Death and the stock market: international evidence from the Spanish Flu

ABSTRACT

The coronavirus pandemic in 2020 was the most devastating worldwide health threat since the 1918–1919 Spanish flu. Panel regression analysis for ten countries suggests that European and US stock markets reacted significantly, and negatively, to the surging death rates that were seen during the Spanish Flu. It is possible that the greater death rates for the Spanish Flu vis-a-vis the coronavirus account for stock market effects being more evident in 1918–1919 than in 2020.

KEYWORDS:

• Stock markets
• Spanish Flu
• Europe
• United States

JEL Codes:

• G15
• N22
• N24

Acknowledgments

The author is grateful to Ran Tao, Pierre Siklos, Bill Brown and two anonymous referees for helpful comments and suggestions and to Sam Harrison for excellent research assistance.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ Longer-run effects of the Spanish Flu and other pandemics are examined by Jordà, Singh, and Taylor (2020), who find evidence of sustained reductions in real rates of return.

² Although division by the size of the population would be required to derive comparable death rates, the empirical work examines year-on-year percentage changes – ensuring that the scaling effect is no longer relevant in the analysis. The October-November peak is itself a feature not only of the ten countries studied here but also the broader set of European countries examined by Ansart et al. (2009), who also rely on the monthly figures from Bunle (1954) but were not limited by the additional availability of stock market data.

³ This in no way represents any means of separating Spanish Flu deaths from other more standard causes of death. However, such precise information would, in any event, have been unavailable to contemporary stock market participants, whose information would have primarily involved knowledge of the epidemic and the accompanying surge in the death rate.

⁴ The unit root tests in Appendix A Table A.1 confirm that the stock returns series have unit roots but that the changes in stock returns do not. Thus, stock returns are I(1) – and cointegration techniques would be inadvisable owing both to the different orders of integration for the other variables and the sample’s relatively short time period. Although the Levin, Lin, and Chu (2002) procedure assumes a common unit root across countries, additional country-by-country unit root testing (not shown) indicates a common pattern for all variables except for the increase in deaths. The change in the increase in the deaths (like the change in the consol yield and change in market returns) remains stationary in every individual case in complete accordance with the Table A.1 aggregated test statistics.

⁵ This monthly series is expressed in per cent per annum and, like the stock market series, is drawn from Global Financial Data. Meanwhile, monthly money supply data were not always available – with Jeng, Butler, and Liu (1990) in any event finding no evidence of causal effects of money supply on stock prices over the 1921–1930 period except in the case of the United Kingdom (which was excluded from the present study due to deaths being available only on a quarterly basis).

⁶ Brown, Burdekin, and Weidenmier (2006) find that, while there is some instability associated with the onset of World War I, major increases in consol volatility are note seen until Britain’s exit from the gold standard in 1931 effectively put an end to debt obligations being fixed in real terms.

⁷ Additional allowance for lagged effects of deaths or the consol yield showed these lags to be uniformly insignificant.

⁸ With the exception of Italy and the United States, all other countries in the sample remained neutral throughout. Although Barro, Ursua, and Weng (2020) include data on estimated war deaths, these series are limited to an annual basis and have missing or zero values for a majority of the countries considered here.