Impact of COVID-19 on jump occurrence in capital markets

The study examines how pandemic-related information, specifically daily confirmed COVID-19 cases and deaths, influences abrupt price jumps in equity index returns. Motivated by the unique uncertainty created by COVID-19 and the prominent role of government interventions (lockdowns, quarantines, mobility restrictions) in shaping economic activity and investor sentiment, the paper asks whether the scale (cases) or severity (deaths) of the pandemic better explains jump occurrences and whether effects are consistent across countries. The context spans six representative markets (China, France, Italy, Germany, the UK, the US) over pre-, during-, and post-pandemic periods, with the goal of understanding heterogeneous market reactions and the role of policy regimes in amplifying or dampening jump risk.

Prior COVID-19-market research largely focuses on volatility dynamics rather than discrete jumps, documenting heightened uncertainty and adverse effects on returns and economic activity (e.g., Albulescu 2021; Chowdhury 2022; Uddin et al. 2021). Studies linking COVID-19 indicators to stock returns generally find cases and deaths depress returns (Al-Awadhi et al. 2020; Ashraf 2020; Ftiti et al. 2021). Some works examine announcements of control measures and short-term return effects (Phan and Narayan 2020; Narayan et al. 2021). Emerging jump-focused studies find elevated jump risk during COVID-19 (Alqahtani et al. 2021; Liu et al. 2022; Zeng et al. 2022; Zhang et al. 2022), but they have not directly tied pandemic monitoring indicators to jump occurrence. Given that jumps reflect overreactions to unexpected information, effects on returns cannot be assumed to translate to jump behavior; government responses can moderate or intensify market jumps. This paper fills the gap by linking pandemic indicators to time-varying jump intensity.

• Model: The study employs the Autoregressive Jump Intensity (ARJI) model (Chan and Maheu, 2002) where returns comprise a diffusion component with GARCH(1,1) volatility and a Poisson jump component. The jump intensity is time-varying and follows an autoregressive process.
• Extension with COVID-19 indicators: The jump intensity equation is augmented with external variables representing pandemic information: daily confirmed cases (NC_t), lagged confirmed cases (NC_{t-1}), daily deaths (ND_t), and lagged deaths (ND_{t-1}). Four specifications are estimated, each including one indicator (contemporaneous or one-day lag) in the jump intensity.
• Estimation: Parameters are estimated by maximum likelihood. The conditional return density integrates over the Poisson-distributed number of jumps with time-varying intensity. A two-step procedure initializes GARCH parameters, followed by full ARJI/extended-ARJI estimation.
• Monte Carlo validation: Simulated series (T=3000) with about 5% jump days validate that MLE recovers true parameters closely (e.g., external variable coefficient ≈1.023 vs true 1) and that inferred jump intensity spikes align with inserted jumps, especially for externally driven shocks.
• Data: Daily close prices for six equity indices (SSE/China, CAC 40/France, DAX 30/Germany, MIB/Italy, FTSE 100/UK, S&P 500/US) from 2013-01-02 to 2023-12-31 (Wind). COVID indicators (daily confirmed cases and deaths) for 2020-01-03 to 2021-12-31 from WHO. Policy stringency from Oxford COVID-19 Government Response Tracker. Periods: pre-COVID (2018-2019), COVID (2020-2021), post-COVID (2022-2023). Earlier data (2013-2017) establish baseline jump intensity.
• Jump frequency extraction: Fit ARJI on the full sample to infer daily jump intensity. Standardize intensity and flag jump days when intensity exceeds a threshold (>3 standard deviations). Count annual jump frequencies by year and aggregate by pre-/during-/post-COVID periods.
• DID experiment: A Difference-in-Differences design compares monthly jump frequency between China (treatment: effective control during second wave) and a control group (Italy, France, Germany, UK, US) across an event window (2020-08-01 to 2021-08-01). Regress monthly jump frequency on treatment dummy, event dummy, and their interaction.
• Robustness: Weekly aggregation test relates weekly counts of extreme return days (returns > 2× own standard deviation) to weekly aggregated (differenced) cases and deaths via linear regression.
• Policy profiling: Descriptive analysis of stringency index and a proposed “stringency cases ratio” (stringency divided by cases per 1000) plus hierarchical clustering (Ward method) on four standardized metrics (cases, deaths, stringency, stringency/cases) to profile country strategies.

• Jump frequency surge: Annual jump frequencies increased during COVID across markets. Average annual jumps across six indices: pre-COVID 6.8, COVID 7.8, post-COVID 5.9.
• Temporal pattern: European markets (France, Germany, Italy, UK) show a common pattern—low pre-2020, peaking in 2020, then returning toward baseline. China is distinct: low in 2021, high in 2022, consistent with its policy U-turn and prolonged strict measures.
• DID evidence: The DID estimator for the China vs control comparison is −0.0218 (significant at 5%), indicating China’s effective control during the second wave is associated with a relative decline in monthly jump frequency versus increases in the control group.
• Indicator effects on jump intensity (extended ARJI): • China: Jump intensity responds to daily confirmed cases; death counts are not significant. • US: Neither cases nor deaths significantly affect jump intensity. • Europe and UK: Daily deaths significantly affect jump intensity. France shows the strongest sensitivity; the UK the weakest (notably significance for lagged deaths only).
• Distributional shifts: During COVID, return skewness becomes more negative and kurtosis increases across all indices, reflecting fatter tails and downside risk.
• Robustness (weekly extremes): Weekly confirmed cases significantly predict extremes in China; weekly deaths significantly predict extremes in France, Italy, and the UK; no significant predictors in the US; mixed/limited significance for Germany. These results align with extended ARJI findings.
• Policy profiling: Mean stringency (2020–2021) highest in China (69.73) and lowest in the US (56.94). The “stringency cases ratio” is extreme in China (495.56) and very low in the US (0.76), with Europe/UK in between (3.26–8.16). Clustering groups France/Germany/Italy/UK together, with China and the US as distinct singletons, reflecting divergent strategies and corresponding market sensitivities.

The results directly address the research question by demonstrating that jumps—discrete, abrupt market moves—were more frequent during COVID-19 and that their occurrence is systematically linked to pandemic monitoring indicators, contingent on country policy regimes. Investor anxiety about prospective control measures, triggered by daily pandemic updates, appears to be a key mechanism driving jumps. Differences across countries map to policy strategies: China’s zero-tolerance approach makes confirmed cases salient for jump risk; Europe/UK’s emphasis on avoiding healthcare overload makes death counts more informative; the US’s decentralized, less restrictive stance attenuates market sensitivity to pandemic indicators. These findings extend volatility-centered literature by isolating jump dynamics and highlight the importance of policy transmission channels in shaping market overreactions. The cross-validated evidence (frequency comparisons, DID, and robustness checks) strengthens causal interpretation that pandemic evolution and policy responses influenced jump risk.

The study extends the ARJI framework to embed COVID-19 indicators into the time-varying jump intensity and shows that market jumps became more frequent during the pandemic than in pre- and post-periods. It identifies heterogeneous informational drivers: confirmed cases matter for China, deaths for European markets and the UK, and neither indicator for the US. These differences align with country-specific strategies—zero tolerance (China), managing to healthcare capacity (Europe/UK), and less tolerance for restrictions (US). Contributions include: pioneering the link between pandemic monitoring indicators and jump occurrence; documenting cross-country heterogeneity tied to policy regimes; and providing policy-relevant insights into how information and anticipated interventions shape jump risk. Future work should broaden market coverage (including emerging markets), refine measurement of health data quality, and explore additional policy and sentiment channels to better understand drivers of market jumps under systemic shocks.

• Market coverage: Only six major indices were analyzed, potentially limiting generalizability and introducing selection bias.
• Data quality: Reliance on WHO-reported cases and deaths may be affected by underreporting or measurement issues, notably debates around COVID-19 death counts in China, which could influence estimates.
• Indicator scope: The study focuses on cases and deaths; other relevant indicators (testing rates, hospitalization, vaccination uptake, variant prevalence) and direct policy announcements could further explain jump dynamics.
• Model assumptions: The ARJI specification and jump detection thresholding may miss non-Poisson jump features or structural breaks; alternative jump models and thresholds could be explored.