Relative Stock Market Performance during the Coronavirus Pandemic: Virus vs. Policy Effects in 80 Countries

Abstract

This paper examines relative stock market performance following the onset of the coronavirus pandemic for a sample of 80 stock markets. Weekly data on coronavirus cases and deaths are employed alongside Oxford indices on each nation’s stringency and government support intensity. The results are broken down both by month and by geographical region. The full sample results show that increased coronavirus cases exert the expected overall effect of worsening relative stock market performance, but with little consistent impact of rising deaths. There is some evidence of significantly negative stock market effects arising from lockdowns as reflected in the Oxford stringency index. There are also positive reactions to government support in March and December in the overall sample—combined with some additional pervasive effects seen in mid-2020 in Latin America.

1. Introduction

The coronavirus pandemic of 2020 represented the most devastating health crisis since the Spanish Flu of 1918–1919. Although stock markets reacted negatively to the spike in death rates at that time (Burdekin 2021), far more violent moves were seen in 2020. The record-breaking 33.7% drop in the S&P 500 stock market index between 19 February and 23 March 2020 was accompanied by massive declines in most other major stock markets around the world. This was quickly followed by record GDP declines in the second quarter of 2020 in the United States and many European countries. Nevertheless, most of these same countries then enjoyed a major stock market boom from the March lows in the face of massive central bank liquidity expansion (Burdekin 2020a).1

The initial US stock market plunge following the onset of the 2020 pandemic greatly outpaced the earlier experience under the Spanish Flu. Even though the Spanish Flu appears to have been far deadlier (cf. Burdekin 2020b), the US stock market initially continued to rise following the outbreak in late 1918 and merely fell back to summer 1918 levels after monthly deaths peaked later in the year. There was a sharp rebound afterwards, however, and the Dow Jones Industrial Average gained around 50% from late February 1919 through the end of the year. Pent-up demand following the end of the pandemic may well have been a factor in this outcome (Burdekin 2020a)—as could prove to be true once again in 2021. Nevertheless, a major difference between the 1918–1919 and 2020–2021 experiences remains the far lesser extent of the lockdowns and government intervention seen during the earlier episode.2

The main goal of this paper is to assess US stock market behavior not in isolation but in comparison with relative performance in other world stock markets. Additionally, in contrast to most prior empirical work, the analysis extends beyond the initial crisis period up until the final trades of December 2020. Following the next section’s discussion of the existing literature on stock market responses to the pandemic, Section 3 outlines the paper’s methodology. The data used to explain relative stock market performance on the basis of virus spread and government intervention are also detailed there. The main findings for our 80-country study are then presented in Section 4. Finally, Section 5 contains the paper’s conclusions.

2. Explaining the 2020 Stock Market Reactions

The enormous market turnaround after March 2020 occurred despite accelerating coronavirus spread around the world, with cases rising in every continent besides Antarctica. Cox et al. (2020, p. 20) conclude that the stock market rebound was largely driven by shifts in sentiment rather than fundamentals: “We find that the most likely candidate for explaining the market’s volatility during the early months of the pandemic is the pricing of stock market risk, driven by big fluctuations in risk aversion or sentiment unrelated to economic fundamentals or interest rates.” The rise in market volatility actually appears to precede the March shutdowns and travel bans, with Just and Echaust (2020), for example, identifying a shift from a low volatility regime to a high volatility regime in late February 2020.3 Alber (2020) and Ashraf (2020) seek to quantify stock market responses to rising virus cases and deaths early in the pandemic across as many as 64 countries. Although they each identify a more significant role for cases than deaths, Al-Awadhi et al. (2020) find both of these variables significant in China over a 10 January—16 March 2020 sample period. Panel data analysis by Khan et al. (2020), using weekly data through the end of March 2020, confirms significant effects of coronavirus cases across sixteen stock markets.4

Although the pandemic continued to play a key role, the extreme shifts seen early on in the pandemic were not typical of the later experience. Indeed, Phan and Narayan (2020) find support for a possible initial stock market over-reaction to the coronavirus that was already evident in data going through just early April 2020. This reflects Phan and Narayan’s (2020) finding that initial negative reactions to new cases and deaths are often followed by positive responses later on, with half of the 25 stock markets seemingly evincing positive reactions after reaching 100,000 cases and 100 virus-related deaths. A case in point is the remarkable US market rally from the March 2020 lows even as virus cases and deaths continued to ratchet higher through the end of the year. Meanwhile, whereas contagion between markets appeared to soar with the onset of the pandemic, Okorie and Lin (2021) find that this initial shock effect was already fading away by the end of March 2020.5 This only points to the importance of considering data extending beyond the initial crisis period.

In addition to the spread of the virus itself, other factors coming into play over time include government reactions to the public health emergency, such as the massive stimulus measures undertaken in the United States and Europe to provide economic support to individuals and businesses. The positive stock market effects themselves appear to have been far from uniform, with Harjoto et al. (2021) finding the beneficial effects to have been concentrated on larger firms.6 These government spending initiatives were accompanied by widespread monetary easing, which, in the US case, resulted in annual M2 money supply growth of 28% after February 2020. This was far greater than during the global financial crisis, and the highest rate of money emission seen since 1943 at the height of World War II (Greenwood and Hanke 2021).7 Australia was just one of many other countries embarking on similarly unprecedented stimulus efforts. Nevertheless, Rahman et al. (2021) find evidence of quite mixed stock market reactions with only the “JobKeeper” package (announced on 22 March 2020) eliciting a clearly significant, favorable stock market reaction.

Shifts in the level of economic support in either direction could be expected to trigger stock market reactions, with Chan-Lau and Zhao (2020), for example, emphasizing negative stock market reactions in cases where government stimulus was withdrawn when the number of daily COVID-19 cases remained relatively high. There is also the impact of government-ordered lockdowns, however, and Baker et al. (2020) conclude that it was these government restrictions on commercial activity, combined with social distancing, that account for the US stock market reacting so much more forcefully to COVID-19 than to the Spanish Flu (or prior pandemics). Given that stock markets would be expected to react to both stimulus measures and lockdowns, this paper provides new comparative evidence on the impact of each assessed over a broad range of world stock markets. The effects of rising coronavirus cases and deaths are also controlled for in the analysis.

We compare relative market performance to each country’s relative levels of virus cases, virus deaths, government stringency and economic support. We examine these relationships not only in the aggregate but also on a continent-by-continent basis for each month between March and December 2020. This encompasses the span between the extreme pessimism evident in market reactions early in the pandemic to the greater optimism emerging later on, spurred also by encouraging vaccine news in November and December 2020. Following the release of favorable clinical trials data in November 2020, the Pfizer COVID-19 vaccine was approved for emergency use in the United Kingdom on 2 December and in the United States on December 12.

3. Methodology and Properties of the Data

This paper sets virus effects on stock market performance against the relative importance of government interventions. Whereas most prior work was limited to data through just the first quarter of 2020, this combination of factors is assessed over a sample period extended from March 2020 through the end of the year.8 Like Khan et al. (2020), we utilize weekly data and employ pooled OLS (ordinary least squares) analysis. In order to allow for varying relationships over the year, panel regression analysis is applied not only for the full sample and also for the individual months between March 2020 and December 2020. In order to make the analysis as broad-based as possible, we incorporate 80 stock market indices drawn from all corners of the globe (Appendix A Table A1). Our focus is on the relative strength of each market during 2020 and we essentially include every available stock market, with just Venezuela and Zimbabwe excluded owing to the distortions associated with their respective hyperinflations that predated the pandemic. The stock market data are drawn from the Bloomberg terminal (with the exception only of the Jamaican market index, which is directly downloaded from the Jamaica Stock Exchange website at www.jamstockex.com, accessed on 10 April 2021).

The paper’s panel estimation assesses how each market’s relative strength is affected by variations in virus conditions and government measures. We utilize publicly available data on coronavirus virus and deaths, in each case scaled by population in order to achieve comparability across countries. The government policy responses are quantified on the basis of the data series available in Oxford University’s “Coronavirus Government Response Tracker” (Hale et al. 2020). The Oxford University data utilizes an ordinal scale reflecting the relative level of economic support and the relative level of stringency. Economic support encompasses income support, debt/contract relief for households, fiscal measures, and provision of international support. Meanwhile, the stringency index collates publicly available information on such policies as school and workplace closures, stay at home restrictions and travel bans to produce an additive score measured on an ordinal scale that varies from 0 to 100 (see Appendix A Table A2).

We regress each stock market’s relative strength index (market_RSI) on (i) increases in cases per 100,000 (g_cases100k), (ii) increases in deaths per 100,000 (g_deaths100k), and (iii) the Oxford series on the levels of government economic support (econsupport) and stringency measures (stringency). A lagged dependent variable and lagged cases and deaths are also included to allow for inertia. Increases in cases and deaths are used instead of cumulative cases and deaths owing to the non-stationarity of the latter series, which continuously rise over the sample period. Although cases and deaths are themselves both manifestations of the same virus spread, the timing of the two series is quite different. Testa et al. (2020), for example, show deaths typically not occurring until between two weeks and eight weeks after the onset of symptoms. Each of the Oxford series is measured on a scale of 0–100, with zero being lowest and 100 representing maximum intensity. The market_RSI puts each stock market on a scale from 1 to 80, with 1 being the top performer and 80 being the bottom performer. Relative strength is assessed on the basis of the dollar returns for each market index.9

In addition to the fluctuations in virus infections and policy responses over time, Table 1 reveals substantial geographical variation in 2020. Summary statistics over the March-December 2020 period for the full sample of 80 countries (Table 1a) are followed by summary statistics broken down according to the following geographical groupings:

Africa: Botswana, Egypt, Ghana, Kenya, Malawi, Mauritius, Morocco, Nigeria, South Africa, Tunisia and Zambia.

Australasia: Australia and New Zealand.

East Asia: China, Hong Kong, Indonesia, Japan, Malaysia, Philippines, Singapore, South Korea, Taiwan, Thailand and Vietnam.

Eastern and Southern Europe: Bulgaria, Croatia, Czech Republic, Cyprus, Estonia, Greece, Hungary, Lithuania, Latvia, Poland, Romania, Russia, Slovakia, Slovenia and Ukraine.

Latin America: Argentina, Barbados, Bermuda, Brazil, Chile, Colombia, Jamaica, Mexico, Peru and Trinidad.

North America: Canada and United Sates.

South Asia and Middle East: Bahrain, India, Israel, Kazakhstan, Kuwait, Lebanon, Oman, Pakistan, Qatar, Saudi Arabia, Turkey and United Arab Emirates.

Western Europe: Austria, Belgium, Denmark, Finland, France, Germany, Iceland, Ireland, Italy, Luxembourg, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and United Kingdom.

Table 1. Summary statistics for full sample and by region.

a. Full Sample
	count	mean	sd	min	max
market_RSI	4240	40.5	23.09493	1	80
g_cases100k	4160	41.44165	90.85629	−179.5596	1203.091
g_deaths100k	4160	0.7668794	1.798592	−3.2232	18.49932
stringency	4240	51.30944	27.16211	0	100
econsupport	4240	48.98585	34.41286	0	100
b. Africa
	count	mean	sd	min	max
market_RSI	583	42.48885	23.49986	1	80
g_cases100k	572	9.616895	22.52485	0	146.5333
g_deaths100k	572	0.2194387	0.6140417	0	5.347495
stringency	583	48.40607	28.56593	0	93.52
econsupport	583	37.92882	32.34679	0	100
c. Australasia
	count	mean	Sd	min	max
market_RSI	106	37.76415	21.59345	2	78
g_cases100k	104	1.502934	2.912689	0	13.40398
g_deaths100k	104	0.0392611	0.100428	−0.0039216	0.5803949
stringency	106	44.64877	26.9876	0	96.3
econsupport	106	60.14151	32.5555	0	100
d. East Asia
	count	mean	sd	min	max
market_RSI	583	39.4837	21.63178	1	80
g_cases100k	572	4.373834	11.17784	0	120.0784
g_deaths100k	572	0.0442898	0.0892606	0	0.6588728
stringency	583	50.63443	23.04173	0	100
econsupport	583	46.93396	34.95796	0	100
e. Eastern and Southern Europe
	count	mean	sd	min	max
market_RSI	795	40.54591	21.73602	1	80
g_cases100k	780	57.9161	117.8906	−0.4040714	748.3035
g_deaths100k	780	1.123939	2.549789	−0.4523048	17.31658
stringency	795	46.34185	26.29474	0	96.3
econsupport	795	53.01887	33.15399	0	100
f. Latin America
	count	mean	sd	min	max
market_RSI	530	42.88679	26.56565	1	80
g_cases100k	520	36.31582	51.12816	0	246.3407
g_deaths100k	520	1.13994	1.776847	0	15.79226
stringency	530	58.97732	30.16572	0	100
econsupport	530	41.58019	32.50223	0	100
g. North America
	count	mean	sd	min	max
market_RSI	106	37.73585	18.72778	7	74
g_cases100k	104	72.89453	102.6356	0	463.8697
g_deaths100k	104	1.402589	1.410535	0	5.45615
stringency	106	54.84009	26.07422	0	75.46
econsupport	106	53.89151	29.04037	0	75
h. South Asia and Middle East
	count	mean	Sd	min	max
market_RSI	636	40.47956	23.21433	1	80
g_cases100k	624	51.01512	86.15512	0	1203.091
g_deaths100k	624	0.3522827	0.4588852	0	2.98075
stringency	636	57.83392	28.00892	0	100
econsupport	636	44.65409	32.02359	0	100
i. Western Europe
	count	mean	sd	min	max
market_RSI	901	39.08768	23.12896	1	80
g_cases100k	884	68.73868	123.657	−179.5596	1075.221
g_deaths100k	884	1.357633	2.388899	−3.2232	18.49932
stringency	901	49.26022	25.26129	0	93.52
econsupport	901	59.43396	36.13918	0	100

The geographical groupings are primarily done by continent, but with Asia and Europe divided into two parts. In the case of Europe, the separation is between the more established nations of Western Europe and the primarily emerging economies of Eastern and Southern Europe. Although East Asia covers a range of development levels, all these nations share relative proximity with China and most are members of ASEAN (Association of Southeast Asian Nations). The countries in the South Asia and Middle East group are distinct both in terms of geographical location and by typically having lesser links with China.

The summary data broken down by region in Table 1b–i show some of the enormous variations in national experiences under the pandemic. For example, North America (Canada and United States) show an average 72.9 increase in cases per hundred thousand and 1.40 deaths per hundred thousand. Even with all these numbers scaled by population, the North America figures are over fifty times greater than the corresponding Australasia values. Cases and deaths in Africa and East Asia are also dramatically lower than in North America, while the other regions represent intermediate cases—but generally closer to the North American levels than Africa, Australasia, and East Asia.

The highest average levels of stringency are seen in Latin America, South Asia and Middle East, North America, and East Asia (in that order). The other regions are all below 50 on a scale of 0–100, with Australasia having the lowest average value of all. However, it should be emphasized that even Australia’s average stringency level of 44.6 is substantial, leaving it within 25% of the highest level of 59.0 registered in Latin America. Interestingly, despite having the lowest virus cases, virus deaths and average stringency levels, Australasia has the highest average level of government economic support (60.1). Western Europe is just behind with 59.4, followed by North America with 59.3 and Eastern and Southern Europe with 59.0. The nations of Africa and Latin America on average offered the least economic support, with index values of 37.9 and 41.6, respectively.

4. Empirical Findings

The panel estimation results have the relative strength of each market index is regressed on its own lagged value, current and lagged levels of additional cases (per hundred thousand of the population), current and lagged values of additional deaths (per hundred thousand of the population) and that country’s Oxford index values for stringency and economic support. As noted earlier, weekly data are employed in the regressions and the results are broken down both by month and by geographical region.10 Given that stronger market performance equates to a lower market_RSI value, the expected signs on virus cases and virus deaths is positive insofar as worse virus spread exerts a negative market effect. The expected sign on stringency would also be positive given that negative effects on economic activity should worsen the relative performance of the stock market. Finally, the expected sign on econsupport would be negative if the additional funding boosts relative market performance.

The full sample results in Table 2 show increased cases to exert the expected overall effect of worsening relative stock market performance. This effect is significant at the 95% confidence level or better overall as well as for April, August and October individually. Although there is some evidence of an offsetting effect of lagged cases, this may simply reflect the market impact of a rise in cases subsiding over the following month. New deaths are insignificant overall, but are significant and positive in May and August (while significant and negative in April and September). Such variation over the sample is unsurprising insofar as Phan and Narayan (2020) noted swings from negative to positive effects even in a study limited to the early months of the pandemic alone. There are similarly mixed findings for lagged deaths and no clear systematic effect.

Table 2. Stock market regressions for all 80 countries.

VARIABLES	March	April	May	June	July	August
L.market_RSI	−0.502 ***	−0.407 ***	−0.234 ***	−0.253 ***	−0.259 ***	−0.228 ***
	(0.044)	(0.055)	(0.039)	(0.064)	(0.043)	(0.050)
g_cases100k	−0.334	0.161 **	0.074	0.220	0.115	0.174 ***
	(0.289)	(0.064)	(0.094)	(0.194)	(0.127)	(0.054)
L.g_cases100k	0.068	−0.220 ***	−0.037	−0.296	0.127	0.140
	(0.487)	(0.051)	(0.107)	(0.242)	(0.116)	(0.168)
g_deaths100k	−3.122	−2.731 ***	6.880 ***	3.378	−1.557	2.776 *
	(4.282)	(0.651)	(2.172)	(3.498)	(2.377)	(1.526)
L.g_deaths100k	−7.161	2.931 ***	−3.302 *	7.889 **	−3.632 **	2.213 *
	(8.475)	(1.034)	(1.666)	(3.851)	(1.495)	(1.279)
stringency	0.161 ***	−0.275	−0.025	0.117	0.640*	−0.371
	(0.060)	(0.333)	(0.161)	(0.254)	(0.381)	(0.274)
econsupport	−0.144 *	−0.015	0.484 *	−0.058	0.104	0.106
	(0.073)	(0.142)	(0.244)	(0.420)	(0.200)	(0.588)
Constant	58.772 ***	81.153 ***	18.766	43.953	4.566	52.597
	(2.870)	(27.107)	(16.711)	(31.508)	(26.160)	(42.888)
Observations	320	320	400	320	400	320
R-squared	0.262	0.190	0.079	0.082	0.100	0.097
VARIABLES	September	October	November	December	Overall
L.market_RSI	−0.365 ***	−0.321 ***	−0.187 ***	−0.205 ***	−0.095 ***
	(0.040)	(0.037)	(0.038)	(0.056)	(0.016)
g_cases100k	0.110	0.107 ***	0.002	−0.003	0.030 **
	(0.131)	(0.039)	(0.048)	(0.011)	(0.012)
L.g_cases100k	0.060	−0.094	−0.019	−0.030 **	−0.035 ***
	(0.108)	(0.075)	(0.036)	(0.011)	(0.011)
g_deaths100k	−18.080 ***	−2.859	−0.086	1.273	−0.627
	(6.454)	(2.196)	(2.373)	(1.715)	(0.460)
L.g_deaths100k	−6.440	1.693	−0.449	−1.008	0.462
	(8.296)	(3.115)	(2.157)	(1.499)	(0.478)
stringency	−1.070 ***	0.158	0.190	−0.241	0.006
	(0.236)	(0.247)	(0.310)	(0.342)	(0.019)
econsupport	−0.114	0.228	0.053	−0.612 ***	−0.018
	(0.169)	(0.190)	(0.200)	(0.220)	(0.014)
Constant	126.146 ***	29.175 *	36.922	104.391 ***	45.238 ***
	(17.053)	(16.411)	(23.421)	(25.052)	(1.212)
Observations	320	400	320	400	4080
R-squared	0.193	0.136	0.061	0.059	0.013

Note: Robust standard errors are in parentheses; and *** p < 0.01; ** p < 0.05; * p < 0.1.

Stringency and econsupport are insignificant overall. This may reflect, in part, measurement error in the available Oxford indices, which would, in turn, bias down the estimated coefficients. Although stringency has the expected positive and significant effect in March and July, the significant negative effect for September suggests a favorable impact on relative market strength later in the year. Meanwhile, econsupport is significant with the expected negative sign in March and December, but significant with the opposite sign in May. With the major first and second waves of the virus in countries like the United States occurring in or around those same March and December months, it may well be that government support was seen as being more critical at those times; this could explain why there appear to be favorable market reactions in those particular months.

Appendix BTable A4, Table A5, Table A6, Table A7, Table A8 and Table A9 present results for the individual geographical groupings. Although Australasia and North America were included in the full sample results reported in Table 2, the number of countries in these two groups is insufficient for separate regression analysis. In terms of the overall findings across the different groupings, increased cases are significant with the expected positive sign for East Asia and Western Europe. Increased cases are insignificant overall in all other regions, with mixed significance observed in individual months. New deaths are significant overall only for Western Europe, with a negative sign. This counterintuitive finding seems to primarily derive from a strong effect indicated for July, when deaths were generally trending down, and may simply reflect a failure of markets to recover when the death rate improved. Elsewhere, there are generally very mixed findings across individual months that leaves little clear pattern. However, Africa has four months (March, April, September and December) for which deaths exert significant positive effects, i.e., weakening stock market RSI.

Stringency is significant overall with the expected positive (weakening RSI) effect only for Africa. Although it is otherwise significant overall just for East Asia, and with a negative sign, in terms of individual months it is significant there with the expected positive sign in March and only significant with a negative sign in September. In Eastern and Southern Europe as well as Latin America, stringency is significant with the expected positive sign in June and insignificant for other months. South Asia and Middle East also evinces a significant positive effect of stringency in June, but with offsetting indicated significant negative effects in May and September. Although the findings remain quite mixed, there is a tendency for stringency measures to hurt stock market performance earlier in the pandemic (as seen for March and June 2020).

Finally, economic support is never significant overall and the only clear case of beneficial stock market effects is seen for Latin America. In this region, econsupport is negative and significant in each of May, June and August (while positive and significant in July). Included here is the Brazilian support package of as much as USD 10 billion per month, more than for any other developing nation, that helped Brazil’s GDP exceed its pre-pandemic January 2020 levels by July—albeit it at the expense of potentially disastrous longer-term fiscal consequences (see Magalhaes and Pearson 2020). The lack of more consistent overall findings for economic support may reflect the fact that such support is usually applied when times are darkest. If the such support is applied in the face of an already weakening market, it will only appear successful on a statistical basis if it is able to immediately turn this trend around. This will not always be the case, even if these same interventions do help lay the groundwork for subsequent recovery.

5. Conclusions

Whereas the existing literature on the stock market effects of the pandemic has primarily focused on just the early crisis period in the first quarter of 2020, this paper assesses virus effects and government policy effects through December 2020 for a broad sample of 80 world stock markets. The overall results offer some support for growth in coronavirus cases and stringency hurting stock market performance, together with some evidence that 2020 government support measures helped the market both in March and late in the year. However, there is considerable variation across the different geographical regions considered in this paper. Clear-cut adverse effects of new coronavirus cases are found only for East Asia and Western Europe, for example. There is more widespread evidence of adverse effects associated with stringency, in overall terms for Africa and during mid-year in Eastern and Southern Europe, Latin America and South Asia and Middle East. Economic support’s favorable effects in March and December in the overall sample are generally not mirrored over the different geographical groupings. However, there are several significant months for the Latin American grouping around mid-2020.

The empirical findings are limited both by the total coverage being only ten months and the impossibility of fully capturing the tumultuous events through the limited array of variables included in the regressions. Nevertheless, markets are seen to generally react to growth in cases in the expected fashion and there is some relatively widespread evidence of adverse stringency effects on relative stock market performance. Although it is possible that positive reactions to government support in March and December in the overall sample may be driven by the US experience, there are also some pervasive effects in mid-2020 evident elsewhere—especially in Latin America. More generally, it is clear that the substantial variation by region would make it unwise to draw too many general conclusions from analysis focused more narrowly on either the United States alone or just more advanced industrial countries. It is hoped that the wide-ranging sample covered in this study can serve as a starting point in unpacking what has driven market performance globally under the coronavirus pandemic. A similarly broad-based approach might be employed in assessing the impact of the vaccine rollout in 2021, which began much more rapidly in countries like Israel, the United Kingdom and the United States than in most of the rest of the world.