[Paper Review] Quantile cross-spectral measures of dependence between economic variables

Jozef Baruník, Tobias Kley|arXiv (Cornell University)|Oct 23, 2015

Market Dynamics and Volatility50 references9 citations

TL;DR

This paper introduces quantile cross-spectral analysis to detect frequency-domain dependence structures in economic time series that are invisible under traditional methods. By modeling dependence across quantiles and frequencies using quantile vector autoregressions, the authors propose new estimators with established asymptotic properties, enabling inference for nonlinear processes and revealing new patterns in bivariate stock market return dependence.

ABSTRACT

In this paper we introduce quantile cross-spectral analysis of multiple time series which is designed to detect general dependence structures emerging in quantiles of the joint distribution in the frequency domain. We argue that this type of dependence is natural for economic time series but remains invisible when the traditional analysis is employed. To illustrate how such dependence structures can arise between variables in different parts of the joint distribution and across frequencies, we consider quantile vector autoregression processes. We define new estimators which capture the general dependence structure, provide a detailed analysis of their asymptotic properties and discuss how to conduct inference for a general class of possibly nonlinear processes. In an empirical illustration we examine one of the most prominent time series in economics and shed new light on the dependence of bivariate stock market returns.

Motivation & Objective

• To address the limitation of traditional spectral analysis in detecting dependence structures that emerge only in specific quantiles of the joint distribution.
• To develop a frequency-domain method capable of capturing general, possibly nonlinear, dependence between economic variables across different parts of their joint distribution.
• To propose new estimators for quantile cross-spectral measures with rigorous asymptotic theory for inference.
• To demonstrate the method’s empirical relevance through an analysis of bivariate stock market returns.

Proposed method

• Proposes quantile cross-spectral analysis as an extension of classical spectral methods to the quantile regression framework.
• Uses quantile vector autoregressive (QVAR) processes to model how dependence varies across quantiles and frequencies.
• Defines new estimators for quantile cross-spectral density that capture dependence structures in the frequency domain.
• Establishes the asymptotic distribution of the estimators under general conditions for possibly nonlinear time series processes.
• Derives inference procedures, including confidence bands and hypothesis tests, for the estimated quantile cross-spectral measures.
• Applies the method to real financial data to detect time-varying, quantile-specific comovement patterns in stock returns.

Experimental results

Research questions

• RQ1 How can dependence between economic time series be detected in the frequency domain when it is confined to specific quantiles of their joint distribution?
• RQ2 What are the asymptotic properties of quantile cross-spectral estimators under general weakly dependent and possibly nonlinear processes?
• RQ3 How does the dependence structure between financial assets vary across different quantiles and frequencies?
• RQ4 Can quantile cross-spectral analysis reveal new comovement patterns in stock market returns not detectable through standard spectral methods?
• RQ5 What is the empirical relevance of quantile cross-spectral dependence in real-world economic data?

Key findings

• The proposed quantile cross-spectral estimators are consistent and asymptotically normal under general conditions, enabling valid statistical inference.
• The method reveals that dependence between bivariate stock market returns is significantly stronger in the lower and upper quantiles of the joint distribution at certain frequencies.
• Frequency-specific dependence structures emerge that are absent in the mean-based spectral analysis, indicating tail dependence at specific business cycle frequencies.
• The quantile cross-spectral approach detects nonlinear comovement patterns in financial returns that are invisible to classical spectral methods.
• Empirical results show that extreme market movements (e.g., crashes or rallies) are associated with heightened dependence at specific frequencies, particularly in the lower and upper tails.
• The method provides a more nuanced understanding of financial risk transmission by identifying frequency-varying, quantile-specific co-movements.