mle(TDist, y1)
## nu = res.ν (this is the Greek letter nu, not Latin v)
## sigma = sqrt(nu/(nu-2))
## VaR5 = -sigma * quantile(TDist(nu), p) * value
sigma = sd(y1)
ES2 = sigma*dnorm(qnorm(p))/p * portfolio_value
print(ES2)
sigma = std(y1)
ES2 = sigma * pdf(Normal(0,1), (quantile(Normal(0,1), p))) / p * value
println("Normal ES: ", round(ES2, digits = 3), " USD")
VaR = -qnorm(p)
integrand = function(q){q*dnorm(q)}
ES = -sigma*integrate(integrand,-Inf,-VaR)$value/p*portfolio_value
print(ES)
using QuadGK;
VaR = -quantile(Normal(0,1), p)
integrand(x) = x * pdf(Normal(0,1), x)
ES3 = -sigma * quadgk(integrand, -Inf, -VaR)[1] / p * value
println("Normal integrated ES: ", round(ES3, digits = 3), " USD")
WE=20
for (t in seq(length(y1)-5,length(y1))){
t1=t-WE+1
window= y1[t1:t] # estimation window
sigma=sd(window)
VaR6 = -sigma * qnorm(p) * portfolio_value
print(VaR6)
}
WE = 20
for t in T-5:T
t1 = t-WE
window = y1[t1+1:t] # estimation window
sigma = std(window)
VaR6 = -sigma*quantile(Normal(0,1),p)*value
println("MA Normal VaR", Int(p*100), "% using observations ", t1, " to ", t, ": ",
round(VaR6, digits = 3), " USD")
end
lambda = 0.94
s11 = var(y1) # initial variance, using unconditional
for (t in 2:length(y1)){
s11 = lambda * s11 + (1-lambda) * y1[t-1]^2
}
VaR7 = -qnorm(p) * sqrt(s11) * portfolio_value
print(VaR7)
lambda = 0.94
s11 = var(y1) # initial variance
for t in 2:T
s11 = lambda * s11 + (1-lambda) * y1[t-1]^2
end
VaR7 = -sqrt(s11) * quantile(Normal(0,1), p) * value
println("EWMA VaR ", Int(p*100), "%: ", round(VaR7, digits = 3), " USD")
s = cov(y) # initial covariance
for (t in 2:dim(y)[1]){
s = lambda*s + (1-lambda)*y[t-1,] %*% t(y[t-1,])
}
sigma = sqrt(t(w) %*% s %*% w)[1] # portfolio vol
## Note: trailing[1] is to convert single element matrix to float
VaR8 = -sigma * qnorm(p) * portfolio_value
print(VaR8)
s = cov(y) # initial covariance
for t in 2:T
s = lambda * s + (1-lambda) * y[t-1,:] * (y[t-1,:])'
end
sigma = sqrt(w'*s*w) # portfolio vol
VaR8 = -sigma * quantile(Normal(0,1), p) * value
println("Two-asset EWMA VaR ", Int(p*100), "%: ", round(VaR8, digits = 3), " USD")
library(rugarch)
spec = ugarchspec(variance.model = list( garchOrder = c(1, 1)),
mean.model = list( armaOrder = c(0,0),include.mean = FALSE))
res = ugarchfit(spec = spec, data = y1)
omega = res@fit$coef['omega']
alpha = res@fit$coef['alpha1']
beta = res@fit$coef['beta1']
sigma2 = omega + alpha * tail(y1,1)^2 + beta * tail(res@fit$var,1)
VaR9 = -sqrt(sigma2) * qnorm(p) * portfolio_value
names(VaR9)="VaR"
print(VaR9)
## We use the ARCHModels package
using ARCHModels;
garch1_1 = fit(GARCH{1,1}, y1; meanspec = NoIntercept);
## In-sample GARCH VaR - Use the VaRs function
garch_VaR_in = VaRs(garch1_1, :0.01)
## Out-of-sample GARCH VaR - Use the predict function
cond_vol = predict(garch1_1, :volatility) # 1-day-ahead conditional volatility
garch_VaR_out = -cond_vol * quantile(garch1_1.dist, p) * value
println("GARCH VaR ", Int(p*100), "%: ", round(garch_VaR_out, digits = 3), " USD")