The impact of the COVID-19 outbreak on the connectedness of the BRICS's term structure

The study investigates how the COVID-19 pandemic altered interdependence and spillovers among the BRICS sovereign yield curves, with implications for diversification and risk management. Globalization has increased cross-market linkages, which tend to intensify during crises, potentially eroding diversification benefits. BRICS markets are important to global portfolios due to their size and growth, yet differ in structures and dynamics. The COVID-19 crisis created an unprecedented shock, with particularly severe effects on emerging markets, making BRICS an apt setting to study time-varying connectedness in bond markets. The research question centers on whether, how, and when pandemic waves influenced connectedness across BRICS term structures—overall and by yield-curve components (level, slope, curvature)—and the role of COVID-19-related media coverage (MCI) in this network. The study aims to quantify total and directional spillovers and to identify net transmitters and receivers of shocks across countries and curve factors, informing investors and policymakers about diversification, contagion, and policy transmission during systemic stress.

The literature is grouped into three strands. (1) COVID-19 and BRICS: Studies document pandemic-induced damage to emerging markets’ liquidity, volatility, exchange rate interdependence, and sovereign bond behavior (e.g., Gubareva, 2021; Haroon and Rizvi, 2020a,b; Topcu and Gulal, 2020; Xu and Lien, 2021; Janus, 2021; Zaremba et al., 2021, 2022; Będowska-Sójkaa and Kliber, 2022; To et al., 2022). Media-based measures such as the Coronavirus Media Coverage Index (MCI) are shown to matter for financial markets. (2) Connectedness within BRICS: Prior work identifies time-varying spillovers across BRICS assets and with global markets, often rising in crises, and heterogeneous roles across countries (e.g., Bouri et al., 2018; McIver and Kang, 2020; Dahir et al., 2020; Jiang et al., 2019; Mensi et al., 2016, 2017; Li et al., 2021; Hung, 2021; Esparcia et al., 2022), with China and Russia often influential. (3) TSIR connectedness: Research on yield curve co-movement and spillovers highlights the usefulness of decomposing the curve into level, slope, and curvature (Nelson-Siegel; Diebold and Li, 2006) and examining their interactions with macro/commodity shocks (e.g., Ioannidis and Ka, 2018; Gabauer et al., 2020; Gupta et al., 2020; Nazlioglu et al., 2020; Mensi et al., 2021; Umar et al., 2022a–d). Related studies use MCI to study TSIR connectedness (Aharon et al., 2022). The present paper fills a gap by focusing on BRICS sovereign yield curve components and COVID-19 uncertainty (MCI) using a TVP-VAR connectedness framework.

Data: Daily zero-coupon sovereign yields for Brazil, Russia, India, China, and South Africa from Bloomberg, with maturities at 3–360 months (3, 6, 12, 24, 36, 48, 60, 72, 84, 96, 108, 120, 180, 240, 360). Sample: 01/01/2020–02/28/2022, encompassing six COVID-19 waves. COVID-19 media data: RavenPack Coronavirus Media Coverage Index (MCI), a daily index (0–100) measuring the share of news about COVID-19. Yield curve estimation: Dynamic Nelson-Siegel per Diebold and Li (2006) to extract the three factors—Level (long-term), Slope (short-term), Curvature (medium-term). The loadings: level is unity across maturities; slope decays monotonically; curvature is hump-shaped, capturing medium-term. Stationarity of factor series is ensured via first differences, validated by ADF, PP, and KPSS tests. Connectedness framework: Time-Varying Parameter VAR (TVP-VAR) of Antonakakis and Gabauer (2017), an enhancement over rolling-window Diebold-Yilmaz connectedness, using stochastic volatility with Kalman filtering and forgetting factors. From the TVP-VAR’s VMA representation, generalized impulse response functions and generalized forecast error variance decompositions (GFEVD) are computed. Connectedness measures: Total Connectedness Index (TCI), directional TO and FROM connectedness for each variable, and NET (TO minus FROM) to identify transmitters (>0) and receivers (<0). The system includes BRICS TSIR components and the MCI, allowing assessment of how COVID-19-related news shocks transmit to yield curves across countries and curve factors over time and across pandemic waves.

- Descriptive characteristics: For level factors, average values are positive for all except China; slope averages are negative for all except Russia, implying generally upward-sloping curves; curvature means are negative for India and China and positive for others. Curvature exhibits the highest volatility among factors. Unit-root and stationarity tests (ADF, PP, KPSS) support stationarity after differencing. - Time-varying total connectedness: Connectedness fluctuates markedly, peaking before and during the first COVID-19 wave (March–June 2020) across level, slope, and curvature. Level and curvature connectedness persist longer; slope spikes and recedes quickly. Minimal increases appear at the onset of the second and third waves, with only marginal curvature spikes, while connectedness rises again in later waves (fifth and sixth), particularly for curvature near end-2021/early-2022. - Directional TO (contributors): During the first wave, Brazil, South Africa, and Russia dominate as net transmitters at the level; India contributes less and China is stable. Slope TO shows early spikes in South Africa and India; Russia’s TO spikes near the sixth wave, potentially foreshadowing geopolitical stress. Curvature TO is elevated pre-pandemic and early in the first wave for South Africa, Brazil, and India, declines mid-sample, and rises again ahead of the sixth wave, led by South Africa and with a final spike in Russia. The MCI’s TO remains relatively flat over most of the period. - Directional FROM (receivers): Level FROM is high and volatile early (Jan–Jun 2020) with successive peaks in India, China, Brazil, Russia; it falls and stabilizes afterwards, with upticks in South Africa and Russia between second and third waves, and a rebound in Russia before the sixth wave. For slope, China and India have high early FROM; Brazil is lowest; Russia, India, and South Africa peak at the first-wave onset; Russia spikes again at the sixth wave. Curvature FROM is generally lower than other components but increases at the start of the first wave (notably India) and shows small rises in later waves, with notable increases in South Africa (fifth wave) and Russia and India (sixth wave). - Net connectedness (NET): Brazil and South Africa are net transmitters; China and India are net receivers. Russia evolves from initially receiving/neutral to transmitting during the second wave and again at the end of the sample, coincident with the Russia–Ukraine invasion. The largest cross-country NET differences occur before and during the first wave; differences compress through much of 2021 and widen again into the sixth wave. - Implications: High connectedness during the first wave underscores diminished diversification benefits in BRICS sovereign bonds during systemic crises; later waves had milder, factor- and country-specific effects, with renewed elevations late in the sample.

The findings directly address the research question by showing that COVID-19 significantly altered the connectedness of BRICS sovereign yield curves, with the strongest effects concentrated around the first wave and renewed elevations in later waves. The decomposition into level, slope, and curvature reveals heterogeneous transmission channels: short-term (slope) shocks were abrupt and transient, while long- and medium-term (level and curvature) linkages were more persistent. Identifying Brazil and South Africa as net transmitters and China and India as net receivers clarifies the directionality of spillovers within BRICS bond markets during crisis periods. The elevated total and directional connectedness indicates that contagion risk and co-movements intensified under stress, eroding diversification benefits in BRICS sovereign debt. The use of MCI contextualizes how pandemic-related news pressure interacts with bond market dynamics. These insights are relevant for policymakers monitoring contagion channels and for investors managing cross-country bond portfolios, highlighting the need for dynamic hedging and allocation adjustments during crisis waves and evolving geopolitical risks.

This paper provides the first evidence on COVID-19-driven connectedness among BRICS term structure components linked to the Coronavirus MCI using a TVP-VAR framework. Contributions include: (1) documenting that total connectedness is time-varying and highest during the first pandemic wave, with renewed increases in late waves; (2) showing that level and curvature spillovers persist longer than slope; (3) identifying Brazil and South Africa as consistent net transmitters and China and India as net receivers, with Russia shifting to a transmitter role during later episodes; and (4) demonstrating diminished diversification benefits for BRICS sovereign bonds during crisis periods. Policy and practice implications involve the need for vigilant monitoring of bond market interdependencies across maturities, contingency planning for shock transmission, and dynamic portfolio strategies that account for shifting transmitter/receiver roles. Future research could extend the sample beyond early 2022, disentangle concurrent geopolitical shocks (e.g., Russia–Ukraine war) from pandemic effects, incorporate additional macro-financial drivers, and assess out-of-sample portfolio performance under varying connectedness regimes.

- Temporal and coverage constraints: The sample spans 01/01/2020–02/28/2022 and excludes Russian market data after 02/28/2022 due to the war, limiting post-invasion analysis. - Scope: Focused on BRICS sovereign yield curves and the MCI; other macroeconomic, policy, and global risk factors are not jointly modeled, which may confound attribution. - Model assumptions: TVP-VAR with generalized variance decompositions assumes linear relationships and relies on stochastic volatility specifications; results may be sensitive to specification choices. - Data frequency and proxies: Zero-coupon curves are model-based; the MCI proxies pandemic salience in media rather than epidemiological severity, which could diverge from actual health or policy shocks.