The stabilizing effect of social distancing: Cross-country differences in financial market response to COVID-19 pandemic policies

Graphical abstract

Abstract

COVID-19 has had far-reaching global effects on the health and wellbeing of individuals on every continent. The economic and financial market response has been equally disastrous with high levels of volatility observed. This study explores the temporal relations between structural breaks, market volatility and government stay-at-home policy interventions and social distancing measures for 28 countries and their respective indices. We present results which indicate the establishment of stay-at-home policies influence sharp discontinuities in 15 of 28 markets (53.57 %) and increase market efficiency in 30 of 49 cases observed (61.22 %). These results indicate a small, statistically significant degree of stabilization in international financial markets responding to government stay-at-home policies and social distancing measures, a promising result for political actors concerned with economic performance during the public health response to the coronavirus 2019 pandemic.

1. Introduction

The coronavirus outbreak has impacted almost all countries and highlighted significant cross-country differences in approach and success in controlling and containing the virus. Governments around the world have been quick to establish strict control measures such as travel restrictions and city-wide lockdowns. But these responses also impact the economy, the banking and insurance system, and the global financial markets, in particular due to COVID-19′s unique global scope as a pandemic (Goodell, 2020).

Following such a significant global impact, literature on the economic effects of COVID-19 has grown rapidly. Extensive media coverage has increased volatility in equity markets and sectors perceived to be most at-risk (e.g., tourism, hospitality and retail) (Haroon and Rizvi, 2020). Global stock market risks have increased substantially and between-country differences are observed correlating with both severity of outbreak and policy interventions (Zhang et al., 2020). Akhtaruzzaman et al. (2020) find the transmission of financial contagion followed a similar pattern to that of the virus, with Chinese and Japanese firms acting as transmitters of shocks to G7 countries. Huber et al. (2020) find higher levels of risk aversion in financial professionals after the COVID-19 market crash. Further, companies with the name “corona” have experienced abnormal losses and sustained periods of trading volatility despite not been connected or responsible for the outbreak (Corbet et al., 2020a,b), suggesting that being even merely associated with coronavirus can spell bad news.

Various studies have explored COVID-19′s impact on financial market volatility (Albulescu, 2020; Baker et al., 2020a,b; Corbet et al., 2020a,b; Onali, 2020). Zaremba et al. (2020), for example, have demonstrated that government social distancing interventions have unanimously increased stock market volatility in international markets. The authors are aware of only one COVID-19 paper on market volatility which reports the Hurst exponent, a measure of long-range dependence/volatility persistence (Aslam et al., 2020). In other words, the degree to which a market displays memory and dependency of past prices on future prices and hence, predictability/opportunity for traders to capitalise on. For this reason, the Hurst exponent is used as a proxy for market inefficiency (Christodoulou-Volos and Siokis, 2006). This volatility (and persistence of it) can also act as an indicator for political actors concerned with financial market health and can, of course, be used by them for their own public (and private) purposes and agenda, and we can also use these as measures for politically induced (un)certainty.

Our study extends previous findings by exploring the temporal relation between the estimated structural breaks, market volatility and government stay-at-home policy interventions for 28 countries and their respective market indices across 6 continents. First, we estimate correlations between value-traded and COVID-19 variables such as population mobility, outbreak severity and policy interventions. Next, we associate the temporal sequence of events (e.g., WHO pandemic declaration, different stages of stay-at-home policy) to the detected and estimated structural breaks in traded value. Finally, we estimate the Hurst exponent as a proxy for market volatility and efficiency. In doing so, we find that in more than half of the observed cases, the establishment of stay-at-home policies appears to have elicited a strong but somewhat stabilising market response – a promising result for political actors concerned with economic performance during the public health response to the COVID-19 pandemic.

2. Materials and methods

2.1. Data description

2.1.1. Mobility

We incorporated country and regional-level mobility measures from the COVID-19 Community Mobility Reports (Google, 2020)(see also Chan et al. (2020a,Chan et al., 2020b for how this index was constructed) from 15 February onwards. Google calculates mobility by observing the percentage change in total visitors to 6 area/functional classifications: Retail and Recreation, Grocery and Pharmacy, Parks, Transit Stations, Workplaces, and Residential. The mobility measure is given as the percentage change in length of stay from the median value (of the corresponding day and location) taken between 3 January and 6 February. For the purpose of this study, we only use mobility measures on the Residential category as a proxy for the degree to which a population conforms to a stay-at-home policy and hence, restrains from travelling far from their place of residence.

2.1.2. Oxford Covid-19 government response tracker (OxCGRT)

The record for each country’s COVID-19 policy interventions was obtained from the Oxford Covid-19 Government Response Tracker (OxCGRT) database (Hale et al., 2021). The database records the level of strictness on various government policy responses such as school closures, cancellation of public events, public information campaigns, and international travel controls, from 1 January 2020 onwards. Each of these is scored on severity from 0 to 3 (3 = most severe) with the exception of public information campaigns, from 0 to 2 (2 = most severe). In our analyses, we only consider the timing and severity of government stay-at-home policies to focus on the impact of social distancing measures.

2.1.3. Market indices

All daily value-traded data was sourced from Bloomberg from 24 May 2019 to 16 June 2020. We use only from 1 January 2020 to 16 June in our analyses. The countries, stock market indices, and number of observations are presented in Table 1 . The observations are clustered by policy strictness (C1, C2 and C3) based on OxCGRT and a binary value (0,1) corresponding to pre/post-policy timing, respectively. C1 has 2631 observations across 22 countries with a 54 % split between pre- and post-policy observations. C2 has 2632 observations across 22 countries with a 49 % split between pre- and post-policy observations. C3 has 676 observations across 5 countries with a 36 % split between pre- and post-policy observations.

Table 1.

List of countries and corresponding stock index.

No.	Country	Index Shorthand	Observations (daily)
			C1	C2	C3
1	Australia	ASX	115	115	0
2	Austria	ATX	0	115	0
3	Belgium	BEL	0	107	0
4	Brazil	IBOVESPA	115	115	0
5	Canada	TSX	118	0	0
6	China	SSE / SZSE*	220	0	220
7	France	CAC	117	117	0
8	Germany	CDAX	117	117	0
9	Great Britain	FTSE	115	115	0
10	Greece	ASE	111	111	0
11	Hong Kong	HangSeng	113	0	0
12	India	BSE	115	115	115
13	Indonesia	IDX	112	0	0
14	Italy	ITLMS	118	118	118
15	Japan	NIKKEI	111	0	0
16	Korea	KOSPI	115	115	0
17	Malaysia	BURSA	113	0	0
18	Mexico	SPBMVIPC	0	116	0
19	New Zealand	NZX	112	112	0
20	Philippines	PSEI	111	111	111
21	Poland	WIG	115	115	0
22	Russia	MOEX	112	112	0
23	Singapore	SGX	115	115	0
24	Spain	IBEX	118	118	0
25	Switzerland	SMW	115	115	0
26	Thailand	SET	116	0	0
27	United States	NASDAQ / NYSE *	0	230	0
28	South Africa	JALSH	116	116	0
	Total	5939	2631	2632	676

Notes: C1, C2, C3 refer to the strictness of the stay-at-home policy with C3 as the strictest level of stay-at-home policy. C1 - recommend not leaving house; C2 - require not leaving house with exceptions for daily exercise, grocery shopping, and 'essential' trips; and, C3 - require not leaving house with minimal exceptions (e.g., allowed to leave once a week, or only one person can leave at a time, etc).

^*Cumulative Sum of Both Markets.

Value-traded (number of shares traded multiplied by price, summed across all securities in the exchange/index) was chosen over volume as it gives a better indication of the value/size of trades taking place; trade volume is a somewhat arbitrary figure when aggregating across securities. Value-traded was sourced at the entire exchange level where possible, in order to capture as much of a country’s trading activities as possible. Where unavailable, the value-traded of a composite index for a given country was used as a proxy for the entire exchange.

2.2. Methodology

2.2.1. Stage 1 – correlation

The correlation between traded value, coronavirus case statistics (number of confirmed and deaths) and population mobility is calculated using a 30-day moving average window. For each focal date, values of the variable pair from 15 days before and after (30 days inclusive) are used and the corresponding correlation is plotted. Relative (percentage) differences are found by taking the first difference of the log time series (daily change) and repeating the above.

2.2.2. Stage 2 – structural breaks

To calculate the structural breaks in traded value, each log time series is regressed to a linear time trend and a Wald test for a structural break with an unknown break is performed. The test is then repeated with the stay-at-home policy dates for each country on the corresponding time series. For countries with multiple policy stages, multiple breaks are tested as well as the individual breaks. For example, for the United Kingdom (GBR), we first conduct a Wald test on both dates (after regression), then another two Wald tests are performed for each date individually.

2.2.3. Stage 3 – daily log return

Converting the daily value-traded data to log returns normalises the market data so that all data are in a comparable metric despite originating from different time series data. Daily log return is defined as:

$$r_i(t)=\text{ln}{v}_i(t)-\text{ln}{v}_i(t-\Delta t)$$ (1)

where the closing value-traded of the index $i$ is represented by $v_i(t)$ and the time interval is $\Delta t$. For estimation of the daily log return, we used the R package with a lag of 1 time period (i.e., 1 day).1

2.2.4. Stage 4 – seasonal and trend decomposition using Loess (STL)

Seasonal and Trend Decomposition using Loess (STL) method is employed to decompose the time series data, as proposed by Cleveland et al. (1990). This operation extracts the seasonal variation ($S_i$), deterministic trend ($T_i$), and stochastic remainder ($R_i$) from the log return data.

For STL decomposition, we used the R package2 . The span (in lags) of the Loess window for seasonal extraction is set to 7, which is the number of trading prices in one week and is the minimum value (Cleveland et al., 1990).

2.2.5. Stage 5 – multifractal detrended fluctuation analysis (MFDFA)

When the Hurst exponent can change over time (Dajcman, 2012), multifractal analysis must be applied for a full description of scaling behaviour. Multifractal detrended fluctuation analysis (MFDFA) is a method used for the multifractal characterisation of non-stationary time series information (Kantelhardt et al., 2002), wherein conventional statistical properties of the time series, such as mean and variance, change over time. This study employs the procedure described by Açikgöz and Günay (2020) to compute the multifractality of the data and uses the MFDFA R package.3 The MFDFA analysis is applied to the stochastic remainder ($R_i$) of the value-traded log return data.

Fig. 1 shows the Generalised Hurst Exponent (top) and Mass Exponent (bottom) for NYSE daily value-traded data (C2 policy) for values of $q$ ranging from -10 to 10. The pre-policy and post-policy estimates are represented as black and red points, respectively. As seen in Fig. 1, there is slightly less multifractal content in the post-policy period indicated by lower variability of the generalised Hurst exponent. This could indicate market stabilisation, perhaps due to the implementation of a stay-at-home, social distancing policy. Similar figures are provided in the Supplementary Materials for all market indices and C1, C2, and C3 policies, where applicable.

Fig. 1.

Estimates of Gen eralised Hurst Exponent, h(q), and Mass Exponent, tau(q), for NYSE.

We calculate the range of the Generalised Hurst Exponent, the Hurst Exponent (when $q=2$), and the Market Deficiency Measure (MDM) for both the pre-policy and post-policy periods as well as the difference between periods. Greater variability indicates richer multifractality and an MDM close to zero indicates an efficient market. The Hurst exponent, $h\left(2\right),$ indicates one of three: persistence if $>0.5$; anti-persistence if $<0.5$; and random-walk behaviour if equal to $0.5$. If a financial market becomes more efficient (as measured by MDM) following the introduction of a stay-at-home policy, it could indicate that the social distancing measure has had a somewhat stabilising effect on the market in question.

MDM is defined as:

$$MDM=\frac{1}{2}\left(\left|H\left(-10\right)-0.5\right|\right)+(\left|H\left(10\right)-0.5\right|)$$ (2)

3. Results

Fig. 2 displays the daily value-traded (in logs) of the NYSE in conjunction with various COVID-19 statistics, including confirmed deaths, and infections (also in logs) through the dates beginning 1 January 2020 and ending near the end of June 2020. Visually, we observe a somewhat positively trending relationship in traded value as the number of confirmed cases (particularly in the United States) increases. As the number of confirmed cases levels out (around mid-March), the trend direction changes to negative. In addition, we also observe that variability of value-traded appears to increase substantially after this point in time, indicating a delayed market response to the COVID-19 pandemic.

Fig. 2.

Daily traded value (in logs) of the New York Stock Exchange (NYSE) overlayed with various COVID-19 statistics, including confirmed cases and deaths, from 1 January 2020 to mid-June 2020.

Fig. 3 plots the correlation between daily log value-traded with daily log of COVID-19 statistics. Also marked on each of the graphs (if applicable) are sets of vertical lines that indicate the date where a stay-at-home policy was introduced in that respective country. A green line represents recommendations by health experts to not leave the house (the least strict policy, C1). Yellow represents a policy which requires citizens to stay home with some exceptions, including exercise, grocery shopping and other “essentials,” (C2). Orange represents very strict stay-at-home measures (C3), where people are required by law to stay home, with minimal exception. In this sense, these three levels (C1, C2 and C3) of stay-at-home policy represent increasing levels of social distancing measures, respectively. Markers on each graph represent a significant correlation at 10 %. Correlation significance seems to predominantly lie at points where correlations are greater than 0.4 and less than -0.4. We also show similar correlation graphs, using differenced COVID-19 log variables with log of value-traded and then again with differenced log COVID-19 variables and differenced log value-traded, in the Supplementary Materials.

Fig. 3.

Log trade volume with daily log COVID-19 case statistics. Markers indicate the correlation is statistically significant below 10 %.

The correlation plots in Fig. 3 for the United States tell the same story as in Fig. 2. Namely, there is a sharp increase in traded value between February and March, which then subsides through April and May while cases and deaths continue to rise, before value-traded once again starts to increase through to June. Evidently, most other countries follow a similar pattern, likely due to the substantial correlation between global stock markets. One clear exception is China, which experiences the initial increase in value-traded earlier in January, likely due to experiencing a local outbreak and corresponding stay-at-home policy introduction before other countries. Value-traded in India, Philippines, and Indonesia display noticeably subdued reactions to US and global cases. South Korea appears to have high trading activity extended through to April, with subsequent reduction also subdued compared to others.

Next, we focus on detected structural breaks in value-traded data. Fig. 4 displays our entire sample of countries’ value-traded through time, with the vertical blue line indicating a detected unknown break in the time series based on significant (p-value < 0.01) supremum Wald tests. For clarification, Table 2 confirms at what dates these occur. Again, countries that employed stay-at-home measures are also shown on their respective graphs by either green, yellow, or orange vertical lines (C1, C2, and C3, respectively) at time the policy was implemented. For half of the countries we analyse (14 out of the 28 countries), the detected unknown breaks occur around late February 2020 (between 21 February and 27 February), approximately 2–3 weeks before WHO declared COVID-19 a pandemic. Interestingly, the structural break for SSE and SZSE (China) occurs on 27 December 2019 and 4 February 2020, respectively, suggesting a somewhat early response in contrast to other international markets.

Fig. 4.

Correlation between daily log traded value with daily log of COVID-19 statistics (deaths, confirmed cases, etc.) for all 28 countries and their respective market index.

Table 2.

Date of identified structural breaks for all of the 28 countries and their respective market index.

Country	Market Index	Break date	p-val
USA	NYSE	24-Feb-20	<0.001
USA	NASDAQ	21-Feb-20	<0.001
CAN	TSX	24-Feb-20	<0.001
GBR	FTSE	24-Feb-20	<0.001
DEU	CDAX	24-Feb-20	<0.001
AUT	ATX	9-Oct-19	0.003
CHE	SMW	21-Feb-20	<0.001
FRA	CAC	24-Feb-20	<0.001
BEL	BEL	20-Apr-20	<0.001
POL	WIG	24-Feb-20	<0.001
GRC	ASE	27-Aug-19	<0.001
ZAF	JALSH	24-Feb-20	<0.001
RUS	MOEX	25-Feb-20	<0.001
IND	BSE	5-Dec-19	0.0006
BRA	IBOVESPA	30-Jan-20	<0.001
CHN	SSE	27-Dec-19	<0.001
CHN	SZSE	4-Feb-20	<0.001
HKG	Hang Seng	9-Jan-20	<0.001
TWN	TWSE	26-Feb-20	<0.001
JPN	NIKKEI	25-Feb-20	<0.001
KOR	KOSPI	8-Jan-20	<0.001
SGP	SGX	27-Feb-20	<0.001
MYS	BURSA	9-Apr-20	<0.001
PHL	PSEI	23-Apr-20	0.0006
IDN	IDX	8-Apr-20	<0.001
AUS	ASX	19-Feb-20	<0.001

Note: Statistical significance (p-val) in the right-most column.

Next, we statistically test for known structural breaks at the point where stay-at-home policies are implemented. Table 3 displays the Chi-square value based on Wald tests at each stay-at-home policy stage to test whether a structural break occurs at this point. Countries that implement more than one stay-at-home policy are tested for structural breaks at each stage individually (individual break column) and collectively (break/s) if that particular country adopted multiple social distancing measures during COVID-19. For some countries, changes in stay-at-home measures were modified too quickly which prevented structural break analysis due to a limited number of observations. From Table 3, we observe significant structural breaks for 15 of the 28 countries at each policy stage (i.e., 53.57 %). Namely, the US, Canada, Germany, Switzerland, France, Spain, Italy, Belgium, Russia, Hong Kong, Japan, Korea, Malaysia, Philippines, and Indonesia. This indicates that in more than half of the observed cases, the introduction of a government social distancing measure has elicited a significant financial market response.

Table 3.

Testing for known structural breaks at the implementation of stay-at-home policies (C1, C2, C3) for all 28 countries and their respective market index.

Country	Policy stage	Date	Break(s)	Individual break	SD test (one-tailed)
USA (NYSE)	C2	15-Mar	χ2 = 22.54; p < 0.001		+ve, p = 0.689
USA (NASDAQ)			χ2 = 0.35; p = 0.841		+ve, p = 0.725
MEX	C2	30-Mar	χ2 = 0.125; p = 0.939		-ve, p = 0.073
CAN	C1	14-mar	χ2 = 22.47; p < 0.001		-ve, p = 0.222
GBR	C1	13-May	χ2 = 14.57; p = 0.006	χ2 = 2.339; p = 0.310	+ve, p = 0.804
GBR	C2	23-Mar		χ2 = 5.536; p = 0.063	-ve, p = 0.409
DEU	C1	09-mar	χ2 = 61.13; p < 0.001	χ2 = 30.21; p < 0.001	-ve, p = 0.142
DEU	C2	21-mar		χ2 = 2.36; p = 0.307	+ve, p = 0.689
AUT	C2	16-Mar	χ2 = 0.74; p = .692		-ve, p<.001
CHE	C1	27-apr	χ2 = 47.25; p < 0.001	χ2 = 18.46; p < 0.001	+ve, p = 0.985
CHE	C2	17-mar		χ2 = 26.32; p < 0.001	+ve, p = 0.950
FRA	C1	11-May	χ2 = 43.41; p < 0.001	χ2 = 9.96; p = 0.007	+ve, p = 0.768
FRA	C2	17-Mar		χ2 = 11.21; p = 0.004	+ve, p = 0.608
ESP	C1	27-May	χ2 = 35.33; p < 0.001	χ2 = 7.15; p = 0.028	+ve, p = 0.590
ESP	C2	14-Mar		χ2 = 3.92; p = 0.141	+ve, p = 0.859
ITA	C1	04-May	Break1 and 2: χ2 = 220.88; p < 0.001	χ2 = 40.33; p < 0.001	+ve, p = 0.865
ITA	C2	23-Feb	Break1 and 3: χ2 = 106.92; p < 0.001	χ2 = 53.84; p < 0.001	-ve, p = 0.457
ITA	C3	20-Mar		χ2 = 47.69; p < 0.001	+ve, p = 0.789
BEL	C2	18-Mar	χ2 = 16.18; p < 0.001		+ve, p = 0.547
POL	C1	9-Apr	χ2 = 12.31; p = 0.015	χ2 = 4.68; p = 0.096	-ve, p = 0.157
POL	C2	31-Mar		χ2 = 7.50; p = 0.024	-ve, p = 0.1
GRC	C1	30-May	χ2 = 32.74; p < 0.001	χ2 = 21.2; p < 0.001	+ve, p = 0.793
GRC	C2	23-Mar		χ2 = 8.63; p = 0.013	+ve, p = 0.785
ZAF	C1	29-may	χ2 = 6.80; p = 0.147	χ2 = 4.464; p = 0.107	+ve, p = 0.849
ZAF	C2	26-mar		χ2 = 0.496; p = 0.781	-ve, p = 0.112
RUS	C2	5-Mar	χ2 = 67.92; p < 0.001	χ2 = 47.75; p < 0.001	-ve, p = 0.021
RUS	C3	30-Mar		χ2 = 3.38; p = 0.184	+ve, p = 0.113
IND	C1	26-jan	χ2 = 20.81; p = 0.002	χ2 = 9.77; p = 0.008	+ve, p = 0.514
IND	C2	04-may		χ2 = 8.44; p = 0.015	+ve, p = 0.999
IND	C3	22-mar		χ2 = 6.38; p = 0.041	+ve, p = 0.999
BRA	C1	13-mar	χ2 = 16.89; p = 0.002	χ2 = 0.203; p = 0.904	+ve, p = 0.644
BRA	C2	05-may		χ2 = 4.436; p = 0.109	-ve, p = 0.133
CHN (SSE)*	C1	23-Jan	Insufficient observations	χ2 = 145.39; p < 0.001	+ve, p = 0.999
CHN (SSE)	C3	1-Feb		χ2 = 142.24; p < 0.001	+ve, p = 0.999
CHN (SZSE)*	C1	23-Jan		χ2 = 127.01; p < 0.001	ve, p = 0.435
CHN (SZSE)	C3	1-Feb		χ2 = 121.60; p < 0.001	+ve, p = 0.552
HKG	C1	8-Feb	χ2 = 21.5; p < 0.001		-ve, p<.001
JPN	C1	07-apr	χ2 = 14.41; p < 0.001		-ve, p = 0.208
KOR	C1	23-Feb	χ2 = 66.21; p < 0.001	χ2 = 65.09; p < 0.001	-ve, p = 0.279
KOR	C2	21-Mar		χ2 = 29.97; p < 0.001	-ve, p = 0.200
SGP*	C1	3-Apr	insufficient observations	χ2 = 7.31; p = 0.026	+ve, p = 0.675
SGP	C2	8-Apr		χ2 = 10.90; p = 0.004	+ve, p = 0.696
MYS	C1	18-Mar	χ2 = 88.43; p < 0.001		-ve, p = 0.107
THA	C1	21-Mar
PHL*	C1	29-May	Break1 and 2: χ2 = 48.16; p < 0.001	χ2 = 33.11; p < 0.001	+ve, p = 0.976
PHL	C2	15-Mar	Break1 and 3: χ2 = 48.25; p < 0.001	χ2 = 2.50; p = 0.29	-ve, p = 0.470
PHL	C3	18-Mar		χ2 = 2.73; p = 0.26	+ve, p = 0.638
IDN	C2	10-Apr	χ2 = 78.48; p < 0.001		-ve, p=<0.001
AUS	C1	24-Mar	χ2 = 9.89; p = 0.042	χ2 = 1.98; p = 0.372	-ve, p = 0.05
AUS	C2	2-Apr		χ2 = 7.62; p = 0.022	-ve, p<.001
NZL*	C1	21-Mar	insufficient observations	χ2 = 9.98; p = 0.007	+ve, p = 0.997
NZL	C2	23-Mar		χ2 = 6.50; p = 0.039	+ve, p = 0.524

Notes: C1 - recommend not leaving house; C2 - require not leaving house with exceptions for daily exercise, grocery shopping, and 'essential' trips; and, C3 - require not leaving house with minimal exceptions (e.g., allowed to leave once a week, or only one person can leave at a time, etc).

^{^}Taiwan and Sweden did not impose any stay at home measures during our study period (Hale et al., 2021).

^*Insufficient observations between policy dates to test for structural breaks.

Fig. 5, Fig. 6, Fig. 7 show the generalised Hurst exponents for the pre-policy and post-policy periods for all 28 countries at policy restrictions of C1, C2, and C3, respectively. As it can be seen, there is a high degree of variability in market volatility across each of the 28 countries and their market indices. In some cases, stay-at-home policies decrease multifractality. In others, they appear to increase volatility.

Fig. 5.

Generalised Hurst exponent for each market index, from q = -10 to 10, for pre- (black) and post-(red) level 1 (C1) stay-at-home policy recommendation.

Fig. 6.

Generalised Hurst exponent for each market index, from q = -10 to 10, for pre- (black) and post-(red) level 2 (C2) stay-at-home policy recommendation.

Fig. 7.

Generalised Hurst exponent for each market index, from q = -10 to 10, for pre- (black) and post-(red) level 3 (C3) stay-at-home policy recommendation.

Table 4, Table 5, Table 6 list the salient measures of market turbulence for C1, C2, and C3 policies, respectively. As it can be seen, the estimated Hurst exponent, $h(2$), varies greatly across countries. Interestingly, the post-policy exponent is not always higher than the pre-policy exponent which indicates a reduction in multifractal content and hence, potential market stabilisation following social distancing policy interventions. The country with the greatest MDM (indicating greatest inefficiency) is given in bolded red text. In all but the C3 policies, the most inefficient market does not remain the same between the pre-policy and post-policy period indicating that at least for the most inefficient and turbulent financial markets during the pre-policy stage, the introduction of a stay-at-home policy appears to elicit a stabilising response. For the C1, C2, and C3 policies, the MDMs for 15 in 22 markets (68.18 %), 14 in 22 markets (63.63 %), and 1 in 5 markets (20 %) fell between the pre-policy and post-policy periods, respectively. From this, the majority (61.22 %) of stay-at-home policies (C1, C2, and C3 combined) have elicited a stabilising response in the global financial markets.

Table 4.

Market turbulence measures including Generalised Hurst exponent range, Hurst exponent estimate h(2), and market deficiency measurements (MDM) for pre- and post- level 1 stay-at-home policy recommendation.

Country	Δh pre-c1	Δh post-c1	DΔh c1	h(2) pre-c1	h(2) post-c1	MDM pre-c1	MDM post-c1
AUS	4.0169	2.9093	−1.1076	0.4144	0.6124	1.9161	1.377
AUT
BEL
BRA	1.4269	1.879	0.4521	0.5772	0.5658	0.63335	0.83505
CAN	2.3451	1.2688	−1.0763	0.6056	0.6063	1.0625	0.533
CHE	3.6024	4.1079	0.5055	0.5003	0.5352	1.66825	1.93735
CHN*	2.2058	2.9535	0.7477	0.3401	0.5202	1.00065	1.40185
DEU	3.8149	1.4878	−2.3271	0.4219	0.3976	1.8045	0.63735
ESP	2.2515	1.6976	−0.5539	0.3644	0.0987	1.0343	0.7583
FRA	0.9419	1.4251	0.4832	0.482	0.5444	0.4122	0.59305
GBR	2.5293	1.3095	−1.2198	0.3203	0.2927	1.1458	0.5819
GRC	2.7952	0.7713	−2.0239	0.5149	0.8875	1.2531	0.5804
HKG	2.8819	3.0253	0.1434	0.7228	0.3576	1.3215	1.395
IDN
IND	4.0353	2.4559	−1.5794	0.3646	0.3991	1.89945	1.07025
ITA	3.3455	1.6372	−1.7083	0.3813	0.4533	1.51185	0.69025
JPN	2.7525	0.9378	−1.8147	0.4537	0.7666	1.22595	0.6108
KOR	1.4429	2.607	1.1641	0.7645	0.4755	0.6337	1.184
MEX
MYS	1.7032	1.3594	−0.3438	0.5408	0.5182	0.74755	0.6068
NZL	2.3061	1.5904	−0.7157	0.5338	0.4516	1.03175	0.64385
PHL	2.6753	1.5037	−1.1716	0.4573	0.5012	1.2177	0.6693
POL	2.7068	0.709	−1.9978	0.4477	0.5998	1.21145	0.2882
RUS
SGP	4.1095	3.6012	−0.5083	0.3266	0.4098	1.903	1.65255
THA	1.8283	2.4459	0.6176	0.3814	0.4785	0.7822	1.1247
USA*
ZAF	2.4087	0.6993	−1.7094	0.4763	0.5224	1.06655	0.2583

^*SZSE and NYSE used.

Table 5.

Market turbulence measures including Generalised Hurst exponent range, Hurst exponent estimate h(2), and market deficiency measurements (MDM) for pre- and post- level 2 stay-at-home policy recommendation.

Country	Δh pre-c2	Δh post-c2	DΔh c2	h(2) pre-c2	h(2) post-c2	MDM pre-c2	MDM post-c2
AUS	3.2377	2.2698	−0.9679	0.34	0.5105	1.51255	1.0451
AUT	2.9772	1.7676	−1.2096	0.0088	0.4079	1.3401	0.78925
BEL	2.0411	2.6361	0.595	0.6674	0.5268	0.8834	1.1797
BRA	3.3571	2.2016	−1.1555	0.4319	0.3223	1.44855	0.99035
CAN
CHE	1.9044	2.7228	0.8184	0.5755	0.5459	0.85735	1.24315
CHN*
DEU	1.845	0.7957	−1.0493	0.4733	0.3496	0.85755	0.2878
ESP	2.0775	1.5814	−0.4961	0.2946	0.5651	0.93895	0.73035
FRA	1.0448	1.8527	0.8079	0.4389	0.4195	0.4424	0.78205
GBR	2.0057	1.5639	−0.4418	0.4536	0.3908	0.91035	0.66475
GRC	1.2149	1.4762	0.2613	0.5787	0.4467	0.5476	0.6331
HKG
IDN	1.1898	1.4703	0.2805	0.4844	0.5812	0.47425	0.6472
IND	3.1909	1.8782	−1.3127	0.3223	0.1729	1.46715	0.8098
ITA	2.459	1.6573	−0.8017	0.3689	0.4318	1.11295	0.70205
JPN
KOR	2.0704	1.6463	−0.4241	0.5017	0.3347	0.876	0.70925
MEX	2.761	1.9099	−0.8511	0.5046	0.3186	1.2823	0.8678
MYS
NZL	2.3061	1.5904	−0.7157	0.5338	0.4516	1.03175	0.64385
PHL	3.6006	3.6253	0.0247	0.6019	0.5871	1.70325	1.6967
POL	3.2403	4.4111	1.1708	0.4485	0.5512	1.4568	2.07245
RUS	0.9156	2.0357	1.1201	0.3064	0.8047	0.3922	0.88545
SGP	4.1158	1.7406	−2.3752	0.3443	0.3767	1.90325	0.7698
THA
USA*	2.387	1.8917	−0.4953	0.3702	0.515	1.0825	0.8311
ZAF	2.4081	3.8609	1.4528	0.3638	0.5839	1.0433	1.79295

^*SZSE and NYSE used.

Table 6.

Market turbulence measures including Generalised Hurst exponent range, Hurst exponent estimate h(2), and market deficiency measurements (MDM) for pre- and post- level 3 stay-at-home policy recommendation.

Country	Δh pre-c3	Δh post-c3	DΔh c3	h(2) pre-c3	h(2) post-c3	MDM pre-c3	MDM post-c3
AUS
AUT
BEL
BRA
CAN
CHE
CHN*	1.7178	2.8185	1.1007	0.3665	0.6715	0.7491	1.2352
DEU
ESP
FRA
GBR
GRC
HKG
IDN
IND	2.2191	2.7046	0.4855	0.5742	0.4106	1.01065	1.2082
ITA	2.9919	2.1533	−0.8386	0.4031	0.363	1.3305	0.9367
JPN
KOR
MEX
MYS
NZL
PHL	3.694	4.5356	0.8416	0.6043	0.3313	1.75035	2.13265
POL
RUS	1.1306	1.4463	0.3157	0.5987	0.5737	0.54305	0.62335
SGP
THA
USA*
ZAF

^*SZSE and NYSE used.

4. Discussion

In interpreting our results, we see it is essentially about how individuals deal with uncertainty brought by the COVID-19 pandemic and the political reaction to it, as well as the feedback between the two. As Underhill (1991) eloquently explains,

The notions of self-regulation and automaticity which permeate the literature on markets implies the existence of a separate and autonomous sphere untainted by mercenary political interests (p. 198).

This untainted version of markets free from political influence is far from reality, as “[t]he organization of finance is a consequence of political decisions, a stake of political struggles, and a source of political interests and conflicts” (Davis, 2012, pp. 38). In other words, at every stage and phase of the political decision-making process: from organization, process, to redistributions of finance, all serve as both an input and output of policy; hence, studying one without due consideration of the other is like trying to understand tidal patterns in isolation from earth’s orbit around the sun and the moon’s orbit around earth – nonsensical. Political decisions and interventions without a doubt have at least some minor and lingering effects on markets if not completely defining “the structures, processes, and institutions within and around organizations that allocate power and resource control among participants” (Davis, 2005, pp. 143). Thus, understanding how politics affects markets is key to greater understanding of adverse market conditions (such as volatility and its persistence), which could serve to arm political actors in their debates (see e.g., Hirschman, 1991 for his triad of political argumentation under the perversity, futility, and jeopardy theses). During COVID-19, things have not been so different with what some would call the politicisation of public health and science more generally (Abbasi, 2020; Halpern, 2020; Peretti-Watel et al., 2020; Shumba et al., 2020), after all, politicians would still need support from their electorate in the upcoming election.

We can also make sense of these results by considering the relationship and interactions between political processes and financial markets during and in response to natural disasters, terrorism, and armed conflict, as these seem to elicit or influence similar contagion and volatility effects onto financial markets (Lee et al., 2018; Kutan and Yaya, 2016; Schneider and Troeger, 2006). In response to such events, governments often act quickly and decisively to preserve the lives of society and restore the former economic activity and social order. However, as Snowden (2019) stresses in his historical overview of pandemics – “major epidemics caught authorities unprepared, leading to confusion, chaos, and improvision” (p. 77) can lead to uncertainty. These events (and political actions) affect investor mentality and decision-making processes, market volatility and persistence of it, and increase uncertainty more generally, stunting economic growth (Baker et al., 2020a,b). Such effects also ripple through global society but are generally limited by some degree of geographical distance to the crisis in question (Ichev and Marinč, 2018) and are heavily influenced by news, media, and reporting (Audrino et al., 2020). Snowden (2019) also points out that pandemic regulations have historically cast a long shadow in political history. They often lead to an extension of state power into the spheres of human life and the economy, which we also observe during COVID-19 (see Eichenberger et al., 2020; Zweifel, 2020; Bickley et al., 2021). Increase in health risks causes citizens to increase their marginal willingness to pay for the control of risk increasing therefore the GDP share under politicians’ power (Zweifel, 2020). Evidence, for example, indicates that regions with higher confidence in the health care systems are more responsive to government interventions that request mobility reduction. Even in regions with low levels of confidence in the health care systems, a strong increase in staying at home is found when there is a high level of trust in the government (Chan et al., 2020a,b).

An interesting question to ask is which factor, e.g., government intervention, media reporting, or the crises event itself, has the strongest (or longest-lasting) influence on market volatility. To see more clearly the relation and feedback between truly exogenous events and the endogenous mechanisms and structures which respond and adapt to them. When making generalisations however, it is important to consider the context of the country (social, political, economic) and the potential influence this may has on market activity. In addition, whether the dynamic between political processes and market activity during and in response to other crises such as natural disasters, acts of terrorism, armed conflict, and social unrest, is similar or dissimilar, and what may be the lasting effects of such political interventions in the long run. For example, van Barneveld et al. (2020) suggest that the COVID-19 pandemic has brought to surface many pre-existing and emerging inequities, inequalities, and social norms and contend that “there can be no return to the ‘old normal’” (p. 133). However, what will be the lasting effects of the coronavirus pandemic and pandemic policies introduced is subject to a high level of uncertainty.

5. Conclusions

In this study, we explored financial market volatility and structural breaks present in response to various stages of COVID-19 stay-at-home policies and social distancing measures. The financial market indices of 28 countries across 6 continents are shown to correlate with country-specific COVID-19 variables such as population mobility, outbreak severity and stay-at-home policy interventions. In general, markets tend to have reacted sharply during the early stages of COVID-19 transmission (February/March) and become more volatile in the following months. In more than half of the observed cases, the establishment of stay-at-home policies appear to have elicited a strong and somewhat stabilising response.

For investors, the evidence of long-range dependency indicates increased predictability in an otherwise unpredictable and volatile market. For policy makers, controls and measures can be established and their efficacy in stabilising economic markets can be empirically determined by employing methods of multifractal analysis used in this paper. Further, policy makers can find solace that in more than half of the observed cases, social distancing measures appear to elicit a strong and stabilising response in the global financial markets. Thus, our results suggest there may be potential to find synergies between public health and economic policy – one just needs to know where and what to look for and with what tools to monitor and track.

The “opening up” of societies and economies from COVID-19 is a promising avenue for future research and for an extension of this current study. In doing so, we can observe whether the lifting or establishment of pandemic policies has a greater stabilising effect on financial markets and whether these demonstrate irreversibility or permanent effects on market multifractality, turbulence, and efficiency. We can also look to identify common mechanisms across the different stages of pandemic policies (i.e., when first established and when abolished and if applicable, when re-introduced and re-abolished) which catalyse volatility, turbulence, and inefficiency in the global financial markets. Future work could also explore the implications of other pandemic policies (aside from stay-at-home requirements) such as those with health (e.g., public information campaigns, testing/contact tracing policies), or economic focus (e.g., social supports, fiscal measures, debt/contract relief) and/or control for mobility along with the other google mobility functional area classifications/dimensions (e.g., Transit Stations, Workplaces).

Ethics approval and consent to participate

Not Applicable.

Consent for publication

Not Applicable.

Availability of data and materials

Data and materials used in the study are available on Open Science Framework, DOI 10.17605/OSF.IO/AEFJ9.

Funding

This research is/was partly funded by an Australian Government Research Training Program (RTP) Scholarship.

CRediT authorship contribution statement

Steve J. Bickley: Conceptualization, Methodology, Software, Formal analysis, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration. Martin Brumpton: Writing - original draft. Ho Fai Chan: Conceptualization, Methodology, Software, Formal analysis, Data curation, Visualization, Supervision. Richard Colthurst: Investigation, Data curation, Writing - original draft. Benno Torgler: Conceptualization, Methodology, Writing - review & editing, Supervision.

Declaration of Competing Interest

The authors report no declarations of interest.

Appendix A. Supplementary data

The following are Supplementary data to this article: