descriptive statistics for financial time series© eric zivot 2006 descriptive statistics for...
TRANSCRIPT
![Page 1: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/1.jpg)
© Eric Zivot 2006
Descriptive Statistics for Financial Time Series
Econ 424Winter 201Eric Zivot
Updated: January 26, 2015
![Page 2: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/2.jpg)
Data for Examples
© Eric Zivot 2006
# Load libraries> library(tseries)> library(PerformanceAnalytics)
# Get adjusted closing price data from Yahoo!> MSFT.prices = get.hist.quote(instrument="msft", start="1998-01-01",+ end="2012-05-31", quote="AdjClose",+ provider="yahoo", origin="1970-01-01",+ compression="m", retclass="zoo")
> SP500.prices = get.hist.quote(instrument="^gspc", start="1998-01-01",+ end="2012-05-31", quote="AdjClose",+ provider="yahoo", origin="1970-01-01",+ compression="m", retclass="zoo")
• Note: MSFT.prices and SP500.prices are “zoo” objects. “zoo” is a special time series class (from the zoo package) that is very useful for financial data.
• See the document “Working with Financial Time Series Data in R” on the class syllabus page.
![Page 3: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/3.jpg)
Price Data
© Eric Zivot 2006
> plot(MSFTSP500.prices, main="Adjusted Closing Prices", lwd=2, col="blue")
![Page 4: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/4.jpg)
Monthly Continuously Compounded Returns
© Eric Zivot 2006
Q: What common features do you see?
![Page 5: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/5.jpg)
© Eric Zivot 2006
MSFT and S&P 500 tend to move together
S&P 500 has much lower volatility than MSFT
![Page 6: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/6.jpg)
Compare Cumulative Returns: Equity Curve
© Eric Zivot 2006
![Page 7: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/7.jpg)
Are Monthly Returns Gaussian White Noise?
© Eric Zivot 2006
> set.seed(123)> gwn = rnorm(length(MSFT),mean=mean(MSFT),sd=std(MSFT))
![Page 8: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/8.jpg)
Estimating the pdf: Histogram
© Eric Zivot 2006
> hist(MSFT,main="Histogram of MSFT monthly cc returns“,
+ col=“slateblue1”)
![Page 9: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/9.jpg)
© Eric Zivot 2006
> hist(gwn,main="Histogram of simulated Gaussian data“,
+ col=“slateblue1”)
Note: Simulated data has the same mean and variance as the returns on Microsoft
![Page 10: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/10.jpg)
© Eric Zivot 2006
> hist(SP500,main="Histogram of SP500 monthly cc returns“,
+ col=“slateblue1”)
![Page 11: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/11.jpg)
© Eric Zivot 2006
Note: MSFT has larger SD (volatility) than S&P 500
![Page 12: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/12.jpg)
© Eric Zivot 2006
> MSFT.density = density(MSFT)> plot(MSFT.density,type="l",xlab="monthly return",+ ylab="density estimate",main="Smoothed histogram for MSFT+ monthly cc returns“, col=“orange”, lwd=2)
![Page 13: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/13.jpg)
© Eric Zivot 2006
> hist(MSFT,main="Histogram and smoothed density of MSFT+ monthly returns", probability=T, col=“slateblue1”,+ ylim=c(0,5))> points(MSFT.density, type="l“, col=“orange”, lwd=2)
![Page 14: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/14.jpg)
Computing quantiles
© Eric Zivot 2006
> quantile(MSFT.ret.mat)0% 25% 50% 75% 100%
-0.421081 -0.050019 0.008343 0.055535 0.342188
1% and 5% quantilesare used for Value-at-Risk calculations
# 1% and 5% empirical quantiles> quantile(MSFT.ret.mat,probs=c(0.01,0.05))
1% 5% -0.2110 -0.1473
# compare to 1% and 5% normal quantiles> qnorm(p=c(0.01,0.05), mean=mean(MSFT.ret.mat), + sd=sd(MSFT.ret.mat))[1] -0.2291 -0.1608# SP500 empirical and normal quantiles> quantile(SP500.ret.mat,probs=c(0.01,0.05))
1% 5% -0.12846 -0.08538 > qnorm(p=c(0.01,0.05), mean=mean(SP500.ret.mat),+ sd=sd(SP500.ret.mat))[1] -0.11107 -0.07804
![Page 15: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/15.jpg)
Monthly VaR Using Empirical Quantiles
© Eric Zivot 2006
> q.01 = quantile(MSFT.ret.mat, probs=0.01)> q.05 = quantile(MSFT.ret.mat, probs=0.05)> q.01
1% -0.211 > q.05
5% -0.1473
# Monthly VaR on $100,000 investment> VaR.01 = 100000*(exp(q.01) - 1)> VaR.05 = 100000*(exp(q.05) - 1)> VaR.01
1% -19020
> VaR.055%
-13694
![Page 16: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/16.jpg)
Summary Statistics
© Eric Zivot 2006
> mean(MSFT.ret.mat)MSFT
0.004127
> var(MSFT.ret.mat)MSFT
MSFT 0.01005
> sd(MSFT.ret.mat)MSFT
0.1003
> skewness(MSFT.ret.mat)[1] -0.09073
> kurtosis(MSFT.ret.mat)[1] 2.082
kurtosis() function is in package PerformanceAnalyticsand computes excess kurtosis
Skewness() function is in package PerformanceAnalytics
![Page 17: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/17.jpg)
Summary Statistics by Column
© Eric Zivot 2006
> apply(MSFTSP500.ret.mat, 2, mean)MSFT SP500
0.004127 0.001687
> apply(MSFTSP500.ret.mat, 2, sd)MSFT SP500
0.10026 0.04847
> apply(MSFTSP500.ret.mat, 2, skewness)MSFT SP500
-0.09073 -0.73988
> apply(MSFTSP500.ret.mat, 2, kurtosis)MSFT SP500
2.082 1.068
Note: MSFT has a higher mean and higher SD than SP500
![Page 18: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/18.jpg)
© Eric Zivot 2006
-0.2 -0.1 0.0 0.1 0.2
0.0
0.2
0.4
0.6
0.8
1.0
Empirical CDF of Gaussian data
sort(gwn)
#x(i)
<= x
> n1 = length(gwn)> plot(sort(gwn),(1:n1)/n1,type="s",ylim=c(0,1), col=“slateblue1”+ main="Empirical CDF of Gaussian data", ylab="#x(i) <= x")
![Page 19: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/19.jpg)
© Eric Zivot 2006
-2 -1 0 1 2
0.0
0.2
0.4
0.6
0.8
1.0
Empirical CDF vs. Normal CDF for Gaussian data
standardized gwn
CD
F
Normal CDFEmpirical CDF
![Page 20: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/20.jpg)
QQ-plots against Normal Distribution
© Eric Zivot 2006
> par(mfrow=c(2,2))> qqnorm(gwn)> qqline(gwn)> qqnorm(MSFT.ret)> qqline(MSFT.ret)> qqnorm(SP500.ret)> qqline(SP500.ret)> par(mfrow=c(1,1))
Nice and straight!
Empirical quantilesdon’t match normal quantiles in the tails!
![Page 21: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/21.jpg)
QQ-Plots for Fat-tailed and Skewed Data
© Eric Zivot 2006
![Page 22: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/22.jpg)
© Eric Zivot 2006
![Page 23: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/23.jpg)
Effect of Outliers on Descriptive Statistics
© Eric Zivot 2006
Outlier
![Page 24: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/24.jpg)
Summary statistics of polluted data
© Eric Zivot 2006
> tmp = cbind(gwn, gwn.new)> apply(tmp, 2, mean)
gwn gwn.new0.0043420 -0.0006391
> apply(tmp, 2, sd)gwn gwn.new
0.09515 0.11746
> apply(tmp, 2, skewness)gwn gwn.new
0.2842 -2.3751
> apply(tmp, 2, kurtosis)gwn gwn.new
0.1557 18.3707
Notice how sample statistics are influenced by the single outlier
# outlier robust measures> apply(tmp, 2, median)
gwn gwn.new-0.0009163 -0.0009163
> apply(tmp, 2, IQR)gwn gwn.new
0.1200 0.1219
![Page 25: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/25.jpg)
© Eric Zivot 2006
-0.4
-0.2
0.0
0.2
Boxplot of monthly cc returns on Microsoftm
onth
ly c
c re
turn
> boxplot(MSFT,outchar=T,main="Boxplot of monthly cc+ returns on Microsoft",ylab="monthly cc return")
medianIQR
Smallest data point at most1.5*IQR from bottom of box
Largest data point at most1.5*IQR from top of box
outliers
![Page 26: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/26.jpg)
© Eric Zivot 2006
> boxplot(gwn,MSFT,SP500,names=c("gwn","MSFT","SP500"),outchar=T,+ main="Comparison of return distributions", ylab="monthly cc+ return")
![Page 27: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/27.jpg)
Four Graph Summary
© Eric Zivot 2006
fourPanelPlot = function(ret) {retName = colnames(ret)ret.den = density(ret)par(mfrow=c(2,2))hist(ret, main=paste(retName, " monthly returns", sep=""),
xlab=retName, probability=T, col="cornflowerblue")boxplot(ret, outchar=T,col="cornflowerblue")plot(ret.den, main="smoothed density",
type="l", lwd=2, xlab="monthly return",ylab="density estimate")
lines(ret.den$x, dnorm(ret.den$x, mean=mean(ret), sd=sd(ret)),col="cornflowerblue", lwd=2)
legend(x="topleft", legend=c("smoothed", "normal"),lty=c(1,1), col=c("black", "blue"), lwd=2)
qqnorm(ret, col="cornflowerblue", pch=16)qqline(ret)par(mfrow=c(1,1))
}
![Page 28: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/28.jpg)
© Eric Zivot 2006
> fourPanelPlot(MSFT)
![Page 29: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/29.jpg)
Scatterplot
© Eric Zivot 2006
> plot(MSFT.mat,SP500.mat,main="Monthly cc returns on MSFT+ and SP500“, pch=16, cex=1.5, col=“blue”)> abline(h=mean(SP500)) # horizontal line at SP500 mean> abline(v=mean(MSFT)) # vertical line at MSFT mean
Cov(MSFT,SP500)=0.0028
Corr(MSFT,SP500)=0.60
![Page 30: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/30.jpg)
Pairwise Scatterplots
© Eric Zivot 2006
> pairs(cbind(gwn,MSFT,SP500), col=“blue”, pch=16,
+ cex=1.5, cex.axis=1.5)
![Page 31: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/31.jpg)
Sample Covariances and Correlations
© Eric Zivot 2006
> var(cbind(gwn,MSFT.mat,SP500.mat))gwn MSFT SP500
gwn 0.0090534 -0.001856 -0.0009685MSFT -0.0018563 0.010044 0.0029993SP500 -0.0009685 0.002999 0.0023494
> cor(cbind(gwn,MSFT.mat,SP500.mat))gwn MSFT SP500
gwn 1.0000 -0.1947 -0.2100MSFT -0.1947 1.0000 0.6174SP500 -0.2100 0.6174 1.0000
![Page 32: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/32.jpg)
Sample ACF for S&P 500
© Eric Zivot 2006
Monthly returns are essentially uncorrelated; i.e., unpredictable!
![Page 33: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/33.jpg)
Sample ACF for MSFT
© Eric Zivot 2006
Small negative 1st order autocorrelation – possibly MA(1) model
![Page 34: Descriptive Statistics for Financial Time Series© Eric Zivot 2006 Descriptive Statistics for Financial Time Series Econ 424 Winter 201 Eric Zivot Updated: January 26, 2015](https://reader034.vdocuments.mx/reader034/viewer/2022042416/5f3201a1a5b6671cdc6b78f0/html5/thumbnails/34.jpg)
© Eric Zivot 2006
Stylized Facts for Monthly CC Returns
• Returns appear to be approximately normally distributed– Some noticeable negative skewness and
excess kurtosis• Individual asset returns have higher SD
than diversified portfolios• Many assets are contemporaneously
correlated• Assets are approximately uncorrelated
over time (no serial correlation)