The Impact of News-Based and Twitter-Based Economic Uncertainty on Realized Volatility: Asymmetric Effect with Threshold Quantile ARX Model

This research employs a two-regime threshold quantile autoregressive model with exogenous variables and a GARCH specification to investigate the dynamic relationship between perceived uncertainty and realized equity volatility. By utilizing Twitter-based and newspaper-based economic uncertainty measures, this paper offers insights into S&P 500 market dynamics across multiple distinct scenarios. The switching mechanism is governed by the previous day’s return. We make inferences and model selection within a Bayesian framework for each quantile level and utilize the multiple-try Metropolis algorithm to sample the threshold value, while other parameters are generated based on an adaptive Markov chain Monte Carlo (MCMC) procedure. Our diagnostic checking supports the proposed Bayesian methods. Findings reveal that these indicators capture various facets of uncertainty and display diverse behavior in stable and volatile markets. Specifically, the daily news-based economic policy uncertainty indicator shows a pronounced positive impact on realized volatility in the lower regime at higher quantile levels, while the Twitter-based market uncertainty indicator positively affects realized volatility across all quantile levels in the upper regime. The model selection results suggest that the news-based indicator outperforms other indicators at higher quantile levels, which correspond to more volatile market conditions, while the Twitter-based indicator serves as an effective explanatory variable when the S&P 500 is in a dormant or rising market. This paper presents the innovative use of Twitter-based and news-based indicators as dual proxies for economic uncertainty.