Skip to main content
Log in

Bayesian inference of multiple structural change models with asymmetric GARCH errors

  • Original Paper
  • Published:
Statistical Methods & Applications Aims and scope Submit manuscript

This article has been updated

Abstract

Structural change in any time series is practically unavoidable, and thus correctly detecting breakpoints plays a pivotal role in statistical modelling. This research considers segmented autoregressive models with exogenous variables and asymmetric GARCH errors, GJR-GARCH and exponential-GARCH specifications, which utilize the leverage phenomenon to demonstrate asymmetry in response to positive and negative shocks. The proposed models incorporate skew Student-t distribution and prove the advantages of the fat-tailed skew Student-t distribution versus other distributions when structural changes appear in financial time series. We employ Bayesian Markov Chain Monte Carlo methods in order to make inferences about the locations of structural change points and model parameters and utilize deviance information criterion to determine the optimal number of breakpoints via a sequential approach. Our models can accurately detect the number and locations of structural change points in simulation studies. For real data analysis, we examine the impacts of daily gold returns and VIX on S&P 500 returns during 2007–2019. The proposed methods are able to integrate structural changes through the model parameters and to capture the variability of a financial market more efficiently.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5

Similar content being viewed by others

Change history

  • 05 December 2020

    The layout of several tables was corrupted in the original publication. The table layout has been corrected.

References

  • Alberg D, Shalit H, Yosef R (2008) Estimating stock market volatility using asymmetric GARCH models. Appl Financ Econ 18:1201–1208

    Article  Google Scholar 

  • Bai J, Perron P (2003) Computation and analysis of multiple structural change models. J Appl Econ 18:1–22

    Article  Google Scholar 

  • Bollerslev T (1986) Generalized autoregressive conditional heteroskedasticity. J Econom 31:307–327

    Article  MathSciNet  Google Scholar 

  • Bradley BO, Taqqu MS (2003) Financial risk and heavy tails. In: Handbook of heavy tailed distributions in finance, North-Holland, pp 35–103

  • Chen CWS, Chiang TC, So MKP (2003) Asymmetrical reaction to US stock-return news: evidence from major stock markets based on a double-threshold model. J Econ Bus 55:487–502

    Article  Google Scholar 

  • Chen CWS, Gerlach R, Liu FC (2011) Detection of structural breaks in a time-varying heteroscedastic regression model. J Stat Plan Inference 141:3367–3381

    Article  Google Scholar 

  • Chen CWS, So MKP (2006) On a threshold heteroscedastic model. Int J Forecast 22:73–89

    Article  Google Scholar 

  • Chiang TC, Chen CWS, So MKP (2007) Asymmetric return and volatility responses to composite news from stock markets. Multinatl Financ J 11:179–210

    Article  Google Scholar 

  • Chib S (1998) Estimation and comparison of multiple changepoint models. J Econom 86:221–242

    Article  Google Scholar 

  • Davis RA, Lee TC, Rodriguez-Yam GA (2006) Structural break estimation for nonstationary time series models. J Am Stat Assoc 101:223–239

    Article  MathSciNet  Google Scholar 

  • Dong MC, Chen CWS, Lee S, Sriboonchitta S (2019) How strong is the relationship among gold and USD exchange rates? Analytics based on structural change models. Comput Econ 53:343–366

    Article  Google Scholar 

  • Engle RF (1982) Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica 50:987–1008

    Article  MathSciNet  Google Scholar 

  • Fearnhead P, Liu Z (2005) On-line inference for multiple changepoints problems. J R Stat Soc Ser B Stat Methodol 69:589–605

    Article  MathSciNet  Google Scholar 

  • Gelman A, Roberts GO, Gilks WR (1996) Efficient metropolis jumping rules. In: Bernardo JM, Berger JO, Dawid AP, Smith AFM (eds) Bayesian statistics, vol 5. Oxford University Press, Oxford, pp 599–608

    Google Scholar 

  • Giordani P, Kohn R (2012) Efficient Bayesian inference for multiple change-point and mixture innovation models. J Bus Econ Stat 196:66–77

    MathSciNet  Google Scholar 

  • Giot P (2005) Relationships between implied volatility indices and stock index returns. J Portf Manag 31:92–100

    Article  Google Scholar 

  • Glosten LR, Jagannathan R, Runkle DE (1993) On the relation between the expected value and the volatility of the nominal excess return on stocks. J Financ 487:1779–1801

    Article  Google Scholar 

  • Hansen BE (1994) Autoregressive conditional density estimation. Int Econ Rev 35:705–730

    Article  Google Scholar 

  • Hastings WK (1970) Monte Carlo sampling methods using Markov chains and their applications. Biometrika 57:97–109

    Article  MathSciNet  Google Scholar 

  • Inclán C (1993) Detection of multiple changes of variance using posterior odds. J Bus Econ Stat 11:289–300

    Google Scholar 

  • Inclán C, Tiao GC (1994) Use of cumulative sums of squares for retrospective detection of changes of variance. J Am Stat Assoc 89:913–923

    MathSciNet  MATH  Google Scholar 

  • Kim CJ, Nelson CR, Piger J (2004) The less-volatile US economy: a Bayesian investigation of timing, breadth, and potential explanations. J Bus Econ Stat 22:80–93

    Article  MathSciNet  Google Scholar 

  • Lai TL, Xing H (2013) Stochastic change-point ARX-GARCH models and their applications to econometric time series. Stat Sin 23:1573–1594

    MathSciNet  MATH  Google Scholar 

  • Mandelbrot BB (1963) The variation of certain speculative prices. J Bus 36:392–417

    Article  Google Scholar 

  • Metropolis N, Rosenbluth AW, Rosenbluth MN, Teller AH, Teller E (1953) Equation of state calculations by fast computing machines. J Chem Phys 21:1087–1092

    Article  Google Scholar 

  • Nelson DB (1991) Conditional heteroscedasticity in asset pricing: a new approach. Econometrica 59:347–370

    Article  MathSciNet  Google Scholar 

  • Pesaran MH, Pettenuzzo D, Timmermann A (2006) Forecasting time series subject to multiple structural breaks. Rev Econ Stud 73:1057–1084

    Article  MathSciNet  Google Scholar 

  • Pesaran MH, Timmermann A (2004) How costly is it to ignore breaks when forecasting the direction of a time series? Int J Forecast 20:411–425

    Article  Google Scholar 

  • Poon SH, Granger CWJ (2003) Forecasting volatility in financial markets: a review. J Econ Lit 41:478–539

    Article  Google Scholar 

  • Ray BK, Tsay RS (2002) Bayesian methods for change-point detection in long-range dependent processes. J Time Ser Anal 23:687–705

    Article  MathSciNet  Google Scholar 

  • Spiegelhalter DJ, Best NG, Carlin BP, Van der Linde A (2002) Bayesian measures of model complexity and fit (with discussion). J R Stat Soc Ser B Stat Methodol 64:83–616

    Article  Google Scholar 

  • Stock J, Watson M (1996) Evidence on structural instability in macroeconomic time series relations. J Bus Econ Stat 14:11–30

    Google Scholar 

  • Than-Thi H, Dong MC, Chen CWS (2019) Bayesian modelling structural changes on housing price dynamics. In: Kreinovich V, Sriboonchitta S (eds) Structural changes and their econometric modeling. Studies in computational intelligence, vol 808. Springer, Cham, pp 83–104

    Chapter  Google Scholar 

  • Tsay RS (2013) An introduction to analysis of financial data with R. Wiley, Hoboken

    MATH  Google Scholar 

  • Zeileis A, Kleiber C, Krämer W, Hornik K (2003) Testing and dating of structural changes in practice. Comput Stat Data Anal 44:109–123

    Article  MathSciNet  Google Scholar 

Download references

Acknowledgements

We thank the editor, the associate editor, and the anonymous referees for their valuable time and careful comments on our paper, which have led to an improved version of it. We acknowledge a valuable discussion with Professor S. Sriboonchitta. Cathy W.S. Chen’s research is funded by the Ministry of Science and Technology, Taiwan (MOST107-2118-M-035-005-MY2).

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Cathy W. S. Chen.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix

Appendix

We assume \({\varvec{\phi }}_{1}\) \(\sim\) N\(({\varvec{\phi }}_{10},{\varvec{\varSigma }}_1)\). Let \({\varvec{\theta }}_{-{\phi }}\) be the parameter vector excluding the element \({\varvec{\phi }}\). The conditional posterior distribution of \({\varvec{\phi }}_{1}\) is :

$$\begin{aligned} P({\varvec{\phi }}_{1} |\varvec{r},{\varvec{\theta }}_{-{\phi }_{1}}) \propto \mathcal{L}(\varvec{r}|{\varvec{\theta }})\times P({\varvec{\phi }}_{1}). \end{aligned}$$

In this case, when \(\varepsilon _t\) is a normal distribution, we can write the conditional posterior as:

$$\begin{aligned}&P({\varvec{\phi }}_{1} | \varvec{r},{\varvec{\theta }}_{-{\phi }_{1}}) \\&\quad \propto |{\varvec{H}}|^{-\frac{1}{2}} {\text{ exp }} \left\{ -\frac{1}{2}(\varvec{r}-{\varvec{Z}}{\varvec{\phi }}_{1})^T{\varvec{H}}^{-1} ({\varvec{r}}-{\varvec{Z}}{\varvec{\phi }}_{1}) -\frac{1}{2}({\varvec{\phi }}_{1}-{\varvec{\phi }}_{10})^T{{\varvec{\varSigma }}}_i^{-1}({\varvec{\phi }}_{1}-{\varvec{\phi }}_{10}) \right\} \\&\quad \propto |{\varvec{H}}|^{-\frac{1}{2}} {\text{ exp }} \left\{ -\frac{1}{2}({\varvec{\phi }}_{1}-{\varvec{\phi }}^*)^T {{\varvec{\varSigma }}}^{*-1}({\varvec{\phi }}_{1}-{\varvec{\phi }}^*) \right\} , \end{aligned}$$
(8)

and

$$\begin{aligned} {\varvec{\phi }}^*={\varvec{\varSigma }}^*({\varvec{Z}}^T{\varvec{H}}^{-1}\varvec{r}+{\varvec{\varSigma }}_1^{-1}{\varvec{\phi }}_{10}), \, \, {\varvec{\varSigma }}^*= ({\varvec{Z}}^T{\varvec{H}}^{-1}{\varvec{Z}}+{\varvec{\varSigma }}_1^{-1})^{-1}. \end{aligned}$$
(9)

The definitions of \({\varvec{Z}}\), \(\varvec{r}\), and \({\varvec{H}}\) for regime I are as follows.

$$\begin{aligned}&{\varvec{Z}}={ \left[ \begin{array}{cccc} 1 &{} r_{1} &{} x_{1,1} &{}x_{2,1} \\ \vdots &{} \vdots &{} \vdots &{}\vdots \\ \vdots &{} \vdots &{} \vdots &{}\vdots \\ 1 &{} r_{T_{1}-1} &{} x_{1,T_{1}-1} &{}x_{2,T_{1}-1} \end{array} \right] },\\&{\varvec{r}}= \left[ {\begin{array}{l} r_{2}\\ \vdots \\ \vdots \\ r_{T_{1}} \end{array}} \right] \quad {\text{and}} \quad {\varvec{H}} = \left[ {\begin{array}{llll} h_{2} &{}0 &{}\cdots &{}0\\ 0 &{}h_{3} &{}\cdots &{}0\\ \vdots &{}\vdots &{}\ddots &{}0\\ 0 &{}0 &{}\cdots &{}h_{T_{1}} \end{array}} \right] . \end{aligned}$$

The conditional posterior in Eq. (8) is not a standard form. We employ it to obtain new estimates. In other words, we draw estimates \({\varvec{\phi }}_{1}\) from a truncated Gaussian density \(N_{B_1}({\varvec{\phi }}^*,{\varvec{\varSigma }}^*)\), where \(B_1=\{ |\phi _1^{(1)}|<1\}\), to get new iterates and apply the random walk MH algorithm to decide whether or not to update the previous estimates. One can usually select a suitable value with good convergence properties by having an acceptance rate of 25–50%, as Gelman et al. (1996) indicate, which represents good convergence performance.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Chen, C.W.S., Lee, B. Bayesian inference of multiple structural change models with asymmetric GARCH errors. Stat Methods Appl 30, 1053–1078 (2021). https://doi.org/10.1007/s10260-020-00549-z

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10260-020-00549-z

Keywords

Navigation