Time-frequency relationship between heterogeneous uncertainties and stock market extreme risk: new evidence from China

ABSTRACT

We empirically investigate the dynamic correlation between different uncertainties and stock market extreme risk across multiple time-frequency domains, providing novel evidence for the multiscale heterogeneity of popular uncertainty measures. Our findings reveal a more significant short-term relationship between the observable news-based economic uncertainty proposed by Baker et al. (2016) and China’s stock market extreme risk, while the latent economic uncertainty suggested by Jurado et al. (2015) dominates the long-term time horizon. Moreover, financial uncertainty is also a crucial source of stock market extreme risk that cannot be ignored.

KEYWORDS:

• Economic uncertainty
• financial uncertainty
• extreme risk
• time-frequency domain
• stock market

JEL CLASSIFICATION:

• G15
• C58

Acknowledgements

This research is supported by the China Postdoctoral Science Foundation (Grant No. 2023M730591). All authors contribute equally to this paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ The definition and distinction of uncertainty and risk were first proposed Knight (1921), which pointed out that the distribution of risks can be derived from prior probability or empirical statistics, while the distribution of uncertainty remains unknown.

² Jurado et al. (2015) believe that the adequacy of the observable uncertainty proxy relies on its correlations with the latent stochastic process, and it may vary over time based on movements in risk aversion or sentiment even in the absence of changes in economic fundamentals.

³ This equation effectively discriminates between the impacts of second- and first-order moments, thereby enabling precise identification of the effects of unknown shocks on volatility. The Markov chain Monte Carlo method is employed to estimate the equation.:

⁴ There are two commonly used ways to model the extremes: Block Maxima (BMM) and Peak Over Threshold (POT) approaches. Due to the volatility-clustering characteristics exhibited in financial time series, we choose the POT method for this study.

⁵ The continuous wavelet transform is an invaluable tool for investigating multiscale correlations, particularly in the analysis of nonstationary time series.:

⁶ http://www.policyuncertainty.com/all_country_data.html

Additional information

Funding

This work was supported by the China Postdoctoral Science Foundation [2023M730591].