Estimating high-frequency foreign exchange rate volatility with nonparametric ARCH models

https://doi.org/10.1016/S0378-3758(97)00144-4Get rights and content

Abstract

High-frequency foreign exchange rate (HFFX) series are analyzed on an operational time scale using models of the ARCH class. Comparison of the estimated conditional variances focuses on the asymmetry and persistence issue. Estimation results for parametric models confirm standard results for HFFX series, namely high persistence and no significance of the asymmetry coefficient in an EGARCH model. To find out whether these results are robust against alternative specifications, nonparametric models are estimated. Local linear estimation techniques are applied to a nonparametric ARCH model of order one (CHARN). The results show significant asymmetry of the volatility function. To allow for both flexibility and persistence, a higher-order multiplicative model is fitted. The results show important asymmetries in volatility. In contrast to the EGARCH specification, the news impact curves have different shapes for different lags and tend to increase slower at the boundaries.

References (32)

  • E.K. Berndt et al.

    Estimation inference in nonlinear structural models

    Ann. Econom. Soc. Measurement

    (1974)
  • P. Bossaerts et al.

    A new method for volatility estimation with applications to foreign exchange rate series

  • P. Bossaerts et al.

    Foreign exchange rates have surprising volatility

  • R. Chen et al.

    Nonlinear additive ARX models

    J. Amer. Statist. Assoc.

    (1993)
  • M.M. Dacorogna et al.

    A geographical model for the daily and weekly seasonal volatility in the foreign exchange market

    J. Internat. Money Finance

    (1993)
  • J. Diebolt et al.

    Tail behavior of the stationary density of general non-linear autoregressive processes of order 1

    J. App. Probab.

    (1993)
  • Cited by (14)

    • A pairs trading strategy based on mixed copulas

      2023, Quarterly Review of Economics and Finance
      Citation Excerpt :

      Besides nonlinear dependence, these empirically verified regularities, known as stylized facts in the financial literature, are often described in a simpler way within the univariate context. Nevertheless one should notice that non-standard features in the univariate distributions are passed on, in a more complex version, to their multivariate counterpart: (1) asymmetric conditional variance with higher volatility for negative returns than for positive returns (Hafner, 1998); (2) conditional skewness (Chen et al., 2001); (3) leptokurdicity (Andreou et al., 2001; Tauchen, 2001); and (4) nonlinear temporal dependence (Campbell et al., 1997; Cont, 2001). To capture more closely the stylized facts cited above and allow for a more close-to-reality description of the multivariate distribution, Liew and Wu (2013) propose a pairs trading strategy based on two-dimensional copulas.

    • The volatility for pre and post global financial crisis: An application of computational finance

      2013, Modeling Applications and Theoretical Innovations in Interdisciplinary Evolutionary Computation
    • Handbook of Volatility Models and Their Applications

      2012, Handbook of Volatility Models and Their Applications
    View all citing articles on Scopus
    View full text