Skip to main content
SpringerLink
Log in

The power of bootstrap tests of cointegration rank

  • Original Paper
  • Published:
Computational Statistics Aims and scope Submit manuscript

Abstract

Bootstrap likelihood ratio tests of cointegration rank are commonly used because they tend to have rejection probabilities that are closer to the nominal level than the rejection probabilities of asymptotic tests. The effect of bootstrapping the test on its power is largely unknown. We show that a new computationally inexpensive procedure can be applied to the estimation of the power function of the bootstrap test of cointegration rank. The bootstrap test is found to have a power function close to that of the level-adjusted asymptotic test. The bootstrap test therefore estimates the level-adjusted power of the asymptotic test highly accurately. The bootstrap test may have low power to reject the null hypothesis of cointegration rank zero, or underestimate the cointegration rank. An empirical application to Euribor interest rates is provided as an illustration of the findings.

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

Access this article

Log in via an institution

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

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8

We’re sorry, something doesn't seem to be working properly.

Please try refreshing the page. If that doesn't work, please contact support so we can address the problem.

Notes

  1. In DGP 1 with cointegration rank \(r_{0}=1\) the condition 2 (ii) for testing \( r=0\) is that \({\varvec{\rho }} (\varvec{\Phi }^{(0)})<1\), where \( {\varvec{\Phi }}^{(0)}= \varvec{\Gamma }_{1}\). Since \(\rho (\varvec{\Phi }^{(0)})=\rho \ (\varvec{\Gamma }_{1})\), the condition is satisfied for \(\gamma =0.8\) and \(\delta =0\), \(0.1\), but not for \(\gamma =0.8\) and \(\delta =0.2\). In DGP 2 with cointegration rank \(r_{0}=2\) the condition 2 (ii) for testing \( r=0\) is that \(\rho (\varvec{\Phi }^{(0)})<1\) and the condition for testing \( r=1\) is that \(\rho (\varvec{\Phi }^{(0)})<1\) and \(\rho (\varvec{\Phi } ^{(1)})<1\). The condition \(\rho (\varvec{\Phi }^{(1)})<1\) is satisfied for \( \gamma =0.8\) and \(\delta =0,\,0.1\), but not for \(\gamma =0.8\) and \(\delta =0.2\).

  2. The maximum number of replications is set to \(10\)m. Consequently, if the failure rate of the bootstrap algorithm exceeds \(99\,\%\) no results are obtained for the rejection probability of the test.

References

  • Ahlgren N, Antell J (2008) Bootstrap and fast double bootstrap tests of cointegration rank with financial time series. Comput Stat Data Anal 52:4754–4767

    Google Scholar 

  • Bollerslev T (1990) Modelling the coherence in short-run nominal exchange rates: a multivariate generalized ARCH model. Rev Econ Stat 72:498–505

    Article  Google Scholar 

  • Cavaliere G, Rahbek A, Taylor AMR (2012) Bootstrap determination of the co-integration rank in vector autoregressive models. Econometrica 80:1721–1740

    Article  MathSciNet  MATH  Google Scholar 

  • Davidson R, MacKinnon JG (2000) Bootstrap tests: how many bootstraps? Econom Rev 19:55–68

    Article  MathSciNet  MATH  Google Scholar 

  • Davidson R, MacKinnon JG (2006) The power of bootstrap and asymptotic tests. J Econom 133:421–441

    Article  MathSciNet  Google Scholar 

  • Doornik JA (1998) Approximations to the asymptotic distribution of cointegration tests. J Econ Surv 12:573–593

    Article  Google Scholar 

  • Hall AD, Anderson HM, Granger CWJ (1992) A cointegration analysis of treasury bill yields. Rev Econ Stat 74:116–126

    Article  Google Scholar 

  • Johansen S (1996) Likelihood-based inference in cointegrated vector autoregressive models. Oxford University Press, Oxford

    Google Scholar 

  • Johansen S (2002) A small sample correction for the test of cointegrating rank in the vector autoregressive model. Econometrica 70:1929–1961

    Article  MathSciNet  MATH  Google Scholar 

  • Juselius K (2006) The cointegrated VAR model. Methodology and applications. Oxford University Press, Oxford

    MATH  Google Scholar 

  • Saikkonen P, Lütkepohl H (1999) Local power of likelihood ratio tests for the cointegrating rank of a VAR process. Econom Theory 15:50–78

    Google Scholar 

  • Swensen AR (2006) Bootstrap algorithms for testing and determining the cointegration rank in VAR models. Econometrica 74:1699–1714

    Article  MathSciNet  MATH  Google Scholar 

  • Swensen AR (2009) Corrigendum to ’Bootstrap algorithms for testing and determining the cointegration rank in VAR models’. Econometrica 77:1703–1704

    Article  MathSciNet  MATH  Google Scholar 

  • van Giersbergen NPA (1996) Bootstrapping the trace statistic in VAR models: Monte Carlo results and applications. Oxf Bull Econ Stat 58:391–408

    Article  Google Scholar 

Download references

Acknowledgments

The authors want to thank two anonymous referees for their comments, and the Editor. Their comments have greatly improved the paper. The paper was presented in the 57th Session of the International Statistical Institute, Durban, 2009, and the 3rd International Conference on Computational and Financial Econometrics, 2009, Limassol. Niklas Ahlgren acknowledges financial support from Magnus Ehrnrooths Stiftelse.

Author information

Authors and Affiliations

  1. Department of Finance and Statistics, Hanken School of Economics, PO Box 479 (Arkadiagatan 22), 00101 , Helsinki, Finland

    Niklas Ahlgren & Jan Antell

Authors
  1. Niklas Ahlgren
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Jan Antell
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Jan Antell.

Appendices

Appendix 1: Computational details

To compute the rejection probability of a bootstrap likelihood ratio test of cointegration rank by (8) in model (10) with \(p=5,\,k=2\) and \(T=100\) when \(M=100000\) and \(B=1000\) requires about \(18\) h \(14\) min CPU time on the Core \(2\) Duo CPU, \(2.4\) GHz, \(2.00\) GB memory machine used to estimate the power functions in the paper. The total computational time for estimating a power function with \(100\) points would be about \(76\) days CPU time, making the computational time intractable. The computational time increases with the series length and the numbers of lags in the VAR model. Models with conditionally heteroskedastic errors require more computational time than models with normal errors.

The method in Davidson and MacKinnon (2006) reduces the computational burden by a factor of \(B/2\) LR statistics, which for \(B=1000\) is \(500\). The computational time to compute the rejection probability of a bootstrap likelihood ratio test of cointegration rank by (9) is about \(2\) min and \(21\) s CPU time. The estimation of a power function with \( 100\) points takes about \(3\) h \(55\) min CPU time.

Appendix 2: Checking the approximation

In order to check the accuracy of the approximation in (9), we estimate the rejection probabilities of the bootstrap tests by (8) with \(M=100000\) and \(B=1000\) for some points on the power functions in Fig. 1. Table 9 compares the estimated rejection probabilities of the bootstrap tests with \(B=1000\) and \(B=1\). The differences in the rejection probabilities are smaller than \(0.02\,\%\) for size and smaller than \(0.5\,\%\) for power.

Table 9 Estimated rejection probabilities in model (10) when \(k=2,\,\gamma = 0\) and \(T=100\)
Full size table

Rights and permissions

Reprints and permissions

About this article

Cite this article

Ahlgren, N., Antell, J. The power of bootstrap tests of cointegration rank. Comput Stat 28, 2719–2748 (2013). https://doi.org/10.1007/s00180-013-0425-6

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00180-013-0425-6

Keywords

Search

Navigation