The QL and AQL estimates and the RMSE of each estimate is stated below that estimate for ARCH model.
Abstract
This chapter considers estimation of autoregressive conditional heteroscedasticity (ARCH) and the generalized autoregressive conditional heteroscedasticity (GARCH) models using quasi-likelihood (QL) and asymptotic quasi-likelihood (AQL) approaches. The QL and AQL estimation methods for the estimation of unknown parameters in ARCH and GARCH models are developed. Distribution assumptions are not required of ARCH and GARCH processes by QL method. Nevertheless, the QL technique assumes knowing the first two moments of the process. However, the AQL estimation procedure is suggested when the conditional variance of process is unknown. The AQL estimation substitutes the variance and covariance by kernel estimation in QL. Reports of simulation outcomes, numerical cases, and applications of the methods to daily exchange rate series and weekly prices’ changes of crude oil are presented.
Keywords
- ARCH model
- GARCH model
- the quasi-likelihood
- asymptotic quasi-likelihood
- martingale difference
- daily exchange rate series
- prices changes of crude oil
1. Introduction
The autoregressive conditional heteroscedasticity (ARCH(q)) process is defined by
and
The generalized autoregressive conditional heteroscedasticity (GARCH(p,q)) process
and
The GARCH model was developed by Bollersev [24] to extend the earlier work on ARCH models by Engle [1]. For estimation and applications of GARCH models, (see, [2, 3, 6, 7, 8, 10, 11, 14]). Moreover, GARCH models have now become the standard textbook material in econometrics and finance as exemplified by, for example, [20, 21, 22, 23].
This chapter considers estimation of ARCH and GARCH models using quasi-likelihood (QL) and asymptotic quasi-likelihood (AQL) approaches. Distribution assumptions are not required of ARCH and GARCH processes by the QL method. But, the QL technique assumes knowing the first two moments of the process. However, The AQL estimation procedure is suggested when the conditional variance of process is unknown. The AQL estimation substitutes the variance and covariance by kernel estimation in QL.
This chapter is structured as follows. Section 2 introduces the QL and AQL approaches. The estimation of ARCH model using QL and AQL methods are developed in Section 3. The estimation of GARCH model using QL and AQL methods are developed in Section 4. Reports of simulation outcomes, numerical cases and applications of the methods to a daily exchange rate series, and weekly prices changes of crude oil are also presented. Summary and conclusion are given in Section 5.
2. The QLE and AQL methods
Let the observation equation be given by
where
2.1 The QL method
For the model given by Eq. (5), assume that
and the quasi-likelihood estimation function (QLEF) can be defined by
where
If the sub-estimating function spaces of
then the QLEF can be defined by
and the estimation of
A limitation of the QL method is that the nature of
2.2 The AQL method
The QLEF (see Eqs. (6) and (7)) relies on the information of
Definition 2.2.1:
is asymptotically nonnegative definite,
Suppose, in probability,
expresses an AQLEF sequence. The solution of
In this chapter, the kernel smoothing estimator of
The estimation of
For estimation of unknown parameters in fanatical models by QL and AQL approaches, see [21, 28, 29, 30, 31, 32, 33]. The next sections present the parameter estimation of ARCH model using the QL and AQL methods.
3. Parameter estimation of ARCH(q) model using the QL and AQL methods
In this section, we will develop the estimation of ARCH model using QL and AQL methods.
3.1 Parameter estimation of ARCH(q) model using the QL method
The ARCH(q) process is defined by
and
The QLEF to estimate
Given
The QLEF, using
The QL estimate of
3.2 Parameter estimation of ARCH(q) model using the AQL method
For ARCH(q) model given by Eqs. (10) and (11) and using the same argument listed under Eq. (11). First, to estimate
Given
Second, by kernel estimation method, we find
Third, to estimate the parameters
The AQL estimate of
3.3 Simulation studies for the ARCH(1) model
The estimation of ARCH(1) model using QL and AQL methods are considered in simulation studies. The ARCH(1) process is defined by
and
3.3.1 Parameter estimation of ARCH(1) model using the QL method
For ARCH(1) given by Eqs. (16) and (17), the martingale difference is
The QLEF to estimate
Given
To estimate the parameters
The solution of
and let
where
The initial values might be affected the estimation results. For extensive discussion on assigning initial values in the QL estimation procedures, see [21, 34].
3.3.2 Parameter estimation of ARCH(1) model using the AQL method
Considering the ARCH(1) model given by Eqs. (16) and (17) and using the same argument listed under Eq. (17). First, we need to estimate
Given
Second, by kernel estimation method, we find
Third, to estimate the parameters
The AQL estimate of
and let
The estimation procedure will be iteratively repeated until it converges.
For each parameter setting, T = 500 observations are simulated from the true model. We then replicate the experiment for 1000 times to obtain the mean and root mean squared errors (RMSE) for
|
|
|
|
|
|
|
|
|
|
---|---|---|---|---|---|---|---|---|---|
True | 0.010 | 0.980 | 1.30 | 0.010 | 0.980 | −1.30 | 0.010 | 0.980 | 0.030 |
QL | 0.009 | 0.989 | 1.299 | 0.009 | 0.989 | −1.30 | 0.009 | 0.989 | 0.029 |
0.001 | 0.010 | 0.006 | 0.001 | 0.010 | 0.006 | 0.001 | 0.010 | 0.006 | |
AQL | 0.009 | 0.989 | 1.30 | 0.009 | 0.989 | −1.29 | 0.009 | 0.989 | 0.030 |
0.001 | 0.010 | 0.0003 | 0.002 | 0.009 | 0.0003 | 0.001 | 0.009 | 0.0003 | |
True | 0.050 | 0.950 | 1.30 | 0.050 | 0.950 | −1.30 | 0.050 | .950 | 0.030 |
QL | 0.049 | 0.949 | 1.29 | 0.049 | 0.940 | −1.30 | 0.049 | 0.94 | 0.029 |
0.001 | 0.0001 | 0.014 | 0.001 | 0.010 | 0.014 | 0.001 | 0.010 | 0.014 | |
AQL | 0.049 | 0.940 | 1.32 | 0.049 | 0.940 | −1.30 | 0.049 | 0.940 | 0.032 |
0.001 | 0.010 | 0.018 | 0.001 | 0.010 | 0.018 | 0.001 | 0.01 | 0.001 | |
True | 0.10 | 0.90 | 1.30 | 0.10 | 0.90 | −1.30 | 0.10 | 0.90 | 0.030 |
QL | 0.098 | 0.910 | 1.29 | 0.098 | 0.910 | −1.30 | 0.098 | 0.910 | 0.023 |
0.002 | 0.010 | 0.019 | 0.002 | 0.010 | 0.020 | 0.002 | 0.010 | 0.029 | |
AQL | 0.098 | 0.910 | 1.31 | 0.098 | 0.910 | −1.32 | 0.098 | 0.910 | 0.031 |
0.002 | 0.010 | 0.012 | 0.002 | 0.010 | 0.021 | 0.001 | 0.010 | 0.001 | |
True | 0.1 | 0.90 | −0.03 | 0.05 | 0.95 | −0.03 | 0.01 | 0.98 | −0.03 |
QL | 0.098 | 0.910 | −0.031 | 0.051 | 0.949 | −0.030 | 0.009 | 0.990 | −0.030 |
0.002 | 0.010 | 0.019 | 0.001 | 0.001 | 0.014 | 0.001 | 0.016 | 0.006 | |
AQL | 0.098 | 0.910 | −0.031 | 0.051 | 0.949 | −0.031 | 0.009 | 0.990 | −0.031 |
0.002 | 0.010 | 0.001 | 0.001 | 0.001 | 0.002 | 0.001 | 0.010 | 0.001 |
We generated N = 1000 independent random samples of size T = 20, 40, 60, 80, and 100 from ARCH(1) model. In Table 2, the QL and AQL estimation methods show the property of consistency, the RMSE decreases as the sample size increases.
|
|
|
|
|
|
||
---|---|---|---|---|---|---|---|
T = 20 | True | 0.010 | 0.980 | −0.030 | 0.05 | 0.950 | 1.3 |
QL | 0.009 | 0.990 | −0.029 | 0.0495 | 0.9485 | 1.300 | |
0.0008 | 0.0100 | 0.0319 | 0.0005 | 0.0015 | 0.0703 | ||
AQL | 0.009 | 0.990 | −0.031 | 0.0495 | 0.9485 | 1.3107 | |
0.0009 | 0.010 | 0.0084 | 0.0005 | 0.0015 | 0.0213 | ||
T = 40 | QL | 0.009 | 0.990 | −0.031 | 0.0495 | 0.9485 | 1.3015 |
0.00089 | 0.010 | 0.0223 | 0.0005 | 0.0015 | 0.0492 | ||
AQL | 0.009 | 0.990 | −0.031 | 0.0495 | 0.9485 | 1.3113 | |
0.00089 | 0.010 | 0.0039 | 0.0005 | 0.0015 | 0.0143 | ||
T = 60 | QL | 0.009 | 0.990 | −0.029 | 0.0495 | 0.9485 | 1.300 |
0.0009 | 0.010 | 0.0180 | 0.0005 | 0.0015 | 0.0404 | ||
AQL | 0.009 | 0.990 | −0.031 | 0.0495 | 0.9485 | 1.311 | |
0.0009 | 0.010 | 0.0027 | 0.0005 | 0.0015 | 0.0128 | ||
T = 80 | QL | 0.009 | 0.990 | −0.029 | 0.0490 | 0.9485 | 1.300 |
0.0009 | 0.010 | 0.016 | 0.0005 | 0.0015 | 0.0353 | ||
AQL | 0.009 | 0.990 | −0.310 | 0.0495 | 0.9485 | 1.3112 | |
0.0009 | 0.010 | 0.0020 | 0.0005 | 0.0015 | 0.0119 | ||
T = 100 | QL | 0.009 | 0.990 | 0.0292 | 0.0495 | 0.9485 | 1.3017 |
0.0009 | 0.010 | 0.0142 | 0.0005 | 0.0015 | 0.0314 | ||
AQL | 0.009 | 0.990 | −0.031 | 0.0495 | 0.9485 | 1.3111 | |
0.0009 | 0.010 | 0.0018 | 0.0005 | 0.0015 | 0.0116 |
3.4 Empirical applications
The first data set we analyze are the daily exchange rate of
We used the S + FinMetrics function archTest to carry out Lagrange multiplier (ML) test for the presence of ARCH effects in the residuals (see [35]). For
and
The estimation of unknown parameters, (
|
|
|
|
|
---|---|---|---|---|
QL | 0.1300 | 0.8387 | −0.00012 | 0.00013 |
AQL | 0.0200 | 0.9599 | −0.00111 | 0.1350 |
4. Parameter estimation of GARCH(p,q) model using the QL and AQL methods
In this section, we developing the estimation of GARCH model using QL and AQL methods.
4.1 Parameter estimation of GARCH(p,q) model using the QL method
The GARCH(p,q) process is defined by
and
The QLEF to estimate
Given
The QLEF, using
The QL estimate of
4.2 Parameter estimation of GARCH(p,q) model using the AQL method
Considering the GARCH(p,q) model given by Eqs. (31) and (32) and using the same argument listed under Eq. (32). First, we need to estimate
Given
Second, by kernel estimation method, we find
Third, to estimate the parameters
The AQL estimate of
4.3 Simulation studies for the GARCH(1,1) model
The estimation of GARCH(1,1) model using QL and AQL methods are considered in simulation studies. The GARCH(1,1) process is defined by
and
4.3.1 Parameter estimation of GARCH(1,1) model using the QL method
For GARCH(1,1) given by Eqs. (37) and (38), the martingale difference is
The QLEF to estimate
Given
To estimate the parameters
The solation of
and let
where
The initial values might be affected the estimation results. For extensive discussion on assigning initial values in the QL estimation procedures, see [21, 34].
4.3.2 Parameter estimation of GARCH(1,1) model using the AQL method
Considering the GARCH(1,1) model given by Eqs. (37) and (38) and using the same argument listed under (Eq. (38)). First, we need to estimate
Given
Second, by kernel estimation method, we find
Third, to estimate the parameters
The AQL estimate of
and let
where
The estimation procedure will be iteratively repeated until it converges.
For each parameter setting, T = 500 observations are simulated from the true model. We then replicate the experiment for 1000 times to obtain the mean and root mean squared errors (RMSE) for
|
|
|
|
|
|
|
|
|
---|---|---|---|---|---|---|---|---|
True | 0.15 | 0.65 | 0.87 | 0.10 | 0.20 | 0.41 | 0.88 | 0.08 |
QL | 0.149 | 0.779 | 0.865 | 0.074 | 0.199 | 0.461 | 0.912 | 0.057 |
0.040 | 0.353 | 0.011 | 0.029 | 0.031 | 0.155 | 0.033 | 0.025 | |
AQL | 0.150 | 0.661 | 0.851 | 0.092 | 0.209 | 0.405 | 0.901 | 0.076 |
0.001 | 0.012 | 0.019 | 0.009 | 0.010 | 0.006 | 0.021 | 0.004 | |
True | −0.10 | 0.48 | 0.89 | 0.08 | 0.16 | 0.37 | 0.9 | 0.08 |
QL | −0.101 | 0.556 | 0.902 | 0.058 | 0.159 | 0.434 | 0.922 | 0.058 |
0.034 | 0.212 | 0.014 | 0.024 | 0.030 | 0.189 | 0.024 | 0.025 | |
AQL | −0.110 | 0.486 | 0.891 | 0.0752 | 0.161 | 0.374 | 0.911 | 0.076 |
0.010 | 0.006 | 0.001 | 0.005 | 0.001 | 0.004 | 0.011 | 0.004 | |
True | 0.18 | 0.39 | 0.88 | 0.08 | 0.09 | 0.50 | 0.89 | 0.05 |
QL | 0.179 | 0.447 | 0.892 | 0.058 | 0.089 | 0.538 | 0.898 | 0.036 |
0.031 | 0.146 | 0.015 | 0.024 | 0.033 | 0.090 | 0.009 | 0.015 | |
AQL | 0.180 | 0.395 | 0.882 | 0.076 | 0.091 | 0.504 | 0.892 | 0.046 |
0.001 | 0.005 | 0.002 | 0.005 | 0.002 | 0.004 | 0.002 | 0.004 |
We generated N = 1000 independent random samples of size T = 20, 40, 60, 80, and 100 from GARCH(1,1) model. In Table 5, The QL and AQL estimation methods show the property of consistency, and the RMSE decreases as the sample size increases.
|
|
|
|
|
|
|
|
||
---|---|---|---|---|---|---|---|---|---|
True | 0.16 | 0.37 | 0.90 | 0.08 | −0.10 | 0.48 | 0.89 | 0.08 | |
QL | 0.17 | 0.42 | 0.89 | 0.07 | −0.09 | 0.51 | 0.90 | 0.06 | |
T = 20 | 0.176 | 0.511 | 0.008 | 0.016 | 0.169 | 0.451 | 0.018 | 0.022 | |
AQL | 0.16 | 0.38 | 0.89 | 0.07 | −0.10 | 0.47 | 0.90 | 0.07 | |
0.037 | 0.012 | 0.007 | 0.014 | 0.066 | 0.014 | 0.013 | 0.018 | ||
QL | 0.16 | 0.42 | 0.89 | 0.07 | −0.09 | 0.51 | 0.91 | 0.06 | |
0.149 | 0.422 | 0.007 | 0.016 | 0.137 | 0.326 | 0.018 | 0.021 | ||
T = 40 | AQL | 0.16 | 0.38 | 0.89 | 0.07 | −0.10 | 0.47 | 0.90 | 0.07 |
0.027 | 0.012 | 0.007 | 0.013 | 0.022 | 0.014 | 0.012 | 0.016 | ||
QL | 0.16 | 0.42 | 0.89 | 0.07 | −0.09 | 0.52 | 0.91 | 0.06 | |
0.121 | 0.289 | 0.007 | 0.018 | 0.119 | 0.307 | 0.018 | 0.021 | ||
T = 60 | AQL | 0.16 | 0.38 | 0.89 | 0.07 | −0.10 | 0.47 | 0.90 | 0.07 |
0.019 | 0.012 | 0.007 | 0.011 | 0.014 | 0.013 | 0.012 | 0.015 | ||
QL | 0.16 | 0.42 | 0.89 | 0.07 | −0.10 | 0.51 | 0.90 | 0.06 | |
0.100 | 0.159 | 0.007 | 0.017 | 0.108 | 0.248 | 0.018 | 0.021 | ||
T = 80 | AQL | 0.16 | 0.38 | 0.89 | 0.07 | −0.10 | 0.47 | 0.90 | 0.07 |
0.012 | 0.012 | 0.007 | 0.011 | 0.012 | 0.013 | 0.012 | 0.015 | ||
QL | 0.16 | 0.42 | 0.89 | 0.07 | −0.10 | 0.51 | 0.90 | 0.06 | |
0.100 | 0.159 | 0.007 | 0.018 | 0.101 | 0.242 | 0.018 | 0.021 | ||
T = 100 | AQL | 0.16 | 0.38 | 0.89 | 0.07 | −0.10 | 0.47 | 0.90 | 0.07 |
0.012 | 0.011 | 0.007 | 0.011 | 0.011 | 0.013 | 0.012 | 0.015 |
4.4 Empirical applications
The second set of data is the weekly price changes of crude oil prices
and
The estimation of unknown parameters, (
|
|
|
|
|
|
---|---|---|---|---|---|
QL | 0.0008 | 0.566 | 0.912 | 0.0004 | 0.002 |
AQL | 0.0089 | 0.630 | 0.972 | 0.041 | 0.185 |
5. Conclusions
In this chapter, two alternative approaches, QL and AQL, have been developed to estimate the parameters in ARCH and GARCH models. Parameter estimation for ARCH and GARCH models, which include nonlinear and non-Gaussian models is given. The estimations of unknown parameters are considered without any distribution assumptions concerning the processes involved, and the estimation is based on different scenarios in which the conditional covariance of the error’s terms are assumed to be known or unknown. Simulation studies and empirical analysis show that our proposed estimation methods work reasonably quite well for parameter estimation of ARCH and GARCH models. It will provide a robust tool for obtaining an optimal point estimate of parameters in heteroscedastic models like ARCH and GARCH models.
This chapter focuses on models in univariate, while it is desirable to consider multivariate extensions of the proposed models.
Acknowledgments
The author would like to acknowledge the helpful comments and suggestion of the editor. This study is conducted in the King Faisal University, Saudi Arabia, during the sabbatical year of the author from the Al Balqa Applied University, Jordan.
References
- 1.
Engle RF. Autoregressive conditional heteroskedasticity with estimates of the variance of U.K. inflation. Econometrica. 1982; 50 :987-1008 - 2.
Engle RF. GARCH 101: The use of ARCH/GARCH models in applied econometrics. The Journal of Economic Perspectives. 2001; 15 :157-168 - 3.
Bollerslev T, Chou RY, Kroner KF. ARCH modeling in finance: A selective review of the theory and empirical evidence. Journal of Econometrics. 1992; 52 :5-59 - 4.
Bera A, Higgins M. ARCH models: Properties, estimation and testing. Journal of Economic Surveys. 1993; 7 :305-366 - 5.
Bollerslev T, Engle RF, Nelson DB, Models ARCH. In: Engle RF, McFadden D, editors. Handbook of Econometrics. Vol. 4. Amsterdam: North-Holland; 1994. pp. 2959-3038 - 6.
Diebold F, Lopez J. Modeling volatility dynamics. In: Hoover K, editor. Macroeconometrics: Developments, Tensions and Prospects. Boston: Kluwer Academic Press; 1995. pp. 427-472 - 7.
Pagan A. The econometrics of financial markets. Journal of Empirical Finance. 1996; 3 :15-102 - 8.
Palm F. GARCH models of volatility. In: Rao CR, Maddala GS, editors. Handbook of Statistics. Vol. 14. Amsterdam: North-Holland; 1996. pp. 209-240 - 9.
Shephard N. Statistical aspects of ARCH and stochastic volatility models. In: Cox DR, Hinkley DV, Barndorff-Nielsen OE, editors. Time Series Models in Econometrics, Finance and Other Fields. London: Chapman Hall; 1996. pp. 1-67 - 10.
Andersen T, Bollerslev T. ARCH and GARCH models. In: Kotz S, Read CB, Banks DL, editors. Encyclopedia of Statistical Sciences. Vol. 2. New York: John Wiley and Sons; 1998 - 11.
Engle R, Patton A. What good is a volatility model? Quantitative Finance. 2001; 1 :237-245 - 12.
Degiannakis S, Xekalaki E. Autoregressive conditional heteroscedasticity (arch) models: A review. Quality Technology and Quantitative Management. 2004; 1 :271-324 - 13.
Diebold F. The Nobel memorial prize for Robert F. Engle. The Scandinavian Journal of Economics. 2004; 106 :165-185 - 14.
Andersen T, Diebold F. Volatility and correlation forecasting. In: Granger CWJ, Elliott G, Timmermann A, editors. Handbook of Economic Forecasting. Amsterdam: North-Holland; 2006. pp. 777-878 - 15.
Engle RF, Gonzalez-Rivera G. Semiparametric ARCH models. Journal of Business & Economic Statistics. 1991; 9 (4):345-359 - 16.
Li DX, Turtle HJ. Semiparametric ARCH models: An estimating function approach. Journal of Business & Economic Statistics. 2000; 18 (2):174-186 - 17.
Linton O, Mammen E. Estimating semiparametric ARCH models by kernel smoothing methods. Econometrica. 2005; 73 (3):771-836 - 18.
Linton OB. Semiparametric and nonparametric ARCH modeling. In: Handbook of Financial Time Series. Berlin Heidelberg: Springer; 2009. pp. 157-167 - 19.
Su L, Ullah A, Mishra S. Nonparametric and semiparametric volatility models: Specification, estimation, and testing. In: Handbook of Volatility Models and Their Applications. Hoboken, New Jersey, USA: John Wiley Sons, Inc.; 2012. pp. 269-291 - 20.
Alexander C. Market Models: A Guide to Financial Data Analysis. Chichester, UK: John Wiley and Sons, Ltd.; 2001 - 21.
Alzghool R. Estimation for state space models: Quasi-likelihood and asymptotic quasi-likelihood approaches [Ph.D. thesis]. Australia: School of Mathematics and Applied Statistics, University of Wollongong; 2008 - 22.
Enders W. Applied Econometric Time Series. Hoboken, NJ: John Wiley and Sons, Inc.; 2004 - 23.
Taylor S. Asset Price Dynamics and Prediction. Princeton, NJ: Princeton University Press; 2004 - 24.
Bollerslev T. Generalized autoregressive conditional heteroscedasticity. Journal of Econometrics. 1986; 31 (3):307-327 - 25.
Hedye CC. Quasi-Likelihood and Its Application: A General Approach to Optimal Parameter Estimation. New York: Springer; 1997 - 26.
Lin Y-X. A new kind of asymptotic quasi-score estimating function. Scandinavian Journal of Statistics. 2000; 27 :97-109 - 27.
Hardle W. Applied Nonparametric Regression. Cambridge: Cambridge University Press; 1991 - 28.
Alzghool R, Lin Y-X. Asymptotic Quasi-Likelihood Based on Kernel Smoothing for Nonlinear and Non-Gaussian State-Space Models. Lecture Notes in Engineering and Computer Science. London, UK: ICCSDE; 2007. pp. 926-932 - 29.
Alzghool R, Lin Y-X. Parameters estimation for SSMs: QL and AQL approaches. IAENG International Journal of Applied Mathematics. 2008; 38 :34-43 - 30.
Alzghool R, Lin Y-X, Chen SX. Asymptotic quasi-likelihood based on kernel smoothing for multivariate heteroskedastic models with correlation. American Journal of Mathematical and Management Sciences. 2010; 30 (1&2):147-177 - 31.
Alzghool R. Estimation for stochastic volatility model: Quasi-likelihood and asymptotic quasi-likelihood approaches. Journal of King Saud University-Science. 2017; 29 :114-118 - 32.
Alzghool R. Parameters estimation for GARCH (p,q) model: QL and AQL approaches. Electronic Journal of Applied Statistical Analysis (EJASA). 2017; 10 (1):180-193 - 33.
Alzghool R, Al-Zubi LM. Semi-parametric estimation for ARCH models. Alexandria Engineering Journal. 2018; 57 :367-373 - 34.
Alzghool R, Lin Y-X. Initial values in estimation procedures for state space models (SSMs). In: Proceedings of World Congress on Engineering, WCE 2011; London, UK: Newswood Limited; 2011 - 35.
Zivot E, Wang J. Modeling Financial Time Series with S-PLUS. New York: Springer; 2006