Covid-19’s adverse effects on a stock market index

ABSTRACT

We perform a panel data analysis of 14 daily stock market indices during 01/21/2020 – 06/30/2020 to document a stock market index’s negative responsiveness to Covid-19’s spread variations. We find that a stock market index’s elasticity estimate is −0.028 (p-value <0.01) for local cumulative confirmed cases. As a stock market index tends to move with Covid-19’s local and non-local spreads, international efforts of containment are expected to pare stock market losses.

KEYWORDS:

• Covid-19
• daily stock market index
• market index elasticities
• panel data analysis

JEL CLASSIFICATION:

• G10
• G18
• I10
• I18

Acknowledgments

C.K. Woo’s research is funded by research grants (#4388 and #4400) from the Education University of Hong Kong. Without implications, all errors are ours.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ https://www.theguardian.com/world/2020/mar/20/chinese-inquiry-exonerates-coronavirus-whistleblower-doctor-li-wenliang

² We decide not to use a CAPM-based approach (e.g., Ramelli and Wagner 2020) that does not yield elasticity estimates that are of our primary interest.

³ To address concerns of spurious regression (Davidson and Mackinnon 1993), we use the panel unit root test proposed by Im et al. (2003) to reject the hypothesis that the regression residuals follow a random walk.

⁴ We perform the Hausman test (Wooldridge 2010, Chapter 10) to reject the hypothesis that equation (1)(1) $\mathrm{l}\mathrm{n}\left(I_{kt}\right)=S_j{\alpha }_j\mathrm{l}\mathrm{n}\left(X_{jkt}\right)+S_j{\beta }_{j}ln\left(Y_{jkt}\right)+S_j{\text{O}}_j\mathrm{l}\mathrm{n}\left(Z_{jkt}\right)+\mathrm{F}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{d}\mathrm{e}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{s}+{\epsilon }_{kt}$(1) should be based on random effects.

⁵ The number of cumulative cases in January 2020 is zero for European countries, see Figure 2 below.

⁶ If X_1kt > 100, α₁ = ∂ln(I_kt)/∂ln(X_1kt) is identical (within two digits) to the traditionally defined elasticity. To illustrate this point, consider the simple example of ln(y) = b ln(1 + n) where n is a non-negative number. As dln(y)/dln(n) = b [n/(1 + n)], b = 0.99 dln(y)/dln(n) when n = 100.

⁷ We have considered treating cumulative deaths and cumulative recoveries as separate regressors. However, ln(1 + number of cumulative cases) is highly correlated (r > 0.9) with ln(1 + number of cumulative deaths) and with ln(1 + number of cumulative recoveries), causing severe multicollinearity that leads to imprecise and counter-intuitive coefficient estimates.

⁸ For better or worse, Covid-19’s spread will naturally subside when a large percentage (e.g., 80%) of the surviving population have developed herd immunity, thus reducing the chance of an uninfected person becoming infected (https://time.com/5810454/coronavirus-immunity-reinfection/).

⁹ ‘While the official 12- to 18-month timeframe still stands, experimental Covid-19 inoculations for high-risk groups could be rolled out much earlier’ (https://www.theguardian.com/world/2020/apr/12/when-will-we-have-a-coronavirus-vaccine).

Additional information

Funding

This work was supported by the Education University of Hong Kong [4388,4400].