Time-varying quantile association regression model with applications to financial contagion and VaR

Wuyi Ye, Kebing Luo, Xiaoquan Liu

    Research output: Journal PublicationArticlepeer-review

    14 Citations (Scopus)


    This paper develops a quantile association regression model, which is able to capture the dynamic quantile dependence in the tails of conditional distributions. The association measure, the quantile-specific odds ratio (qor), captures the tendency of two random variables being simultaneously below specific quantiles. It is independent of marginal distributions and invariant to monotonic transformation, and enjoys methodological advantages over popular alternatives such as the copula. The ability of the qor measure to capture and forecast a range of different dependence structures is first shown via simulations. In the financial application, we implement the model and compute the qor on a daily basis to assess contagion for 10 stock markets during two recent crises. Our empirical results show that contagion exists during the US banking crisis between the US and all tested markets and between Greece and the tested European markets during the Euro crisis. Hence the model is able to capture the changes in quantile dependence between stock markets and offer a vivid description of market events. In addition, the model provides an accurate valuation of daily value-at-risk (VaR).

    Original languageEnglish
    Pages (from-to)1015-1028
    Number of pages14
    JournalEuropean Journal of Operational Research
    Issue number3
    Publication statusPublished - 1 Feb 2017


    • Copula
    • Finance
    • Financial crisis
    • Local polynomial regression

    ASJC Scopus subject areas

    • Computer Science (all)
    • Modelling and Simulation
    • Management Science and Operations Research
    • Information Systems and Management


    Dive into the research topics of 'Time-varying quantile association regression model with applications to financial contagion and VaR'. Together they form a unique fingerprint.

    Cite this