key: cord-0782581-eaagu99i
authors: Aslam, Faheem; Mohmand, Yasir Tariq; Ferreira, Paulo; Memon, Bilal Ahmed; Khan, Maaz; Khan, Mrestyal
title: Network Analysis of Global Stock Markets at the beginning of the Coronavirus Disease (Covid-19) Outbreak
date: 2020-09-16
journal: nan
DOI: 10.1016/j.bir.2020.09.003
sha: bc5196b36d8067afd10c3c6d91408eff5504ad00
doc_id: 782581
cord_uid: eaagu99i

The Coronavirus (COVID-19) outbreak has become one of the biggest threats to the global economy and financial markets. This study aims to analyze the effects of COVID-19 on 56 global stock indices from October 15, 2019 to August 7, 2020 by using a complex network method. Furthermore, the change of the network structure is analyzed in depth by dividing the stock markets into developed, emerging and frontier markets. The findings reveal a structural change in the form of node changes, reduced connectivity and significant differences in the topological characteristics of the network, due to COVID-19. A contagion effect is also identified in the network structure of emerging markets, with the nodes behaving synchronously. The findings also reveal substantial clustering and homogeneity in the world stock market network, based on geographic positioning. Besides, the number of positive correlations in the global stock indices increased during the outbreak. The stock markets of France and Germany seem to be the most relevant developed markets, while Taiwan and Slovenia have this relevance in emerging and frontier markets. The findings of this study help regulators and practitioners to design important strategies in the light of varying stock market dynamics during COVID-19.

According to the World Health Organization (WHO), by 9 th August 2020, COVID-19 had led to more than 19,824,060 confirmed infections and 729,910 deaths in 215 countries, and numbers continue to increase. 1 In the absence of preventative measures, this outbreak would probably infect 7.0 billion people worldwide, causing 40 million deaths (Walker et al., 2020) . Besides the immediate tragedies of death and disease, indirect effects through fear are taking hold around the world. The fear associated with the number of deaths reported has fostered a sense of emergency and globally panic is spreading faster than the spread of the virus itself (Aslam, Awan, Syed, Kashif & Parveen, 2020) . The virus outbreak is becoming the most defining economic and social event in human history with far-reaching economic implications (Laing, 2020) . represents unprecedented threats to financial stability and reduced economic activity worldwide (Boot et al., 2020) . According to UN International Labor Organization estimates, nearly 25 million jobs could be lost due to COVID-19. Globally, dramatic economic impacts are noted, for example, the possible loss of $2.7 trillion in output, as referred to by Orlik or Bloomberg, the Asian Development Bank stating damages of $4.1 trillion, and an OECD report declaring global economic growth being cut by half. 2,3 Furthermore, a pandemic like COVID-19 remains unknown and presents unprecedented uncertainty, which makes it difficult for governments to formulate an appropriate economic policy response (McKibbin & Vines, 2020) . Global stock markets reacted to the rapid emergence of COVID-19. For instance, the Dow Jones Industrial Average dropped by 2,353 points on March 12, 2020. Within a week, the DJIA fell by almost 3,000 points, the biggest one-day plunge since the "Black Monday" crash of 1987. In just one month, the UK-FTSE fell by 29.72%, Germany's DAX by 33.37%, France's CAC by 33.63%, Japan's NIKKEI by 26.85% and the Indian SUNSEX by 17.74%.

The world has already faced other crises in the form of diseases, i.e., Severe Acute Respiratory Syndrome (SARS), Middle East Respiratory Syndrome (MERS) or Ebola, among others, with the respective negative financial impacts. There is evidence that the SARS virus in 2003 not only affected human health but also damaged business and educational activities in China (Lee, 2003) . The work of Nippani and Washer (2004) examined the impact of SARS on several affected stock markets including China, Canada, the Philippines, Thailand, Singapore, Vietnam, Indonesia and Hong Kong and found an impact on those countries except for China and Vietnam. The Ebola outbreak affected investors' decisions, harming equity capital in African stock markets (Del Giudice & Paltrinieri, 2017) and also had a negative influence on the US stock market (Ichev & Marinč, 2018) .

Different aspects of COVID-19 have been widely observed and commented on by governments, researchers and the public alike. Particularly, the global financial market reacted very strongly to this immense black swan event (Nicola et al., 2020) . Baker et al. (2020) confirmed that COVID-19 is a more severe event than previous infectious disease outbreaks for US stock markets. As a consequence of COVID-19, stock indices fell drastically (McKibbin & Vines, 2020) and stock market volatility increased during the pandemic (Ali, Alam & Rizvi, 2020; Barro, Ursua & Weng, 2020) , causing huge investment losses (Zhang, Hu & Ji, 2020) . Aslam, Mohti and Ferreira (2020) confirmed a decline in the intraday efficiency of European stock markets during this pandemic. A comparative study documented that COVID-19 has had more effect on the US stock market than on Asian and Australian stock markets (Ammy-Driss & Garcin, 2020).

Somewhat similar findings are confirmed by Garcin, Klein and Laaribi (2020) . On the other J o u r n a l P r e -p r o o f hand, Topcu and Gulan (2020) find a greater impact on Asian stock markets than on European ones. From an international perspective, Ashraf (2020) confirms that stock markets react more to the number of confirmed cases than the number of deaths. Czech, Wielechowski, Kotyza, Benešová and Laputková (2020) applied a TGARCH model to analyze the short-term impact of COVID-19 on Visegrad countries' financial markets. The authors also confirm a negative relationship between Visegrad stock market indices, and the spread of COVID-19. In a similar study, Ali et al. (2020) applied an EGARCH model to the stock markets of the 9 countries most affected by COVID-19, concluding that stock markets deteriorated when the disease changed from an epidemic to a pandemic. Commodity markets also suffered when the pandemic moved into the US.

Using network analysis, Zhang et al. (2020) analyze the impact of COVID-19 on the stock markets of the ten countries with most COVID-19 cases. The authors report that European stock markets remain connected during this pandemic and the USA stock market failed to take the leading role before and during the outbreak. Although a few studies have compared the financial impacts of COVID-19 on different regional stock markets by applying diverse statistical techniques, no comprehensive study has addressed the impacts of COVID-19 on stock market networks. From an investment and regulatory point of view the main advantage of analyzing a large sample of global stock markets is that equity valuations should include both local and global risk factors.

The COVID-19 pandemic is creating fear of increasing global poverty levels due to lower consumer spending resulting in lower firm confidence (Lucas, 2020; Sumner, Hoy & Ortiz-Juarez, 2020) . The lockdowns and negative sentiments contributed to an increase in market illiquidity and volatility resulting in the deterioration of market stability (Baig, Butt, Haroon & J o u r n a l P r e -p r o o f Chen, Liu & Zhao, 2020) . Besides this, the news regarding COVID-19 created panic sentiments which affected investor behavior and generated uncertainty and high volatility in financial markets . Particularly, during bearish trends in stock markets, investors become more cautious (Lu & Lai, 2012; Omay & Iren, 2019) . Similar trading behavior is reported by Allam, Abdelrhim and Mohamed (2020) , emerging during the COVID-19 outbreak. With the backdrop of ambiguity and uncertainty, investors search for safe havens to avoid possible financial losses and are reluctant to trade, which affects financial markets adversely (Epstein & Wang, 1994; Mukerji & Tallon, 2001; Levy & Galili, 2006) . Moreover, the sentiments related to COVID-19 news affected the price movements of the financial market (Mamaysky, 2020) and investors' sentiments on future investments also affected the stock market during COVID-19 (Liu, Manzoor, Wang, Zhang & Manzoor, 2020) .

This study is unique and makes three main contributions to the literature. First, the study is more comprehensive. We apply a complex network approach to analyze the effects of the COVID-19 outbreak on 56 global stock indices. Second, the findings of this study are more detailed.

International institutional investors have different evolving investment requirements and use different types of instruments. To address this requirement, we performed an in-depth analysis by examining the effect of COVID-19 using developed, emerging and frontier markets. Third, this study constructs a complex network of all stock correlations and then analyzes in depth the variations in the network structure before and during COVID-19. The findings of this study will be helpful to protect stock markets by revealing the impacts of COVID-19 on global stock markets.

J o u r n a l P r e -p r o o f

In the space of a few weeks, Coronavirus (COVID-19) was officially declared a pandemic and shaved off nearly a third of the global market capitalization. 4 In order to analyze the influence of this disease, this study uses daily stock index prices of 56 global stock market indices. The data is collected from Yahoo Finance from 15 th October 2019 to 7 th August 2020, a total of 204 observations. Stock markets are divided into developed (23 markets), emerging (22 markets) and frontier (11 markets) markets using the Morgan Stanley Capital International (MSCI) classification. The list of countries, stock market index and classification are presented in Table   1 .

According to the emergence of COVID-19 and following Zhu et al. (2020) , the stock data is divided into two periods of 102 trading days each. Due to the global consequence, the World Health Organization (WHO) declared the COVID-19 epidemic as a global pandemic on March 11, 2020 (Maier & Brockmann, 2020) . Using the same date, the closing prices ranging from 15 th October 2019 to 10 th March 2020 refer to the period before COVID-19 and prices from 11 th March 2020 to 7 th August for the period during COVID-19. J o u r n a l P r e -p r o o f

To build the network, we started by defining the traditional log returns:

where the closing price of the index at moments t and t-1 are represented respectively by and − 1 .

To measure the dependence between stock market indices, the long-run correlation coefficient is estimated based on the heteroskedasticity and autocorrelation consistent variance-covariance matrix introduced by Andrews (1991) . This choice was guided by the fact that return series are subject to mild levels of autocorrelation and to excess volatility during crises (Výrost, Lyócsa, & Baumöhl, 2019) .

For a given sample size T, Andrews' (1991) estimate takes the following form:

where,

and where t=1,2,...,T, Z t =[r i,t ,r j,t ] T , and k(.) is the quadratic spectral kernel weighting function that together with band width parameter B weighs lagged variances and co-variances. In our empirical work, we choose automatic choice for the band width parameter which is 4, corresponding to 4 working days that attain the greatest weight. The quadratic spectral kernel function is defined as: 

Finally, the long-run correlation< = , is estimated as:

The correlations for the two different periods are represented in Figure It is important to note that COVID-19 has varying impacts on developed, emerging and frontier markets. Overall, the largest number of changes in correlations is found among the frontier markets during COVID-19, rather than in developed and emerging markets. To calculate the weight in the network, we calculate the distance between all pairs of correlation coefficients, as proposed by Mantegna, (1999) and Stanley and Mantegna (2000) . The transformation function for the distance can be expressed as:

Based on Equation (4), we obtain different distance matrices with dimensions of 56×56

(considering all markets), 23×23 (for developed markets), 22×22 (for emerging markets) and

11×11 (for frontier markets).

Due to the mutual flow of information and synchronization among investors, stock returns show high cross-dependence, even across countries. Since the seminal study by Mantegna (Mantegna, 1999) , devoted to correlation-based networks, it has been observed that the structures of such networks contain significant economic information and also important independent information A spanning tree is a sub-graph which contains all the nodes of the network but with fewer edges.

Different algorithms exist for the extraction of MST from a network, with the Kruskal and Prim algorithm (Kruskal, 1956) being the most popular. In this paper, the stock markets' MSTs are created using the Kruskal algorithm for an undirected graph (G=N,E,W) in forming MST (Kruskal, 1956) . Following this, communities are identified using the Girvan-Newman algorithm (Girvan & Newman, 2002) .

Several topological properties are used to calculate the most important node in a network. The most common include degree centrality, closeness centrality, betweenness centrality and eigenvector centrality. In this study, betweenness and closeness centralities are used.

Betweenness detects the nodes acting as bridges for the flow of information from one end to another. It can be expressed mathematically as

Where V being the node set, λ(a, b) represents the number of shortest paths and λ(a, b|i) is the shortest path passing through i. On the other hand, closeness is the average shortest distance from node 6 to all the other nodes, which reflects the importance of the node relative to other nodes in the network. Mathematically:

is the shortest distance between V i and V j and is equal to the minimum stations from V i to V j in the network, whereas (N-1) is the normalization factor. The average shortest path length is also studied, characterized by the minimum number of edges passing through one node to another, defined as:

Where V is the set of nodes in G, d(s,t) is the shortest path from s to t and n is the number of nodes in MST.

With Apart from these two nodes, the rest of the tree in the MST has a decreased level of connectivity.

Thus, the MST has a star-like structure, which is commonly found during crisis periods . Before COVID-19, Germany and France were the most connected markets.

However, Germany remains in the leading nodes during COVID-19, but Taiwan replaces France J o u r n a l P r e -p r o o f as the most connected market. The number of communities also increased from 7 to 8. Looking at the Asian markets, Taiwan emerges as the central hub followed by South Korea. Before, these two markets were part of same community, but after the pandemic, the community changes.

Taiwan is no longer connected to the Asian markets.

J o u r n a l P r e -p r o o f For a detailed insight, we also created separate MSTs with communities in Figures 6, 7 and 8 .

The MSTs before and during COVID-19 of developed markets are presented in Figure 6 and show a reduction in the number of communities from five before COVID-19 to four during COVID-19. Furthermore, the community changed significantly.

Before COVID-19, most connected nodes remain with European stock market indices dominating the developed countries market with high degree of connections. France holds a key central position in the MST of developed countries stock market with eight degrees of connection, followed by Germany and the Netherlands with four connections each. In addition, the shortest distance is observed among European stock markets, representing the highest correlation among developed stock markets before COVID-19. However, after the COVID-19 outbreak the connections of major hub nodes of France and Germany drop, which may be due to turbulence in the financial markets. Before COVID-19, Spain was connected to Sweden and

Finland, but during COVID-19 it is connected to Belgium only. The topological properties of the MSTs are presented in Table 2 , identifying several measures for the MST of the network for the whole set of indices (AllMST), for developed indices (DevMST), for emerging indices (EmMST) and for frontier indices (FrMST). The identification of "19" refers to the period before COVID-19 and "20" to the period during COVID-19. As shown in Table 2 , the average shortest paths of all markets during the pandemic have increased when compared to before. This means that before COVID, markets were closely connected.

Similarly, betweenness and closeness centralities of the minimum spanning tree of all 56 stock markets during COVID-19 have decreased compared to the minimum spanning trees of the period before COVID-19. The degree distributions of all the markets hold true to power law distribution given by p(k) ~ k -β (except frontier), meaning a few nodes having highest degree centrality (a few markets such as Germany, France, Poland and South Korea are highly connected, whereas others such as China, Hong Kong and Pakistan are less so).

One of our interesting findings was that Germany remained in the leading nodes during COVID-19 and Taiwan became one of the most connected markets, which could be related to some measures adopted in these countries. For instance, in Germany, the first COVID-19 death was reported on 9 th March 2020. Thereafter, Germany imposed very strict actions to control the spread which included closing its borders. By 22 nd March, curfews were imposed in some German states, while others prohibited physical contact with more than one person from outside households. By the first week of April, the number of confirmed cases started to decline in Germany. Additionally, the German parliament introduced economic stabilization funds with the main purpose of supporting the real economy. Regarding Taiwan, this is a country with a very small number of COVID-19 cases, just 481 confirmed cases and only 7 deaths. Taiwan has been admired for its relatively successful control of the first wave of the epidemic. For example, the country started COVID-19 RT-PCR tests at the start of the outbreak and tried to control the emerging virus through science, technology and democratic governance. Several measures were adopted to preserve economic and financial market stability. This could be seen as a sign of confidence by investors, making Taiwan the center of the emerging markets during COVID-19.

In agreement with Zhang et al. (2020) , the findings confirm that the US index (DOW 30) has failed to lead global stock markets before and during the COVID-19 period.

J o u r n a l P r e -p r o o f 27 J o u r n a l P r e -p r o o f

Coronavirus (COVID-19) has had a devastating effect on world economies, with many of them going into lockdown in order to contain the spread of the virus. Besides the economic impacts, it has also had a significant impact on financial markets. The purpose of this study is to provide a perspective of global stock markets during the turbulence caused by this new disease.

This study compares the network properties of 56 global stock markets before and during the coronavirus (COVID-19) outbreak by applying a complex network approach. To address institutional investors and regulators' requirements, we performed an in-depth analysis by examining the effect of COVID-19 on developed, emerging and frontier markets.

Stock network MSTs were created using the Kruskal Algorithm from a long run correlations distance matrix. The findings show that COVID-19 had a significant impact on financial networks with a structural change in the form of node changes and reduced connectivity, along with significant differences in the topological characteristics. Furthermore, the impact of COVID-19 varies with respect to the level of stock market development. In the case of emerging stock markets, a contagion effect is also identified in the network structure, with the nodes behaving synchronously. Based on geographic positioning, substantial clustering and homogeneity is found in the world stock market network. Besides, COVID-19 changes the sign and intensity of the correlation structure among global stock markets. The community structure reveals that the stock markets of France and Germany seem to be crucial for developed, Taiwan for emerging and Slovenia for frontier markets.

Although developed markets remain positively correlated before and during COVID-19, the strength of the relationship declines during COVID-19. For less mature markets, such as emerging and frontier ones, their returns are naturally lower. Normally, this could also be J o u r n a l P r e -p r o o f considered as a possibility for diversification, but our results show that for these indices in many cases the correlations became positive. This means these markets are now positively correlated with other markets, which was not the case in the past, implying less possibility of diversification and increased risk for investors. During COVID-19, market risk has increased substantially due to the great uncertainty of the pandemic. This market risk is raised by economic losses and is highly volatile during COVID-19 .

The main limitation of this study is the relative unavailability of data after COVID-19, as it is still spreading. Despite this limitation, the study has important theoretical and managerial implications. For example, policies and regulations depend on better understanding of the topological network structures of financial markets (Tang, Xiong, Jia, & Zhang, 2018) .

In the light of these findings, this study suggests integrated policy-making to cope with the financial impacts of the COVID-19 outbreak. Particularly, the regulators of stock markets in the same community should design combined policies to improve stock market stability. Individual and institutional investors can design their portfolios and risk management strategies in the light of these findings. The results suggest that diversification opportunities could be lower now, which should be considered by investors. On the other hand, using the information about stock market communities, the investor can design pair trading strategies by considering the movement of correlated indices (Elliott, Van Der Hoek, & Malcolm, 2005; Gatev, Goetzmann, & Rouwenhorst, 2006) . Furthermore, optimal portfolio optimization could be achieved using the topology of networks (Tang et al., 2018) . In order to address the limitations of this paper, future research could use a longer sample of the COVID-19 period, and obviously, information after COVID-19, to investigate how the networks behave after the end of the pandemic.

J o u r n a l P r e -p r o o f

Coronavirus (COVID-19)-An epidemic or pandemic for financial markets

The Effect of the COVID-19 Spread on Investor Trading Behavior On The Egyptian Stock Exchange

Efficiency of the financial markets during the COVID-19 crisis: time-varying parameters of fractional stable dynamics

Heteroskedasticity and autocorrelation consistent covariance matrix estimation

Stock markets' reaction to COVID-19: cases or fatalities?

Sentiments and Emotions Evoked by News Headlines of Coronavirus Disease (COVID-19) Outbreak

Evidence of Intraday Multifractality in European Stock Markets during the recent Coronavirus (COVID-19) Outbreak

Deaths, panic, lockdowns and US equity markets: The case of COVID-19 pandemic

The unprecedented stock market impact of COVID-19

The coronavirus and the Great Influenza epidemic: Lessons from the

Coronavirus and Financial Stability 3.0: Try Equity: Vox

Fear sentiment, uncertainty, and bitcoin price dynamics: The case of COVID-19

The evolution of interdependence in world equity markets-Evidence from minimum spanning trees

Shaking Stability: COVID-19 Impact on the Visegrad Group Countries' Financial Markets

The impact of the Arab Spring and the Ebola outbreak on African equity mutual fund investor decisions

Pairs trading

Intertemporal asset pricing under Knightian uncertainty Uncertainty

Estimation of time-varying kernel densities and chronology of the impact of COVID-19 on financial markets

Pairs trading: Performance of a relative-value arbitrage rule

Community structure in social and biological networks

Network analysis of the Chinese stock market during the turbulence of 2015-2016 using log-returns, volumes and mutual information

COVID-19: Media coverage and financial markets behavior-A sectoral inquiry

Stock prices and geographic proximity of information: Evidence from the Ebola outbreak

On the shortest spanning subtree of a graph and the traveling salesman problem

The economic impact of the Coronavirus 2019 (Covid-2019): Implications for the mining industry

State and Network Structures of Stock Markets Around the Global Financial Crisis

The SARS epidemic in Hong Kong

Terror and trade of individual investors

The COVID-19 outbreak and affected countries stock markets response

Relationship between stock indices and investors' sentiment index in Chinese financial market

Impacts of Covid-19 on Inclusive Economic Growth in Middle-income Countries

A Network Analysis of Shariah-Compliant Stocks across Global Financial Crisis: A Case of Malaysia

Effective containment explains subexponential growth in recent confirmed COVID-19 cases in China

Topology of the South African stock market network across the 2008 financial crisis

Hierarchical structure in financial markets

Financial Markets and News about the Coronavirus

Global macroeconomic cooperation in response to the COVID-19 pandemic: a roadmap for the G20 and the IMF. Forthcoming in 'The Economics of the COVID-19 Pandemic

Structural change and dynamics of Pakistan stock market during crisis: A complex network perspective

Network analysis of Pakistan stock market during the turbulence of economic crisis

Ambiguity aversion and incompleteness of financial markets

Dynamic topology and allometric scaling behavior on the Vietnamese stock market

The socio-economic implications of the coronavirus pandemic (COVID-19): A review

SARS: a non-event for affected countries' stock markets?

Behavior of foreign investors in the Malaysian stock market in times of crisis: A nonlinear approach

Minimum spanning tree application in the currency market

An introduction to econophysics

Estimates of the Impact of COVID-19 on Global Poverty

Complexities in financial network topological dynamics: Modeling of emerging and developed stock markets

The impact of COVID-19 on emerging stock markets

Network-based asset allocation strategies. The North Am

Report 12: The global impact of COVID-19 and strategies for mitigation and suppression

Correlation Structure and Evolution of World Stock Markets: Evidence from Pearson and Partial Correlation-Based Networks

Analysis of linkage effects among industry sectors in China's stock market before and after the financial crisis

Financial markets under the global pandemic of COVID-19

A novel coronavirus from patients with pneumonia in China