QMLE for Periodic Time-Varying Asymmetric $\log $ GARCH Models

Abstract

This paper establishes probabilistic and statistical properties of the extension of time-invariant coefficients asymmetric $\log $ GARCH processes to periodically time-varying coefficients ($P\log $ GARCH) one. In these models, the parameters of $\log -$volatility are allowed to switch periodically between different seasons. The main motivations of this new model are able to capture the asymmetry and hence leverage effect, in addition, the volatility coefficients are not a subject to positivity constraints. So, some probabilistic properties of asymmetric $P\log $ GARCH models have been obtained, especially, sufficient conditions ensuring the existence of stationary, causal, ergodic (in periodic sense) solution and moments properties are given. Furthermore, we establish the strong consistency and the asymptotic normality of the quasi-maximum likelihood estimator (QMLE) under extremely strong assumptions. Finally, we carry out a simulation study of the performance of the QML and the $P\log $ GARCH is applied to model the crude oil prices of Algerian Saharan Blend.

Keywords

QML / Periodicity / $\log $ GARCH')">Asymmetric $\log $ GARCH / EGARCH / Stationarity / Asymptotic properties

References

[1] Aknouche A, Bibi A. Quasi-maximum likelihood estimation of periodic GARCH and periodic ARMA-GARCH processes. J. Time Ser. Anal.. 2008, 30 1 19-46

[2] Aknouche A, Guerbyenne H. Periodic stationarity of random coefficient periodic autoregressions. Stat. Probab. Lett.. 2009, 79 7 990-996

[3] Berkes I, Horváth L, Kokoszka P. GARCH processes: structure and estimation. Bernoulli. 2003, 9 2 201-227

[4] Bibi A, Ghezal A. On periodic time-varying bilinear processes: structure and asymptotic inference. Stat. Methods Appl.. 2016, 25 3 395-420

[5] Bibi A, Ghezal A. Markov-switching BILINEAR-GARCH processes: structure and estimation. Commun. Stat. Theory Methods. 2018, 47 2 307-323

[6] Bibi A, Lescheb I. Strong consistency and asymptotic normality of least squares estimators for PGARCH and PARMA-PGARCH. Stat. Probab. Lett.. 2010, 80 19–20 1532-1542

[7] Bollerslev T, Ghysels E. Periodic autoregressive conditional heteroskedasticity. J. Bus. Econ. Stat.. 1996, 14 2 139-151

[8] Bougerol P, Picard N. Strict stationarity of generalized autoregressive processes. Ann. Probab.. 1992, 20 4 1714-1730

[9] Chan NH, Ng CT. Statistical inference for non-stationary GARCH (p, q) models. Electron. J. Stat.. 2009, 3 956-992

[10] Francq C, Zakoïan JM. Quasi-maximum likelihood estimation in GARCH processes when some coefficients are equal to zero. Stoch. Process. Their Appl.. 2004, 117 9 1265-1284

[11] Francq C, Zakoïan JM. GARCH Models: Structure, Statistical Inference and Financial Applications. 2010 Hoboken: Wiley

[12] Francq C, Wintenberger O, Zakoïan JM. GARCH models without positivity constraints: exponential or $\log $ GARCH ?. J. Econom.. 2013, 177 34-46

[13] Francq C, Sucarrat G. An exponential chi-squared QMLE for $\log $-GARCH models via the ARMA representation. J. Financ. Econom.. 2018, 16 1 129-154

[14] Horst U. The stochastic equation $Y_{t+1}=A_{t}Y_{t}+B_{t}$ with non-stationary coefficients. J. Appl. Probab.. 2001, 38 1 80-94

[15] Jensen ST, Rahbek A. Asymptotic normality of the QML estimator of ARCH in the non-stationary case. Econometrica. 2004, 72 2 641-646

[16] Kesten H, Spitzer F. Convergence in distribution for products of random matrices. Zeitschrift fur Wahrscheinlichkeitstheorie und Verwandte Gebiete. 1984, 67 4 363-386

[17] Martin DEK, Kedem B. Estimation of the period of periodically correlated sequences. J. Time Ser. Anal.. 1993, 14 2 193-205