Municipal bond volatility spillover modeling with COVID-19 effects by hybrid integration of GARCH and machine learning: The connectedness of U.S. states and South African bond markets

Abstract

After the passage of the United States Africa Growth and Opportunity Act (AGOA), the emerging market economy of South Africa (SA) recorded substantial increases in duty-free exports to the U.S. Any loss or disruption of AGOA privileges could undermine SA as a trade partner with designated states in the U.S. For U.S. states most affected by the passage of the 2017-2018 Tax Cuts and Jobs Act (e.g., SALT states) along with the COVID-19 pandemic effects, the elimination of the tax exemption impacted how state governments assist in funding international trade. This paper presents a novel hybrid EGARCH conditional volatility and artificial neural network model to map the COVID-19 effect on volatility spillovers between the SA government bond market and state-issued U.S. municipal bonds traded in secondary markets. Our empirical investigation provides three innovations. First, this study addresses illiquidity and informed trading impairments by observing over 11.5 million municipal bond trades. Second, we provide new evidence that a radial basis function artificial neural network enhanced by conditional volatility and fundamental factors can effectively map the geographical bond market spillover transmission between SA and individual U.S. states. Third, on state-based muni trades, we report how the AR (1) process is weighted positively across all East coast states during the global pandemic. Lastly, the study details how COVID-19 and South African conditional volatility impacted the returns of recorded trades of municipal bonds in SALT states. A global summary of the findings concludes with a discussion of how precious metals trading contributes to the performance of state-based municipal bonds.