Lecture Note of Bus 41202, Spring 2017:

Univariate Volatility Models. Mr. Ruey Tsay

Conditional Heteroscedastic Models

What is the volatility of an asset?

Answer: Conditional standard deviation of the asset return (price)

Why is volatility important?

Has many important applications:
• Option (derivative) pricing, e.g., Black-Scholes formula
• Risk management, e.g. value at risk (VaR)
• Asset allocation, e.g., minimum-variance portfolio; see pages 184-
185 of Campbell, Lo and MacKinlay (1997).
• Interval forecasts
A key characteristic: Not directly observable!!

How to calculate volatility?

There are several versions of sample volatility, but the conditional
standard deviation is commonly used.
1. Use high-frequency data: French, Schwert & Stambaugh (1987);
see Section 3.15.
• Realized volatility of daily log returns: use intraday high-
frequency log returns.
• Use daily high, low, and closing (log) prices, e.g. range =
daily high - daily low.
2. Implied volatility of options data, e.g, VIX of CBOE. Figure 1.
1
 3. Econometric modeling: use daily or monthly returns
We focus on the econometric modeling first. Use of high frequency
data to compute realized volatility will be discussed later.
Note: In most applications, volatility is annualized. This can easily
be done by taking care of the data frequency. For instance, if we use
daily returns in econometric modeling, then the annualized volatility
(in the U.S.) is √
σt∗ = 252σt,
where σt is the estimated volatility derived from an employed model.
If we use monthly returns, then the annualized volaitlity is
√
σt∗ = 12σt,
where σt is the estimated volatility derived from the employed model
for the monthly returns. Our discussion, however, continues to use
σt for simplicity.
Basic idea of econometric modeling:
Shocks of asset returns are NOT serially correlated, but dependent,
implying that the serial dependence in asset returns is nonlinear.
As shown by the ACF of returns and absolute returns of some assets
we discussed so far.
Basic structure
p q
rt = µt + at, µt = φ0 + φirt−i − θiat−i,
X X

i=1 i=1

Volatility models are concerned with time-evolution of

σt2 = Var(rt|Ft−1) = Var(at|Ft−1),
the conditional variance of the return rt.

2
 VIXCLS [1990−01−02/2017−04−06]
Last 12.39
80

70

60

50

40

30

20

10

Jan 02 1990 Jan 02 1996 Jan 02 2002 Jan 02 2008 Jan 02 2014

Figure 1: Time plot of the daily VIX index from January 2, 1990 to April 6,2017.

Consider the daily closing index of the S&P500 index from January
03, 2007 to April 06, 2017. The log returns follow approximately an
MA(2) model
rt = 0.0002 + at − 0.109at−1 − 0.053at−2, σ 2 = 0.00017.
The residuals show no strong serial correlations. [plot not shown.]
R Demonstration
> require(quantmod)
> getSymbols("^GSPC",from="2007-01-03",to="2017-04-06")
[1] "GSPC"
> dim(GSPC)
[1] 2584 6
> head(GSPC)
GSPC.Open GSPC.High GSPC.Low GSPC.Close GSPC.Volume GSPC.Adjusted
2007-01-03 1418.03 1429.42 1407.86 1416.60 3429160000 1416.60
..
2007-01-10 1408.70 1415.99 1405.32 1414.85 2764660000 1414.85
> spc <- log(as.numeric(GSPC[,6]))
> rtn <- diff(spc)
> acf(rtn)

3
 0.10
0.05
residuals−sprtn

0.00
−0.05

0 500 1000 1500 2000 2500

Time

Figure 2: Time plot of residuals of an MA(2) model fitted to daily log returns of the S&P
500 index from January 3, 2007 to April 06, 2017.

Series m1$residuals^2
1.0
0.8
0.6
ACF

0.4
0.2
0.0

0 5 10 15 20 25 30 35

Lag

Figure 3: Sample ACF of the squared residuals of an MA(2) model fitted to daily log returns
of the S&P 500 index from January 3, 2007 to April 06, 2017.

4
> m1 <- arima(rtn,order=c(0,0,2))
> m1
Call: arima(x = rtn, order = c(0, 0, 2))

Coefficients:
ma1 ma2 intercept
-0.1088 -0.0530 2e-04
s.e. 0.0198 0.0205 2e-04

sigma^2 estimated as 0.0001676: log likelihood = 7562.82, aic = -15117.65

> acf(m1$residuals)
> acf(m1$residuals^2)

The residuals are shown in Figure 2.

Is volatility constant over time?
NO! Figure 2 shows a special feature, which is referred to as the
volatility clustering.
How to model the evolving volatility? See the ACF of the squared
residuals in Figure 3.
Two general categories
• “Fixed function” and
• Stochastic function
of the available information.

Univariate volatility models discussed:

1. Autoregressive conditional heteroscedastic (ARCH) model of En-
gle (1982),
2. Generalized ARCH (GARCH) model of Bollerslev (1986),
3. GARCH-M models,
4. IGARCH models (used by RiskMetrics),
5. Exponential GARCH (EGARCH) model of Nelson (1991),
5
 6. Threshold GARCH model of Zakoian (1994) or GJR model of
Glosten, Jagannathan, and Runkle (1993),
7. Asymmetric power ARCH (APARCH) models of Ding, Granger
and Engle (1994), [TGARCH and GJR models are special cases
of APARCH models.]
8. Stochastic volatility (SV) models of Melino and Turnbull (1990),
Harvey, Ruiz and Shephard (1994), and Jacquier, Polson and
Rossi (1994).

ARCH model
at = σtt, σt2 = α0 + α1a2t−1 + · · · + αma2t−m,
where {t} is a sequence of iid r.v. with mean 0 and variance 1,
α0 > 0 and αi ≥ 0 for i > 0.
Distribution of t: Standard normal, standardized Student-t, gener-
alized error dist (ged), or their skewed counterparts.
Properties of ARCH models
Consider an ARCH(1) model
at = σtt, σt2 = α0 + α1a2t−1,
where α0 > 0 and α1 ≥ 0.
1. E(at) = 0
2. Var(at) = α0/(1 − α1) if 0 < α1 < 1
3. Under normality,
3α02(1 + α1)
m4 = ,
(1 − α1)(1 − 3α12)
provided 0 < α12 < 1/3.
6
The 3rd property implies heavy tails.

Advantages
• Simplicity
• Generates volatility clustering
• Heavy tails (high kurtosis)
Weaknesses
• Symmetric between positive & negative prior returns
• Restrictive
• Provides no explanation
• Not sufficiently adaptive in prediction
Building an ARCH Model
1. Modeling the mean effect and testing for ARCH effects
Ho: no ARCH effects versus Ha : ARCH effects
Use Q-statistics of squared residuals; McLeod and Li (1983) &
Engle (1982)
2. Order determination
Use PACF of the squared residuals. (In practice, simply try some
reasonable order).
3. Estimation: Conditional MLE
4. Model checking: Q-stat of standardized residuals and squared
standardized residuals. Skewness & Kurtosis of standardized
residuals.
7
 R provides many plots for model checking and for presenting the
results.
5. Software: We use R with the package fGarch. (Other software
available).
Estimation: Conditional MLE or Quasi MLE
Special Note: In this course, we estimate volatility models using
the R package fGarch with garchFit command. The program is
easy to use and allows for several types of innovational distributions:
The default is Gaussian (norm), standardized Student-t distribution
(std), generalized error distribution (ged), skew normal distribution
(snorm), skew Student-t (sstd), skew generalized error distribution
(sged), and standardized inverse normal distribution (snig). Except
for the inverse normal distribution, other distribution functions are
discussed in the textbook. Readers should check the book for details
about the density functions and their parameters.
Example: Monthly log returns of Intel stock
R demonstration: The fGarch package. Output edited.
> library(fGarch)
> da=read.table("m-intc7303.txt",header=T)
> head(da)
date rtn
1 19730131 0.01005
.....
6 19730629 0.13333
> intc=log(da$rtn+1) <== log returns
> acf(intc)
> acf(intc^2)
> pacf(intc^2)
> Box.test(intc^2,lag=10,type=’Ljung’)
Box-Ljung test

data: intc^2
X-squared = 59.7216, df = 10, p-value = 4.091e-09

8
> m1=garchFit(~garch(3,0),data=intc,trace=F) <== trace=F reduces the amount of output.
> summary(m1)
Title: GARCH Modelling
Call: garchFit(formula = ~garch(3, 0), data = intc, trace = F)

Mean and Variance Equation:

data ~ garch(3, 0)
[data = intc]

Conditional Distribution: norm

Coefficient(s):
mu omega alpha1 alpha2 alpha3
0.016572 0.012043 0.208649 0.071837 0.049045

Std. Errors:
based on Hessian

Error Analysis:
Estimate Std. Error t value Pr(>|t|)
mu 0.016572 0.006423 2.580 0.00988 **
omega 0.012043 0.001579 7.627 2.4e-14 ***
alpha1 0.208649 0.129177 1.615 0.10626
alpha2 0.071837 0.048551 1.480 0.13897
alpha3 0.049045 0.048847 1.004 0.31536
---
Standardised Residuals Tests:
Statistic p-Value
Jarque-Bera Test R Chi^2 169.7731 0
Shapiro-Wilk Test R W 0.9606957 1.970413e-08
Ljung-Box Test R Q(10) 10.97025 0.3598405
Ljung-Box Test R Q(15) 19.59024 0.1882211
Ljung-Box Test R Q(20) 20.82192 0.40768
Ljung-Box Test R^2 Q(10) 5.376602 0.864644
Ljung-Box Test R^2 Q(15) 22.73460 0.08993976
Ljung-Box Test R^2 Q(20) 23.70577 0.255481
LM Arch Test R TR^2 20.48506 0.05844884

Information Criterion Statistics:

AIC BIC SIC HQIC
-1.228111 -1.175437 -1.228466 -1.207193

> m1=garchFit(~garch(1,0),data=intc,trace=F)
> summary(m1)
Title: GARCH Modelling

9
Call: garchFit(formula = ~garch(1, 0), data = intc, trace = F)

Mean and Variance Equation:

data ~ garch(1, 0)
[data = intc]

Conditional Distribution: norm

Coefficient(s):
mu omega alpha1
0.016570 0.012490 0.363447

Std. Errors:
based on Hessian

Error Analysis:
Estimate Std. Error t value Pr(>|t|)
mu 0.016570 0.006161 2.689 0.00716 **
omega 0.012490 0.001549 8.061 6.66e-16 ***
alpha1 0.363447 0.131598 2.762 0.00575 **
---
Log Likelihood:
230.2423 normalized: 0.6189309

Standardised Residuals Tests:

Statistic p-Value
Jarque-Bera Test R Chi^2 122.4040 0
Shapiro-Wilk Test R W 0.9647629 8.274158e-08
Ljung-Box Test R Q(10) 13.72604 0.1858587 <=== Meaning?
Ljung-Box Test R Q(15) 22.31714 0.09975386 <==== implication?
Ljung-Box Test R Q(20) 23.88257 0.2475594
Ljung-Box Test R^2 Q(10) 12.50025 0.2529700
Ljung-Box Test R^2 Q(15) 30.11276 0.01152131
Ljung-Box Test R^2 Q(20) 31.46404 0.04935483
LM Arch Test R TR^2 22.036 0.0371183

Information Criterion Statistics:

AIC BIC SIC HQIC
-1.221733 -1.190129 -1.221861 -1.209182

> plot(m1)
Make a plot selection (or 0 to exit):

1: Time Series
2: Conditional SD
3: Series with 2 Conditional SD Superimposed

10
 4: ACF of Observations
5: ACF of Squared Observations
6: Cross Correlation
7: Residuals
8: Conditional SDs
9: Standardized Residuals
10: ACF of Standardized Residuals
11: ACF of Squared Standardized Residuals
12: Cross Correlation between r^2 and r
13: QQ-Plot of Standardized Residuals

Selection: 13

Make a plot selection (or 0 to exit):

1: Time Series
2: Conditional SD
3: Series with 2 Conditional SD Superimposed
4: ACF of Observations
5: ACF of Squared Observations
6: Cross Correlation
7: Residuals
8: Conditional SDs
9: Standardized Residuals
10: ACF of Standardized Residuals
11: ACF of Squared Standardized Residuals
12: Cross Correlation between r^2 and r
13: QQ-Plot of Standardized Residuals

Selection: 0

The fitted ARCH(1) model is

rt = 0.0176 + at, at = σtt, t ∼ N (0, 1)
σt2 = 0.0125 + 0.363σt−1
2
.
Model checking statistics indicate that there are some higher order
dependence in the volatility, e.g., see Q(15) for the squared standard-
ized residuals. It turns out that a GARCH(1,1) model fares better
for the data.

Next, consider Student-t innovations.

R demonstration
11
 qnorm − QQ Plot

2
Sample Quantiles

0
−2
−4

−3 −2 −1 0 1 2 3

Theoretical Quantiles

Figure 4: QQ-plot for standardized residuals of an ARCH(1) model with Gaussian innova-
tions for monthly log returns of INTC stock: 1973 to 2003.

> m2=garchFit(~garch(1,0),data=intc,cond.dist="std",trace=F)
> summary(m2)
Title: GARCH Modelling

Call: garchFit(formula = ~garch(1, 0), data = intc, cond.dist = "std",

trace = F)

Mean and Variance Equation:

data ~ garch(1, 0)
[data = intc]

Conditional Distribution: std <====== Standardized Student-t.

Coefficient(s):
mu omega alpha1 shape
0.021571 0.013424 0.259867 5.985979

Error Analysis:
Estimate Std. Error t value Pr(>|t|)
mu 0.021571 0.006054 3.563 0.000366 ***
omega 0.013424 0.001968 6.820 9.09e-12 ***
alpha1 0.259867 0.119901 2.167 0.030209 *

12
shape 5.985979 1.660030 3.606 0.000311 *** <== Estimate of degrees of freedom
---
Log Likelihood:
242.9678 normalized: 0.6531391

Standardised Residuals Tests:

Statistic p-Value
Jarque-Bera Test R Chi^2 130.8931 0
Shapiro-Wilk Test R W 0.9637529 5.744026e-08
Ljung-Box Test R Q(10) 14.31288 0.1591926
Ljung-Box Test R Q(15) 23.34043 0.07717449
Ljung-Box Test R Q(20) 24.87286 0.2063387
Ljung-Box Test R^2 Q(10) 15.35917 0.1195054
Ljung-Box Test R^2 Q(15) 33.96318 0.003446127
Ljung-Box Test R^2 Q(20) 35.46828 0.01774746
LM Arch Test R TR^2 24.11517 0.01961957

Information Criterion Statistics:

AIC BIC SIC HQIC
-1.284773 -1.242634 -1.285001 -1.268039

> plot(m2)
Make a plot selection (or 0 to exit):

1: Time Series
2: Conditional SD
3: Series with 2 Conditional SD Superimposed
4: ACF of Observations
5: ACF of Squared Observations
6: Cross Correlation
7: Residuals
8: Conditional SDs
9: Standardized Residuals
10: ACF of Standardized Residuals
11: ACF of Squared Standardized Residuals
12: Cross Correlation between r^2 and r
13: QQ-Plot of Standardized Residuals

Selection: 13 <== The plot shows that the model needs further improvements.

> predict(m2,5) <===== Prediction

meanForecast meanError standardDeviation
1 0.02157100 0.1207911 0.1207911
2 0.02157100 0.1312069 0.1312069
3 0.02157100 0.1337810 0.1337810
4 0.02157100 0.1344418 0.1344418

13
5 0.02157100 0.1346130 0.1346130

The fitted model with Student-t innovations is

rt = 0.0216 + at, at = σtt,  ∼ t5.99
σt2 = 0.0134 + 0.260a2t−1.
We use t5.99 to denote the standardized Student-t distribution with
5.99 d.f.
Comparison with normal innovations:
• Using a heavy-tailed dist for t reduces the ARCH effect.
• The difference between the models is small for this particular
instance.
You may try other distributions for t.

GARCH Model
at = σtt,
m s
σt2 = α0 + αia2t−i + 2
βj σt−j
X X

i=1 j=1
where {t} is defined as before, α0 > 0, αi ≥ 0, βj ≥ 0, and
Pmax(m,s)
i=1 (αi + βi) < 1.
Re-parameterization:
Let ηt = a2t − σt2. {ηt} un-correlated series.
The GARCH model becomes
max(m,s) s
a2t = α0 + (αi + βi)a2t−i + ηt − βj ηt−j .
X X

i=1 j=1

This is an ARMA form for the squared series a2t .

Use it to understand properties of GARCH models, e.g. moment
equations, forecasting, etc.
14
Focus on a GARCH(1,1) model
σt2 = α0 + α1a2t−1 + β1σt−1
2
,
• Weak stationarity: 0 ≤ α1, β1 ≤ 1, (α1 + β1) < 1.
• Volatility clusters
• Heavy tails: if 1 − 2α12 − (α1 + β1)2 > 0, then
E(a4t ) 3[1 − (α1 + β1)2]
= > 3.
[E(a2t )]2 1 − (α1 + β1)2 − 2α12
• For 1-step ahead forecast,
σh2 (1) = α0 + α1a2h + β1σh2 .
For multi-step ahead forecasts, use a2t = σt22t and rewrite the
model as
2
σt+1 = α0 + (α1 + β1)σt2 + α1σt2(2t − 1).
2-step ahead volatility forecast
σh2 (2) = α0 + (α1 + β1)σh2 (1).
In general, we have
σh2 (`) = α0 + (α1 + β1)σh2 (` − 1), ` > 1.
This result is exactly the same as that of an ARMA(1,1) model
with AR polynomial 1 − (α1 + β1)B.

Example: Monthly excess returns of S&P 500 index starting from

1926 for 792 observations.

15
The fitted of a Gaussian AR(3) model
r̃t = rt − 0.0062
r̃t = .089r̃t−1 − .024r̃t−2 − .123r̃t−3 + .007 + at,
σ̂a2 = 0.00333.
For the GARCH effects, use a GARCH(1,1) model, we have
A joint estimation:
rt = 0.032rt−1 − 0.030rt−2 − 0.011rt−3 + 0.0077 + at
σt2 = 7.98 × 10−5 + .853σt−1
2
+ 0.124a2t−1.
Implied unconditional variance of at is
0.0000798
= 0.00352
1 − 0.853 − 0.1243
close to the expected value. All AR coefficients are statistically in-
significant.
A simplified model:
rt = 0.00745 + at, σt2 = 8.06 × 10−5 + .854σt−1
2
+ .122a2t−1.
Model checking:
For ãt: Q(10) = 11.22(0.34) and Q(20) = 24.30(0.23).
For ã2t : Q(10) = 9.92(0.45) and Q(20) = 16.75(0.67).
Forecast: 1-step ahead forecast:
σh2 (1) = 0.00008 + 0.854σh2 + 0.122a2h
Horizon 1 2 3 4 5 ∞
Return .0074 .0074 .0074 .0074 .0074 .0074
Volatility .054 .054 .054 .054 .054 .059
R demonstration:
16
> sp5=scan("sp500.txt")
Read 792 items
> pacf(sp5)
> m1=arima(sp5,order=c(3,0,0))
> m1
Call: arima(x = sp5, order = c(3, 0, 0))

Coefficients:
ar1 ar2 ar3 intercept
0.0890 -0.0238 -0.1229 0.0062
s.e. 0.0353 0.0355 0.0353 0.0019

sigma^2 estimated as 0.00333: log likelihood = 1135.25, aic=-2260.5

> m2=garchFit(~arma(3,0)+garch(1,1),data=sp5,trace=F)
> summary(m2)
Title: GARCH Modelling
Call:
garchFit(formula = ~arma(3,0)+garch(1, 1), data = sp5, trace = F)

Mean and Variance Equation:

data ~ arma(3, 0) + garch(1, 1)
[data = sp5]

Conditional Distribution: norm

Error Analysis:
Estimate Std. Error t value Pr(>|t|)
mu 7.708e-03 1.607e-03 4.798 1.61e-06 ***
ar1 3.197e-02 3.837e-02 0.833 0.40473
ar2 -3.026e-02 3.841e-02 -0.788 0.43076
ar3 -1.065e-02 3.756e-02 -0.284 0.77677
omega 7.975e-05 2.810e-05 2.838 0.00454 **
alpha1 1.242e-01 2.247e-02 5.529 3.22e-08 ***
beta1 8.530e-01 2.183e-02 39.075 < 2e-16 ***
---
Log Likelihood:
1272.179 normalized: 1.606287

Standardised Residuals Tests:

Statistic p-Value
Jarque-Bera Test R Chi^2 73.04842 1.110223e-16
Shapiro-Wilk Test R W 0.985797 5.961994e-07
Ljung-Box Test R Q(10) 11.56744 0.315048
Ljung-Box Test R Q(15) 17.78747 0.2740039
Ljung-Box Test R Q(20) 24.11916 0.2372256
Ljung-Box Test R^2 Q(10) 10.31614 0.4132089

17
Ljung-Box Test R^2 Q(15) 14.22819 0.5082978
Ljung-Box Test R^2 Q(20) 16.79404 0.6663038
LM Arch Test R TR^2 13.34305 0.3446075

Information Criterion Statistics:

AIC BIC SIC HQIC
-3.194897 -3.153581 -3.195051 -3.179018

> m2=garchFit(~garch(1,1),data=sp5,trace=F)
> summary(m2)
Title: GARCH Modelling
Call: garchFit(formula = ~garch(1, 1), data = sp5, trace = F)

Mean and Variance Equation:

data ~ garch(1, 1)
[data = sp5]

Conditional Distribution: norm

Error Analysis:
Estimate Std. Error t value Pr(>|t|)
mu 7.450e-03 1.538e-03 4.845 1.27e-06 ***
omega 8.061e-05 2.833e-05 2.845 0.00444 **
alpha1 1.220e-01 2.202e-02 5.540 3.02e-08 ***
beta1 8.544e-01 2.175e-02 39.276 < 2e-16 ***
---
Log Likelihood:
1269.455 normalized: 1.602848

Standardised Residuals Tests:

Statistic p-Value
Jarque-Bera Test R Chi^2 80.32111 0
Shapiro-Wilk Test R W 0.9850517 3.141228e-07
Ljung-Box Test R Q(10) 11.22050 0.340599
Ljung-Box Test R Q(15) 17.99703 0.262822
Ljung-Box Test R Q(20) 24.29896 0.2295768
Ljung-Box Test R^2 Q(10) 9.920157 0.4475259
Ljung-Box Test R^2 Q(15) 14.21124 0.509572
Ljung-Box Test R^2 Q(20) 16.75081 0.6690903
LM Arch Test R TR^2 13.04872 0.3655092

Information Criterion Statistics:

AIC BIC SIC HQIC
-3.195594 -3.171985 -3.195645 -3.186520

> plot(m2)

18
Make a plot selection (or 0 to exit):

1: Time Series
2: Conditional SD
3: Series with 2 Conditional SD Superimposed
4: ACF of Observations
5: ACF of Squared Observations
6: Cross Correlation
7: Residuals
8: Conditional SDs
9: Standardized Residuals
10: ACF of Standardized Residuals
11: ACF of Squared Standardized Residuals
12: Cross Correlation between r^2 and r
13: QQ-Plot of Standardized Residuals

Selection: 3

> predict(m2,6)
meanForecast meanError standardDeviation
1 0.007449721 0.05377242 0.05377242
2 0.007449721 0.05388567 0.05388567
3 0.007449721 0.05399601 0.05399601
4 0.007449721 0.05410353 0.05410353
5 0.007449721 0.05420829 0.05420829
6 0.007449721 0.05431038 0.05431038

Turn to Student-t innovation. (R output omitted.)

Estimation of degrees of freedom:
rt = 0.0085 + at, at = σtt, t ∼ t7
σt2 = .000125 + .113a2t−1 + .842σt−1
2
,
where the estimated degrees of freedom is 7.00.
Forecasting evaluation
Not easy to do; see Andersen and Bollerslev (1998).

IGARCH model
An IGARCH(1,1) model:
at = σtt, σt2 = α0 + β1σt−1
2
+ (1 − β1)a2t−1.
19
 Series with 2 Conditional SD Superimposed

0.4
0.2
x

0.0
−0.2

0 200 400 600 800

Index

Figure 5: Monthly S&P 500 excess returns and fitted volatility

For the monthly excess returns of the S&P 500 index, we have
rt = .007 + at, σt2 = .0001 + .806σt−1
2
+ .194a2t−1
For an IGARCH(1,1) model,
σh2 (`) = σh2 (1) + (` − 1)α0, ` ≥ 1,
where h is the forecast origin.
Effect of σh2 (1) on future volatilities is persistent, and the volatility
forecasts form a straight line with slope α0. See Nelson (1990) for
more info.
Special case: α0 = 0. Volatility forecasts become a constant.
This property is used in RiskMetrics to VaR calculation.
Example: An IGARCH(1,1) model for the monthly excess returns
of S&P500 index from 1926 to 1991 is given below via R.
rt = 0.0069 + at, at = σtt
20
 σt2 = 0.099a2t−1 + .901σt−1
2
.
R demonstration: Using R script Igarch.R.
> source("Igarch.R")
> sp5=scan(file="sp500.txt")
> Igarch(sp5,include.mean=T)
Estimates: 0.006874402 0.9007153
Maximized log-likehood: -1258.219

Coefficient(s):
Estimate Std. Error t value Pr(>|t|)
mu 0.0068744 0.0015402 4.46332 8.07e-06 ***
beta 0.9007153 0.0158018 57.00082 < 2e-16 ***
Another R package: rugarch can be used to fit volatility models
too.
> sp5=scan("sp500.txt")
> require(rugarch)
>spec1=ugarchspec(variance.model=list(model="iGARCH",garchOrder=c(1,1)),
mean.model=list(armaOrder=c(0,0)))
> mm=ugarchfit(data=sp5,spec=spec1)
> mm
*---------------------------------*
* GARCH Model Fit *
*---------------------------------*
Conditional Variance Dynamics
-----------------------------------
GARCH Model : iGARCH(1,1)
Mean Model : ARFIMA(0,0,0)
Distribution : norm

Optimal Parameters
------------------------------------
Estimate Std. Error t value Pr(>|t|)
mu 0.007417 0.001525 4.8621 0.000001
omega 0.000051 0.000018 2.9238 0.003458
alpha1 0.142951 0.021443 6.6667 0.000000
beta1 0.857049 NA NA NA

Robust Standard Errors:

Estimate Std. Error t value Pr(>|t|)
mu 0.007417 0.001587 4.6726 0.000003
omega 0.000051 0.000019 2.6913 0.007118
alpha1 0.142951 0.024978 5.7230 0.000000
beta1 0.857049 NA NA NA

21
LogLikelihood : 1268.238

Information Criteria
------------------------------------
Akaike -3.1950
Bayes -3.1773
Shibata -3.1951
Hannan-Quinn -3.1882

Weighted Ljung-Box Test on Standardized Residuals

------------------------------------
statistic p-value
Lag[1] 0.5265 0.4681
Lag[2*(p+q)+(p+q)-1][2] 0.5304 0.6795
Lag[4*(p+q)+(p+q)-1][5] 2.5233 0.5009
d.o.f=0
H0 : No serial correlation

Weighted Ljung-Box Test on Standardized Squared Residuals

------------------------------------
statistic p-value
Lag[1] 1.166 0.2803
Lag[2*(p+q)+(p+q)-1][5] 2.672 0.4702
Lag[4*(p+q)+(p+q)-1][9] 4.506 0.5054
d.o.f=2
.....

22