Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/de-gruyter/world-capital-markets-facing-the-first-wave-of-covid-19-traditional-3z3LQxic00
References
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
- Publisher
- de Gruyter
- Copyright
- © 2022 Pedro Luis Angosto-Fernández et al., published by Sciendo
- eISSN
- 2450-0097
- DOI
- 10.18559/ebr.2022.4.2
- Publisher site
- See Article on Publisher Site
Abstract
Economics and Business Review, Vol. 8 (22), No. 4, 2022: 5-38 DOI: 10.18559/ebr.2022.4.2 World capital markets facing the first wave of COVID-19: Traditional event study versus sensitivity to new cases 2 3 Pedro Luis Angosto-Fernández , Victoria Ferrández-Serrano Abstract: e a Th im of the paper is to analyse the impact of the new coronavirus on fi - nancial markets. The sample comprises returns from 80 countries, across all regions and incomes for the period known as the first wave. By combining event study meth - odology and time series analysis of new COVID-19 cases it is found that the negative price ee ff ct is widespread but unequal across regions. It is also noted that the distribu - tion of the impact is also uneven with a high concentration in the week ae ft r the first local case but especially in the weeks around the pandemic declaration. Finally, it has been shown at die ff rent levels how the markets most ae ff cted by the crisis are not nec - essarily the most sensitive to the virus. Keywords: financial markets, event study, COVID-19, coronavirus, stock returns. JEL codes: G01, G14, G15, F65, C32. Introduction On 31 December 2019 China reported the r fi st case of the new coronavirus and since then the world has experienced an unprecedented situation. It is nei- ther the r fi st nor the worst pandemic sue ff red by humanity, but it is the most important one to have existed in the last century. Above all this pandemic is different because it has occurred in a highly globalised and interdependent world economy. As a result, not only has the virus spread rapidly, but the meas- ures taken to contain it and the respective consequences have also turned this health crisis into a political and economic one. During the period covered, from 31 December to 1 June 2020, the virus rapidly infected equity markets causing cumulative declines of more than a quarter of total capitalisation in Austria, Article received 6 July 2022, accepted 21 September 2022. Casa del Paso Building. Universidad Miguel Hernández de Elche. Plaza de Las Salesas, s/n. Zip Code: 03300, Orihuela, Spain, corresponding author: pangosto@umh.es, https://orcid. org/0000-0001-6960-074X. La Galia Building. Universidad Miguel Hernández de Elche. Avenida de la Universidad s/n. Zip Code: 3202, Elche, Spain, v.ferrandez@umh.es, https://orcid.org/0000-0003-2978-9765. 6 Economics and Business Review, Vol. 8 (22), No. 4, 2022 Brazil, Egypt, or Indonesia, among others, and causing daily falls in prices that were higher than those during the global financial crisis. Academic studies on COVID-19 and capital markets have been published continuously since the mid-2020s (Ashraf, 2020, 2021; Baker et al., 2020; Gormsen & Koijen, 2020; Spatt, 2020; Ramelli & Wagner, 2020; Rizwan, Ahmad, & Ashraf, 2020; Zaremba, Kizys, Aharon, & Demir, 2020). They report the statis - tically negative ee ff ct on asset returns and positive ee ff ct on volatility, examine the ee ff ct of government measures with controversial results and try to explain the die ff rent levels of risk exposure at country and firm level. i Th s paper contributes to this recent literature in several ways. A global study of the short- and medium-term ee ff cts of the first wave of the pandemic on equity markets with a sample of 80 countries divided into eight regions is pre- sented. This experiment is based on two approaches that have been the order of the day in finance research: event study and analysis of daily case time series and their inu fl ence on markets. Both approaches are treated by regressing the time series of index returns under a system of simultaneous equations called seemingly unrelated equations (Zellner, 1962; Karafiath, 1988) and using an extended market model and the 3-factor model by Fama and French (1993). e f Th ormer is divided into two distinct events: from the day each country detected its first infection and from the day the WHO declared COVID-19 a pandemic. This makes it possible to assess the significance of these events and their evolution over time. The latter evaluates the sensitivity of investors in each country to the information provided by the health authorities as well as being an experiment to assess the ee ff ct of information about the pandemic on each country and the eci ffi ency of markets in general. The comparison of both methodologies and the level of disaggregation provided makes it possi- ble to present a very detailed and comprehensive study of the first months of the pandemic. From the results the overall signic fi ant negative ee ff ct on equity markets and its concentration around the days when the pandemic was declared are high- lighted. Especially in the regions of Europe, Eastern Europe and South America and the Caribbean. The inverse relationship between case growth and index returns is also proven which is signic fi ant in 56 out of 80 markets. Notably, the comparison of the two experiments shows an avenue for future research, namely that the countries with the lowest cumulative abnormal returns are not the countries most ae ff cted by a growth in cases. e p Th aper continues with Section 1, a review of the literature where a link between this event and the ee ff ct of natural disasters and unexpected events in general on equity markets is established. In the same section there is a discus- sion of the main findings of the emerging literature on COVID-19 and stock markets. Subsequently, the methodology of the two experiments is presented in detail in Section 2, the first being a classical event study approach and the second the application of a time series model where daily returns are related to P. L. Angosto-Fernández, V. Ferrández-Serrano, World capital markets facing the first wave 7 the growth of the cases. Despite the notable die ff rences both experiments are conducted through the use of simultaneous equation models. Ae ft r that, the main results of the research are presented in Section 3, with special emphasis on the comparison of the two methods and finally in Section 4, some conclu - sions and their policy implications are highlighted. 1. Literature review and research questions Unanticipated events ae ff ct stock markets and research in this area has been prolic i fi n recent years. The level of uncertainty may ae ff ct both future divi - dends (negatively) and the expected rate of returns (positively) at least until the contingency is resolved and uncertainty disappears (Brown, Harlow, & Tinic, 1988). There is already evidence that the current crisis has ae ff cted equities through both channels (Gormsen & Koijen, 2020). One of the most widely recognised contributions in this field is the one by Baker, Bloom and Davies (2016) who developed an index capturing moments of high economic policy uncertainty. During the first wave of COVID-19 this index reached an all-time global peak in April (Economic Policy Uncertainty, n.d.). It is a perfect tool for directly studying the relationship between uncer- tainty and stock markets, but it is on a monthly basis and the sample of coun- tries is still reduced. Thanks to this index and other indicators, Baker and others (2020) found that the uncertainty generated by this pandemic is unprecedented. e Th y hypothesise that this reaction is caused by the restrictions implemented by the governments and the preventive behaviour of individuals themselves as this occurred once the virus was detected in each country and not before. In the event literature a division could be made between events that are more related to natural disasters and those that are more politically induced. Obviously, this line is blurred, and coronavirus is the best proof of this; it be- longs to the first category by den fi ition, but its duration and its ee ff ct on the measures taken by countries also make it a political event. In a theoretical and empirical work Barro (2006) developed a model which explains that despite the low probability of rare disasters (such as wars) they are able to explain the high equity premium during the twentieth century. With respect to natural disasters and their effects on stock markets Bourdeau -Brien and Kryzanowski (2017) found that only few events cause significant effects on returns and volatility in USA markets. They also discov - ered that the most adverse effects on the stock market are felt two to three months after the peak of media coverage. Valizadeh, Karali and Ferreira (2017) showed how a disaster, such as the Japan earthquake of 2011, not only affects the national stock market, but it also rapidly extends to related mar - kets and its negative impact partly remains in the long run. In the same vein Papakyriakou, Sakkas and Taoushianis (2019) found that countries which 8 Economics and Business Review, Vol. 8 (22), No. 4, 2022 experienced higher stock declines after terrorist attacks also experienced higher economic losses. More recently, on this connection to the real econo- my, Iheonu and Ichoku (2022) found that terrorism in Africa has a negative effect on domestic investment but even more so on FDI. As a final example Kaplanski and Levy (2010) found that the stock market reacts negatively to aircraft crashes with increases in volatility and decreases in returns. In addi - tion, they found that the market reaction, measured as capital loss can be as much as sixty times the actual economic loss. Special attention should be given to a paper published previous to the cur- rent pandemic by Donadelli, Kizys and Riedel (2017). They studied the phar - maceutical stock reactions to oci ffi al WHO announcements and found that in a r fi st stage there is a fall in prices caused by fear and over-information, but there is also a second stage of growth induced by government intervention and investment opportunities. They also report an abnormal and persistent growth in volatility. While these are interesting results the experiment only sampled pharmaceutical companies where extraordinary returns can be obtained due to potential vaccines or treatments. In the second group, articles analysing unexpected outcomes from elections (Goodell & Vähämaa, 2013; Wagner, Zeckhauser, & Ziegler, 2018), referendums (Angosto-Fernández & Ferrández-Serrano, 2020; Schiereck, Kiesel, & Kolaric, 2016) and other political events (He, Nielsson, & Wang, 2017; Liu, Shu, & Wei, 2017; Hillier & Loncan, 2019) are found. The literature regarding uncertain po - litical events presents key findings that can be extended to neighbouring disci - plines (Brooks, Patel, & Su, 2003). First, there is a negative relationship between uncertainty and returns (Angosto-Fernández & Ferrández-Serrano, 2020; He et al., 2017; Schiereck et al., 2016). Second, there is a positive relationship between uncertainty and volatility (Goodell & Vähämaa, 2013; Smales, 2016; Chiang, 2019), and finally, there is a high dispersion on returns showing that the ef - fects of uncertainty are not homogeneous among firms or countries (Davies & Studnicka, 2018; Shahzad, Rubbaniy, Lensvelt, & Bhatti, 2019). Additionally, and not surprisingly, academic work on the influence of COVID-19 on the stock market has been booming for some months now (Ashraf, 2020, 2021; Ramelli & Wagner, 2020; Zhang, Hu, & Ji, 2020; Zaremba et al., 2020, among others). As in the literature on unanticipated events many researchers report abnormal negative returns (Ashraf, 2021; Heyden & Heyden, 2021; Pandey & Kumari, 2021; Ramelli & Wagner, 2020) and others report an unusual increase in volatility and market contagion (Baker et al., 2020; Contessi & De Pace, 2021; Li et al., 2022; Liu, Wei, Wang, & Liu, 2022; Samitas, Kampouris, & Polyzos, 2022; Zhang et al., 2020; Zaremba et al., 2020). In Liu and others (2022) they find that the cross-market contagion ee ff ct caused by the pandemic lasted between six and eight months, which is important in determining which model and methods to use to conduct any research on returns and/or volatility. Finally, in one of the most interesting papers as it will open the door to future P. L. Angosto-Fernández, V. Ferrández-Serrano, World capital markets facing the first wave 9 debates within the field, Uddin, Chowdhury, Anderson and Chaudhuri (2021) n fi d that the level of economic strength of the country helps to mitigate the ef - fects of COVID-19 on market volatility. Regarding this experiment there is previous evidence of the negative and significant influence of the growth in cases on stock prices worldwide (Ashraf, 2020a; Seven & Yilmaz, 2020; Pandey & Kumari, 2020). In this respect the pa- pers by Ashraf (2021) and Fernández-Peréz, Gilbert, Indriawan and Nguyen (2021) relate the sensitivity of caseload growth to a national cultural ee ff ct be - cause countries with a higher degree of risk aversion seem to be more ae ff cted by the increased incidence of the virus. These results are maintained ae ft r the introduction of control variables and other robustness checks. As well as other researchers O’Donnell, Shannon and Sheehan (2021) found a signic fi ant rela - tionship between the cases and the returns of six of the world’s major indices, but they also found that ae ft r controlling for some of the variables that most inu fl ence capital markets that two of these relationships were no longer signifi - cant (the one for the Chinese index and the one for the world market). More recently Alkhatib, Almahmood, Elayan and Abualigah (2022) conr fi med the negative relationship between the increase in COVID-19 cases and the stock market points of the GCC countries and they used the coeci ffi ents obtained from the time series models to determine which markets are most ae ff cted al - though, as will be seen throughout this article, using only this indicator may be limiting when determining the total ee ff ct. Finally, Yu, Xiao and Liu (2022) construct their own indices from information on new cases and deaths. It is a study with a longer time span which is probably why they find that this rela - tionship is volatile, and that it becomes very weak especially ae ft r the first an - nouncement of the vaccine. With regard to event studies, Narayan, Khan and Liu (2021) used daily dum- mies to control for lockdowns, government stimulus and border closures in G7 countries and found that lockdowns are the events that most severely af- fect stock markets. In some cases, there are mixed signs, but there is an overall positive ee ff ct in returns while the ee ff ct of stimulus is only positive and signifi - cant in three countries. Pandey and Kumari (2020) took a sample of forty-nine markets and conr fi med the evidence regarding lockdowns. They also present - ed additional evidence of the negative ee ff ct on returns from the declaration of a public health emergency (pre-pandemic) by the WHO (three and seven days later), with Asia being the most ae ff cted region. Interestingly, developed countries appear to have anticipated the declaration with signic fi ant abnormal returns prior to the event. Heyden and Heyden (2021) focus on four die ff rent events in the USA and EU countries: first case, first death, fiscal stimulus and monetary stimulus. They found negative abnormal returns for first death and for fiscal stimulus while monetary stimulus provided positive abnormal returns. i Th s result is contra - dicted by Seven and Yilmaz (2020) where fiscal stimulus is related to stock mar - 10 Economics and Business Review, Vol. 8 (22), No. 4, 2022 ket rallies while all other interventions have no signic fi ant ee ff ct. In the latter study, the sample comprises seventy-eight countries, so it seems that there are notable die ff rences in the ee ff ct of stimulus around the world which is an am - biguity also suggested by Narayan and others (2021). e Th present research seeks to complement the information provided by these investigations. Thus, the main objective of this research could be den fi ed as the quantic fi ation of the impact of the first wave of COVID-19 on global capital markets and the comparative analysis of two die ff rent methods to do so. To this end, a series of questions are proposed: – Are the accumulated losses in global capital markets signic fi ant and are they significant in all regions and countries? How significant are price declines aer ft discounting for expected asset returns? – How are markets ae ff cted by the evolution of the pandemic? In which weeks are the bulk of losses concentrated? – Are markets sensitive to new epidemiological information and are there geographical differences? – What information could be obtained from the event study methodology that is not obtainable from studying the time series of growth in cases and its inu fl ence on stock market indices? Some of these questions have been addressed in previous articles, but this research brings new elements to the debate. First, to answer these questions stock market data are collected from major indices from eighty countries for an event window from 31 December to 1 June. One of the longest samples and study periods to date. In addition, the sample selected includes countries such as Iraq, Ghana, Tanzania, Myanmar and Jamaica, whose markets are considered “underdeveloped” and are oe ft n excluded by default in other studies. Second, the event period is die ff rent for each country as it starts from the day the first case was detected. This permits testing for abnormal returns for those days and also observe the evolution of the pandemic over weeks thereby detecting where the bulk of the losses are globally and regionally. Additionally, another event study is carried out, starting from the week when the WHO de- clared COVID-19 a pandemic which allows an insight into the singularity of this unique political and economic event. iTh rd, the event study is based on a multivariate equation system and not on global indices or die ff rent panel study methods. This method provides an interesting level of disaggregation to observe what proportion of national eq- uity markets actually sue ff red signic fi ant ee ff cts. In the same vein, regional data at eight levels is presented: Africa, Asia, North America, South America and the Caribbean, Europe, Eastern Europe, MENA (Middle East & North Africa), and Oceania. Fourth, building on the research by Ashraf (2020), an additional experi- ment is incorporated to observe which investors at country and regional level P. L. Angosto-Fernández, V. Ferrández-Serrano, World capital markets facing the first wave 11 are more sensitive to growth in COVID-19 cases. The data is also comprehen - sively explained and directly comparable to the event study with the intention of comparing both methodologies and finding similarities and differences at all levels. This comparison also enriches the literature on the role of culture and its ee ff cts on the stock market as it is directly observable that the markets most ae ff cted by the pandemic are not necessarily those that react the most to an increase in the rate of infection and vice versa. 2. Data and event description Table 1 lists the different countries in the sample and their indices including other important details to better understand this research. The experiment is split into two parts: a traditional event study and an analysis of the sensitivity of returns to increases in cases. The objective is to answer a single question in two ways: How signic fi ant was the first wave of the pandemic with respect to the capital markets of the die ff rent countries in the world? To do so, the daily quotations of the stock market indices are collected (one per country) provided that there were data from at least one hundred sessions before 31 December 2019, the day when the r fi st case was detected. e Th n, they are used to compound logarithmic returns. The data was obtained from Investing (Investing, 2022), and by asking each stock exchange individually when the data was not on the website. This procedure gives a preliminary sam - ple of more than ninety countries, but ae ft r applying the requirement that no more than 25% of their returns should be 0, the sample was reduced to eighty countries, ten of them being traditionally Jewish or Muslim where the business week goes from Sunday to Thursday. e Th se details can be seen in Table 1. As is well known the first case occurred in China and the country that was the last to detect its first case was Myanmar (also known as Burma) on 24 March 2020. Counting from 31 December the country that experienced the greatest stock market decline was Cyprus with –41.12%, but Sri Lanka followed very closely behind. Conversely, Zimbabwe and Venezuela experienced a stock market growth higher than 100% mainly driven by hyperinflation. If it is determined that the outbreak occurred when the first national case appeared Sri Lanka is clearly the most damaged while the winners are exactly the same. Finally, considering the loss of capital per day, Colombia is the most ae ff cted with a daily fall of more than 0.52%. Logarithmic returns are used in both experiments. Additionally, the MSCI World Index is the benchmark index of the world and the SMB (Small Minus Big market capitalization) and HML (High Minus Low book-to-market ratio) are risk factors collected from the Kenneth French website (Kenneth R. French, 2022). Finally, the number of cases by country and date were collected from the European Union Open Data Portal (European Union Open Data Portal, 2020). [12] Table 1. Average and accumulated raw returns of the sample indices Average daily Average Special Accumulated Accumulated Country Index 1st case return (31st daily return Region week (31st DEC) (1st case) DEC) (1st case) Argentina S&P Merval No 03/03/2020 –7.0850 –0.0668 6.8440 0.1104 SA&C Australia S&P ASX 200 No 25/01/2020 –10.5114 –0.0992 –14.6230 –0.1662 O Austria ATX No 26/02/2020 –35.6028 –0.3359 –28.5620 –0.4328 E Bahrain BAX Yes 24/02/2020 –23.6265 –0.2229 –27.3110 –0.4016 M Bangladesh DESEX Yes 09/03/2020 –10.7476 –0.1014 –6.9510 –0.1198 AS Belgium BEL20 No 04/02/2020 –19.3868 –0.1829 –18.4266 –0.2247 E Brazil Ibovespa No 26/02/2020 –27.3580 –0.2581 –26.3786 –0.3997 SA&C Bulgaria SOFIX No 08/03/2020 –20.8840 –0.1970 –13.9217 –0.2400 EE Cambodia CSX No 28/01/2020 5.5072 0.0520 6.6239 0.0761 AS Canada S&P TSX No 26/01/2020 –9.7988 –0.0924 –12.4917 –0.1420 NA Chile S&P IPSA No 04/03/2020 –24.6290 –0.2323 –16.1310 –0.2644 SA&C China SZSE Component No 31/12/2019 9.1806 0.0866 9.1806 0.0866 AS Colombia COLCAP No 07/03/2020 –40.3364 –0.3805 –30.2446 –0.5215 SA&C Côte d’Ivore BRVM Composite No 12/03/2020 –15.3238 –0.1446 –7.6679 –0.1394 A Cyprus CYMAIN No 10/03/2020 –41.1177 –0.3879 –21.7709 –0.3819 E Czech Republic PX No 02/03/2020 –20.6349 –0.1947 –7.3773 –0.1171 EE Denmark OMX-C20 No 27/02/2020 9.2665 0.0874 4.8471 0.0746 E Egypt EGX 30 Yes 02/03/2020 –30.9497 –0.2920 –17.8220 –0.2829 M [13] Finland OMX-H25 No 30/01/2020 –5.4207 –0.0511 –7.6095 –0.0895 E France CAC 40 No 25/01/2020 –22.7960 –0.2151 –23.4963 –0.2670 E Germany DAX No 28/01/2020 –13.4052 –0.1265 –13.0707 –0.1502 E Ghana GSE Composite No 13/03/2020 –14.2610 –0.1345 –13.0362 –0.2414 A Greece ATG No 27/02/2020 –33.2232 –0.3134 –20.8950 –0.3215 E Hong Kong HSI No 23/01/2020 –17.8151 –0.1681 –17.8915 –0.1988 AS Hungary SE No 05/03/2020 –25.0380 –0.2362 –18.8737 –0.3146 EE Iceland ICEX Main No 29/02/2020 –3.6588 –0.0345 5.1202 0.0813 E India BSE Sensex No 30/01/2020 –18.2949 –0.1726 –17.2998 –0.2035 AS Indonesia IDX Composite No 02/03/2020 –27.6139 –0.2605 –13.1774 –0.2092 AS Iraq ISX 60 Yes 25/02/2020 –15.0737 –0.1422 –9.0080 –0.1344 M Ireland ISEQ Overall No 01/03/2020 –18.1012 –0.1708 –6.5153 –0.1034 E Israel TA125 Yes 24/02/2020 –15.1277 –0.1427 –15.8570 –0.2332 M Italy FTSE MIB No 31/01/2020 –23.8220 –0.2247 –24.9840 –0.2974 E Jamaica JSE All Index No 12/03/2020 –29.1257 –0.2748 –10.1079 –0.1838 SA&C Japan Nikkei 225 No 15/01/2020 –2.5162 –0.0237 –4.0621 –0.0423 AS Jordan SE All Share Yes 03/03/2020 –14.8670 –0.1403 –11.6800 –0.1884 M Kazakhstan KASE No 15/03/2020 –3.8191 –0.0360 9.2898 0.1753 AS Kenya NASI No 14/03/2020 –18.7256 –0.1767 2.5630 0.0484 A Malaysia KLCI No 25/01/2020 –7.0805 –0.0668 –4.3920 –0.0499 AS Malta MSE No 07/03/2020 –17.3032 –0.1632 –14.1092 –0.2433 E Mauritius Semdex No 20/03/2020 –30.5445 –0.2882 1.9864 0.0405 A [14] Mexico IPC No 29/02/2020 –16.7301 –0.1578 –11.2377 –0.1784 NA Morocco MASI No 03/03/2020 –21.4104 –0.2020 –21.3721 –0.3447 M Myanmar Myanpix No 24/03/2020 0.0545 0.0005 0.4657 0.0099 AS Namibia FTSE NSX Overall No 15/03/2020 –27.7432 –0.2617 2.1981 0.0415 A Netherlands AEX No 28/02/2020 –11.5882 –0.1093 –3.7184 –0.0581 E New Zealand NZSX 50 No 28/02/2020 –4.6577 –0.0439 –4.1803 –0.0653 O Nigeria NSE All Share No 28/02/2020 –4.9617 –0.0468 –5.7064 –0.0892 A Norway OBX No 27/02/2020 –16.2236 –0.1531 –10.8240 –0.1665 E Oman MSM 30 Yes 25/02/2020 –11.6506 –0.1099 –16.1940 –0.2417 M Pakistan Karachi All Share No 27/02/2020 –15.8590 –0.1496 –6.2798 –0.0966 AS Peru S&P Lima No 07/03/2020 –30.1284 –0.2842 –18.2753 –0.3151 SA&C Philippines PSEi No 30/01/2020 –29.4035 –0.2774 –24.7822 –0.2916 AS Poland WIG20 No 04/03/2020 –21.5356 –0.2032 –8.6248 –0.1414 EE Portugal PSI-20 No 03/03/2020 –16.8206 –0.1587 –8.4504 –0.1363 E Qatar QE General Yes 01/03/2020 –14.7193 –0.1389 –5.0840 –0.0794 M Romania BET No 27/02/2020 –15.7892 –0.1490 –13.0203 –0.2003 EE Russia RTS No 01/02/2020 –18.9560 –0.1788 –16.8783 –0.2034 EE Saudi Arabia TASI Yes 03/03/2020 –26.7662 –0.2525 –14.4220 –0.2326 M Serbia Belex 15 No 07/03/2020 –17.1791 –0.1621 –15.9967 –0.2758 EE Singapore FTSE Singapur No 24/01/2020 –21.8902 –0.2065 –22.8759 –0.2570 AS South Africa SWIX No 06/03/2020 –16.2334 –0.1531 –7.6838 –0.1302 A South Korea KOSPI No 20/01/2020 –5.6521 –0.0533 –8.0307 –0.0864 AS Th Th Th [15] Spain IBEX 35 No 01/02/2020 –28.6026 –0.2698 –26.0240 –0.3135 E Sri Lanka S&P Sri Lanka 20 No 28/01/2020 –41.0538 –0.3873 –37.0945 –0.4264 AS Sweden OMXS30 No 01/02/2020 –7.1625 –0.0676 –7.8044 –0.0940 E Switzerland SMI No 26/02/2020 –7.6860 –0.0725 –6.3736 –0.0966 E Taiwan Taiwan Weighted No 21/01/2020 –8.2292 –0.0776 –8.7698 –0.0953 AS Tanzania DSE ASI No 17/03/2020 –12.9610 –0.1223 –9.3440 –0.1797 A ailand SETI No 13/01/2020 –17.6959 –0.1669 –17.7459 –0.1811 AS Trinidad and Tobago TTSE Composite No 13/03/2020 –13.8106 –0.1303 –14.3146 –0.2651 SA&C Tunisia TUNINDEX No 03/03/2020 –8.9901 –0.0848 –10.0041 –0.1614 M Turkey BIST 100 No 12/03/2020 –7.2494 –0.0684 5.5510 0.1009 M Uganda Uganda All Share No 22/03/2020 –28.2914 –0.2669 –7.5805 –0.1579 A United Kingdom FTSE 100 No 31/01/2020 –18.3636 –0.1732 –15.6232 –0.1860 E United Arab Emirates ADX General Yes 27/01/2020 –19.9507 –0.1882 –22.6320 –0.2572 M Uruguay BVM No 15/03/2020 –2.2646 –0.0214 0.3362 0.0063 SA&C United States S&P 500 No 21/01/2020 –1.4147 –0.0133 –4.7223 –0.0513 NA Venezuela IBC No 15/03/2020 112.6701 1.0629 96.8292 1.8270 SA&C Vietnam VN No 24/01/2020 –10.0905 –0.0952 –12.7925 –0.1437 AS Zimbabwe Zimbabwe No 21/03/2020 171.3466 1.6165 96.6448 2.0134 A Industrial Notes: Returns are multiplied by 100. Special week means that trading days go from Sunday to ursday. e names of the indices appear as in Investing (Investing, 2022). A is Africa, AS is Asia, E is Europe, EE is Eastern Europe, M is MENA, NA is North America, SA&C is South America & Caribbean, and O is Oceania. Source: Own elaboration on the basis of Investing (Investing, 2022). 16 Economics and Business Review, Vol. 8 (22), No. 4, 2022 2.1. Classic event study e e Th vent period covers from 4 August 2019 until 1 June 2020. The return se - ries are estimated all together using a multivariate system called seemingly un- related regressions (Zellner, 1962; Karafiath, 1988). This methodology permits abnormal returns to be obtained in a single step with no die ff rence between estimation and event window and it considers contemporaneous dependence on disturbances by taking into consideration one of the main problems of clus- tered events: cross-sectional correlation. e Th refore, dummy variables are used to estimate these abnormal returns and each dummy coefficient corresponds to one week and its value is the daily average abnormal return of that week. i Th s was done in response to the specic l fi ength of this event to make the data easier to handle and interpret. Furthermore, as zero moment has been established in the week in which the first COVID-19 case was detected in each country the length of the event varies according to the country. As an example, China has twenty-two weeks of abnormal returns while Myanmar has only ten. It was carried out in this way with the intention of assessing the direct ee ff ct of the virus in each coun - try assuming that at national level investors would act as the virus permeated each particular region (Baker et al., 2020). This decision was based on previous papers and observation of the data because initially no one had given any im- portance to the information and the various control measures were only taken once the virus had permeated the country in question. e Th model used to describe the normal path of returns is an extended ver - sion of the market model. This extended version considers the autocorrelation of the returns for each country and a lag of the market variable. This modi - c fi ation has been made to describe the usual evolution of returns in the best possible way and ae ft r various tests the explanatory power of this model was much higher in the vast majority of countries than the traditional market mod- el. For each country: N=Y r = α + α *r + β *r + β *r + δ * D + ε (1) it i0 i1 it−1 i1 WORLDt i2 WORLDt−1 ∑ ij j it j=0 r is the logarithmic return of the index (country) i on day t; α is the constant it i0 of the model for the index i; r , r and r are the autocorrelation it–1 WORLDt WORLDt–1 of r , the logarithmic return of the world market index on day t and its lag, re- it spectively. α , β and β are their associated coeci ffi ents. δ is the average daily i1 i1 i2 ij abnormal return for index i over week j, D is a binary variable that takes the value of one in any of the days of week j of the event, and ε is the disturbance it term. The weeks of the event are den fi ed as Y since they take different values depending on the country. P. L. Angosto-Fernández, V. Ferrández-Serrano, World capital markets facing the first wave 17 e Th se coeci ffi ents ( δ ) are used to perform joint tests on the global signifi - ij cance of the returns and to analyse the evolution of the pandemic in markets worldwide from the start to the end of the first wave in each country. It also allows an estimation of the abnormal loss accumulated during this period and its statistical signic fi ance. Additionally, the traditional cross-sectional t-test for global and regional signic fi ance was performed. As a robustness check, the Fama and French 3-factor model (1993) is also used by adding the SMB and HML factors to verify whether these risk factors are capable of absorbing and explaining the ee ff ct of the pandemic on listed companies. For each country: N=Y (2) r = α + β *r + φ *SMB + φ * HML + δ * D + ε it i0 i1 WORLDt i1 t i2 t ∑ ij j it j=0 SMB and HML being the risk factors Small Minus Big and High Minus Low t t related to the premium associated with small and value companies respectively. φ and φ are the coeci ffi ents of each factor for each index i. i1 i2 2.2. Market sensitivity to new cases As the information on new daily cases is available almost worldwide another way of testing on financial markets arises. This is basically an analysis of whether the stock market indices were sensitive to the new information given by health authorities it being understood that an increase in the number of cases should have a negative impact on index returns. Carrying out this analysis serves as a robustness check of the event study as well as a study of market eci ffi ency in the face of new daily information. It also permits an estimate as to which coun- tries and regions were more sensitive to the pandemic with a single indicator. In summary, this is a mean model, but a coeci ffi ent is added for daily increases in cases. For each country: r = θ + γ * ∆Cases + ε (3) it i i it it r has been defined above; θ is the constant of the model or the average daily it i return when there is zero growth in cases; γ is the sensitivity of the index to Cases −Cases t t−1 growth in new cases; ∆Cases is the growth in cases: . Cases it Cases t−1 being the accumulated cases of a given country on day t or t–1. Once again, ε it is the disturbance term. This equation is also extended as a robustness check including first the extended market model and then likewise the 3-factor mod - el as explained in the previous section. This allows the identic fi ation of which nations and regions are still sensitive to an increase in cases ae ft r discounting all these risk factors. Th [18] Table 2. Returns by region and other descriptive statistics of indices Asym- n Mean SD Min Q1 Median Q3 Max Kurtosis metry Europe 4.009 –0.0551 1.8672 –18.5411 –0.5945 0.0438 0.7752 10.4143 14.0898 –1.8048 Eastern Europe 1.477 –0.0665 1.7733 –14.2456 –0.5118 0.0108 0.6035 8.8251 14.0214 –1.8984 South America & Caribbean 1.899 0.0399 2.7256 –47.6922 –0.5573 0.0000 0.6378 15.5390 56.8139 –2.9799 North America 633 –0.0118 2.0050 –13.1758 –0.4628 0.0489 0.6349 11.2945 12.4726 –1.0445 Asia (Not MENA) 3.794 –0.0384 1.5842 –14.3224 –0.5369 0.0000 0.5527 9.7984 11.7185 –0.9488 MENA 2.496 –0.0716 1.5061 –28.7827 –0.3658 0.0000 0.3727 21.4684 76.6867 –2.5485 Africa (Not MENA) 2.110 0.0229 1.6985 –14.5260 –0.4473 0.0000 0.4291 15.3517 18.8995 0.7590 Oceania 422 –0.0227 1.6322 –10.2030 –0.4377 0.0700 0.6133 6.9366 9.3494 –1.1878 World Index 211 0.0109 1.8826 –10.4412 –0.3800 0.0944 0.6229 8.4062 10.7265 –1.2267 SMB 211 –0.0160 0.6852 –5.3700 –0.2350 0.0100 0.2800 2.0500 17.5090 –2.3785 HML 211 –0.1123 0.6883 –2.7900 –0.4150 –0.1200 0.2150 2.3800 3.5084 –0.1669 Notes: Statistics multiplied by 100 (except kurtosis and asymmetry). Kurtosis is the excess of kurtosis. e number of observations change according to the number of countries included. Source: Own elaboration. P. L. Angosto-Fernández, V. Ferrández-Serrano, World capital markets facing the first wave 19 Table 2 shows the main statistics for returns by region and for the MSCI World Index, and the Fama and French risk factors. The sample is character - ized by negative returns in mean (not in median), which may suggest a high concentration of losses due to a relatively high volatility and fat tails. 3. Results e Th most relevant results of the research are presented below. The r fi st point will present the results of the event study divided into the study originating from the first local case of coronavirus and the one originating from the declaration of a pandemic. The second will present the results of the time series analysis taking into account the different performance models used for this purpose. Finally, a brief comparison of the two methods will be presented highlighting which markets are winners and losers. Throughout the section the results are presented at a global level and subdivided by region in order to have a better view of the differences between countries. Also, in each table the significance of the global (and regional) average for each of the periods and coeci ffi ents is tested, as well as a detailed analysis of the negative cases. 3.1. Traditional event study In this section, the average abnormal return of week T is den fi ed as AR( T) and the accumulated abnormal returns between time 0 and week T as CAR(0,T). Firstly, the question of how to interpret the data needs to be clarie fi d as each country has a different event period. Two different approaches are taken. One is to take the week of the first case in each country as zero moment and ac - cumulate the respective abnormal returns in the same way. Thus, the AR(0) for Cambodia is directly comparable to the AR(0) for Denmark although the former corresponds to the week of 27 January and the latter to the week of 24 February. In the corresponding tables only the first twelve weeks are shown because ae ft r that the sample drops dramatically. The other is to establish an - other reference point this being the week when the WHO declared the new coronavirus a pandemic as this was the most outstanding event to analyse. To do this it is necessary to order the data chronologically. a Th t is to say that the zero moment is den fi ed as 31 December 2019 and the weeks surrounding the declaration made on 11 March are analysed. It was during that same week in March when the highest number of lockdowns by country occurred (especially in Europe), so in addition to the ee ff ct recorded in reaction to the announce - ment is the ee ff ct of lockdown on investors’ expectations. Table 3 shows the abnormal returns for the first approach. Apart from the F-test to verify the joint hypothesis of global signic fi ance (abnormal returns dif - ferent from zero), of particular interest is the number of individually negative 20 Economics and Business Review, Vol. 8 (22), No. 4, 2022 as well as negative and signic fi ant countries. In the view of the authors, these indicators provide an overview of the number of countries ae ff cted. Firstly, the daily average AR during week zero was –0.43%. This represents an important abnormal loss, where a country with a loss of 3 standard devia- tions above the mean would have sue ff red a daily fall of 3.39%. This bad per - formance is also verifiable through the number of countries with positive ab - normal returns, just 28 out of 80. During weeks zero and one there is a con- centration of nineteen national minimums, which is higher than the number that would be reached if they were equally distributed during the pandemic. i Th s makes sense if weekly developments are observed. It is true that cumula - tive abnormal returns continue to decline throughout the event, but not at the same rate. During the first two weeks 34.21% of total accumulated returns had already been lost which is also much higher than the corresponding figure if the loss were accumulated equally. iTh s means that a very signic fi ant part of the information that investors considered was present as the virus appeared in the different countries. At the end of the period studied the global average CAR is –11.64%. Here a country with a loss of three standard deviations represented an abnormal loss of 53.38%. The number of countries with a negative performance increased consistently with only eleven countries above zero at the end and with more than 40% of CARs being negative and statistically signic fi ant. As shown in the rest of the panels abnormal performance is not equally distributed among the regions. Europe, Eastern Europe, and South America and the Caribbean worse than average data in all respects with Eastern Europe being the worst per- former in the sample. Conversely, North America, Asia, MENA region, Africa and Oceania present better data than the global average. Nevertheless, North America and Oceania are the only ones with positive averages at some point during the event period. Finally, the number of negative and signic fi ant coeci ffi ents may be mis - leading and appear small, but this is because it represents which countries had signic fi ant falls in that particular cumulative week or weeks. Taking the entire event period, it is observable that 92.50% of the countries had at least one week of signic fi ant negative abnormal returns, and 73.75% had at least two. The sig - nic fi ance and generality of these poor results at the global level is in line with the results obtained by Ashraf (2021), Heyden and Heyden (2021), Pandey and Kumari (2021), or Ramelli and Wagner (2020), among others. As the results using the 3-factor model do not alter the extended market model results to any great extent, the details are not reported here, but the graph illustrating the main die ff rence (the average abnormal returns) is. Figure 1 shows that CARs in the 3-factor model are always higher than the extended market model, so the risk factors related to the size and book value explain a little more about loss due to the first wave of COVID-19. Overall, the results are still nega - tive and statistically equally significant, but their economic significance is lower. [21] Table 3. Abnormal and cumulative abnormal returns. First case week as week zero Panel A: World Panel B: Europe AR/CAR (0) (0.1) (0.2) (0.9) (0.10) (0.11) (0) (0.1) (0.2) (0.9) (0.10) (0.11) Sample 80 80 80 80 77 70 19 19 19 19 19 19 Mean –0.429 –3.982 –4.420 –8.895 –10.292 –11.641 –0.649 –4.829 –5.631 –9.322 –12.724 –13.884 SD 0.987 7.455 7.824 13.967 14.260 13.912 0.930 7.191 9.164 8.842 9.731 9.479 Negative (%) 65.0 72.5 73.8 77.5 81.8 84.3 73.7 84.2 78.9 84.2 94.7 94.7 Negative & 36.3 38.8 32.5 37.5 40.3 41.4 57.9 42.1 47.4 31.6 52.6 52.6 significant (%) Avg Adj R2 0.435 0.645 F-test 4.383*** 5.976*** 6.153*** 4.725*** 4.604*** 4.494*** CS t-test –3.886*** –4.777*** –5.052*** –5.696*** –6.333*** –7.001*** –3.045*** –2.927*** –2.678** –4.595*** –5.699*** –6.384*** Panel C: Eastern Europe Panel D: South America and the Caribbean AR/CAR (0) (0.1) (0.2) (0.9) (0.10) (0.11) (0) (0.1) (0.2) (0.9) (0.10) (0.11) Sample 7 7 7 7 7 7 9 9 9 9 9 7 Mean –1.214 –10.652 –10.290 –18.905 –20.005 –20.340 –0.941 –9.381 –8.561 –8.848 –8.627 –14.179 SD 0.444 6.869 9.656 7.749 9.939 9.614 0.853 9.707 7.760 22.463 25.608 23.403 Negative (%) 100.0 100.0 85.7 100.0 100.0 100.0 88.9 88.9 88.9 77.8 77.8 85.7 Negative & 85.7 85.7 71.4 57.1 57.1 57.1 33.3 44.4 44.4 44.4 44.4 57.1 significant (%) Avg Adj R2 0.511 0.407 CS t-test –7.227*** –4.103*** –2.819** –6.455*** –5.325*** –5.597*** –3.309*** –2.899** –3.309*** –1.182 –1.011 –1.602 [22] Panel E: North America Panel F: Asia AR/CAR (0) (0.1) (0.2) (0.9) (0.10) (0.11) (0) (0.1) (0.2) (0.9) (0.10) (0.11) Sample 3 3 3 3 3 3 18 18 18 18 17 16 Mean 0.000 1.150 0.078 –4.321 –6.446 –7.916 –0.194 –1.133 –1.188 –8.500 –8.228 –8.180 SD 0.389 1.861 2.931 10.762 12.161 13.747 1.001 6.302 4.845 17.058 15.665 16.382 Negative (%) 66.7 33.3 33.3 33.3 66.7 66.7 55.6 66.7 66.7 66.7 70.6 68.8 Negative & 0.0 0.0 0.0 33.3 33.3 33.3 16.7 22.2 5.6 33.3 35.3 31.3 significant (%) Avg Adj R2 0.789 0.314 CS t-test –0.002 1.069 0.046 –0.695 –0.918 –0.997 –0.821 –0.763 –1.040 –2.114** –2.166** –1.997* Panel G: MENA Panel H: Africa AR/CAR (0) (0.1) (0.2) (0.9) (0.10) (0.11) (0) (0.1) (0.2) (0.9) (0.10) (0.11) Sample 12 12 12 12 12 12 10 10 10 10 8 4 Mean 0.056 –1.997 –4.233 –9.664 –8.865 –9.781 –0.306 –3.495 –2.982 –2.914 –6.629 –6.458 SD 0.504 3.073 5.334 7.498 8.006 8.524 1.343 6.614 7.318 12.770 8.681 8.147 Negative (%) 41.7 66.7 83.3 91.7 83.3 91.7 60.0 60.0 70.0 70.0 75.0 75.0 Negative & 16.7 33.3 41.7 58.3 25.0 25.0 30.0 30.0 20.0 20.0 25.0 25.0 significant (%) Avg Adj R2 0.286 0.285 CS t-test 0.388 –2.251** –2.749** –4.465*** –3.836*** –3.975*** –0.720 –1.671 –1.289 –0.722 –2.160* –1.585 fi fi fi [23] Panel I: Oceania AR/CAR (0) (0.1) (0.2) (0.9) (0.10) (0.11) Sample 2 2 2 2 2 2 Mean 0.440 4.035 2.132 –5.736 –7.195 –5.796 SD 0.396 3.068 1.098 7.523 10.322 10.953 Negative (%) 0.0 0.0 0.0 50.0 50.0 50.0 Negative & 0.0 0.0 0.0 0.0 50.0 50.0 significant (%) Avg Adj R2 0.519 CS t-test 1.571 1.860 2.745 –0.985 –0.748 –1.203 Notes: Abnormal returns and standard deviations multiplied by 100. Average data from the cross-section. F-test critical values for the joint hypothesis of global signicance. CS test is the cross-sectional critical values for the global and regional signicance hypothesis. ***, ** and * means signicant at 1%, 5% and 10%, respectively. Source: Own elaboration. 24 Economics and Business Review, Vol. 8 (22), No. 4, 2022 –2,000 0 2 4 6 8 10 12 –3,000 –4,000 –5,000 –6,000 –7,000 –8,000 –9,000 –10,000 –11,000 –12,000 Figure 1. Extended market model (dashed) vs. 3-factor model (solid) Notes: Average CAR (0,1) to CAR (0,11). All multiplied by 100. Source: Own elaboration. Table 4 shows the results regarding the weeks surrounding the pandem- ic declaration by the WHO. In this case four reference weeks are taken, from one week before the declaration to two weeks ae ft r. e r Th eason for taking the previous week is because of the high concentration of minimums during that week which may lead the reader to suspect a certain anticipation on the part of economic agents. This would be consistent with the findings of Pandey and Kumari (2020) when they similarly analysed the announcement of a health emergency prior to the announcement of a pandemic. 59 out of 80 local mini- mums are concentrated in these four weeks and the high level of negative and negative and signic fi ant cases support this view as well. Ae ft r analysing the full sample of the first wave of the pandemic, the worst weeks for the stock mar - ket worldwide are located—from 2 March 2 up to 27 March. These findings do not quite fit with those obtained by Narayan and others (2021) since it is in these weeks where the bulk of lockdowns are concentrated, a measure that according to that research is related to positive abnormal returns. Although it is possible that two opposite ee ff cts coexist (a positive one derived from the lockdowns and a negative one due to the announcement) and that depending on the sample, one predominates. It is noteworthy that the most striking results correspond to week two af- ter the pandemic announcement. It is the worst week globally and in five out of eight regions a maximum of 90% negative cases is reached with a maxi- mum of almost 50% of the cases being statistically significant. The mean for this week implies an average additional loss of –0.86% daily during that week (–4.30% accumulated), with seventeen countries exceeding the –1.5% figure. i Th s is especially true for the Europe panel. Meanwhile, the week corresponding to the announcement has an average of –0.61%, which in the case of Eastern Europe is almost three times lower and with a relatively low standard devia- fi fi [25] Table 4. Abnormal and cumulative abnormal returns. Pandemic announcement as week zero Panel A: World Panel B: Europe AR/CAR (–1) (0) (2) (–1,2) (0,2) (–1) (0) (2) (–1,2) (0,2) Sample 58 70 80 80 80 17 19 19 19 19 Mean –0.536 –0.612 –0.861 –8.500 –6.559 –1.114 –0.825 –1.281 –5.394 –0.411 SD 0.790 1.113 1.087 8.433 8.148 0.377 1.212 0.673 7.699 8.306 Negative (%) 72.4 74.3 90.0 85.0 76.3 100.0 84.2 100.0 84.2 47.4 Negative & signicant (%) 48.3 42.9 48.8 55.0 45.0 94.1 68.4 78.9 57.9 15.8 Avg Adj R2 0.435 0.645 F-test 8.546*** 12.497*** 14.210*** 10.335*** 8.769*** CS t-test –5.161*** –4.602*** –7.085*** –9.015*** –7.200*** –12.178*** –2.967*** –8.292*** –3.053*** –0.216 Panel C: Eastern Europe Panel D: South America and the Caribbean AR/CAR (–1) (0) (2) (–1,2) (0,2) (–1) (0) (2) (–1,2) (0,2) Sample 5 7 7 7 7 3 7 9 9 9 Mean –0.819 –1.655 –0.997 –16.070 –13.144 –0.766 –0.366 –0.717 –13.913 –12.636 SD 0.890 0.684 0.742 8.340 7.796 0.929 0.918 1.773 5.725 4.667 Negative (%) 80.0 100.0 100.0 100.0 85.7 66.7 71.4 77.8 100.0 100.0 Negative & signicant (%) 60.0 71.4 42.9 85.7 57.1 33.3 14.3 33.3 77.8 77.8 Avg Adj R2 0.511 0.407 CS t-test –2.058* –6.405*** –3.558** –5.098*** –4.460*** –1.428 –1.054 –1.213 –7.291*** –8.122*** fi fi [26] Panel E: North America Panel F: Asia AR/CAR (–1) (0) (2) (–1,2) (0,2) (–1) (0) (2) (–1,2) (0,2) Sample 3 3 3 3 3 15 16 18 18 18 Mean –0.264 0.248 –1.087 –7.438 –6.119 –0.163 –0.181 –0.679 –8.295 –7.616 SD 0.389 0.515 0.832 7.601 5.844 0.757 1.273 1.455 9.727 8.232 Negative (%) 66.7 33.3 100.0 66.7 66.7 46.7 56.3 77.8 77.8 77.8 Negative & signicant (%) 33.3 0.0 66.7 66.7 66.7 20.0 31.3 44.4 44.4 44.4 Avg Adj R2 0.789 0.314 CS t-test –1.174 0.835 –2.263 –1.695 –1.814 –0.834 –0.570 –1.981* –3.618*** –3.925*** Panel G: MENA Panel H: Africa AR/CAR (–1) (0) (2) (–1,2) (0,2) (–1) (0) (2) (–1,2) (0,2) Sample 11 12 12 12 12 2 4 10 10 10 Mean –0.287 –0.689 –0.375 –7.712 –6.398 0.719 –0.348 –0.895 –4.870 –5.589 SD 0.548 0.722 0.567 5.606 5.097 0.536 0.747 0.440 5.800 5.313 Negative (%) 81.8 83.3 83.3 91.7 91.7 0.0 50.0 100.0 70.0 80.0 Negative & signicant (%) 27.3 41.7 25.0 50.0 50.0 0.0 25.0 40.0 30.0 40.0 Avg Adj R2 0.286 0.285 CS t-test –1.735 –3.307*** –2.293** –4.765*** –4.348*** 1.899 –0.931 –6.440*** –2.656** –3.327*** fi [27] Panel I: Oceania AR/CAR (–1) (0) (2) (–1,2) (0,2) Sample 2 2 2 2 2 Mean –0.393 –0.616 –1.084 –13.479 –11.516 SD 0.977 0.159 0.463 8.681 3.795 Negative (%) 50.0 100.0 100.0 100.0 100.0 Negative & signicant (%) 50.0 0.0 50.0 50.0 100.0 Avg Adj R2 0.519 CS t-test –0.568 –5.484 –3.313 –2.196 –4.292 Notes: See Table 3. Source: Own elaboration. 28 Economics and Business Review, Vol. 8 (22), No. 4, 2022 tion. Compared with the previous table any one of these three weeks repre- sents lower abnormal returns than those estimated using the week of the first infections as week zero. With respect to the CARs, the global minimum corresponds to the estimate including week minus one. During these weeks a country with a CAR three standard deviations below the average obtained a result of –33.80%, a loss of one third of its total value. Once again, this underperformance is not equally distributed and is particularly negative for Eastern Europe, South America and the Caribbean and Oceania. To illustrate the economic importance of these results it is observable that the size of the CARs associated with these three to four weeks are very similar to those accumulated over ten weeks starting from the first local contagion. For instance, if the previous CAR (0,11) represents the total abnormal loss associated with this period, 73% of it occurred during these four weeks. It should be remembered that the majority of total lockdowns also took place during these weeks. Finally, these data show that three months ae ft r the r fi st global contagion in China, investors were still discounting the ee ff cts of the pandemic and the measures taken to deal with it illustrating that the severity of the pandemic could not have been foreseen in the financial markets at the outset. As discussed previously, it is not the first time that such a sustained reaction over time to an extreme event is documented and in fact Bourdeau-Brien and Kryzanowski (2017) showed that the worst ee ff cts of natural disasters were noticeable in the market up to two and three months ae ft r the tragedy. 3.2. Market sensitivity to new cases e f Th ollowing section shows and discusses the results for the sensitivity of stock market indices to the daily increase in Coronavirus cases. This part of the research has a major advantage over the traditional estimation of the event study in that it provides a single indicator of how this health, economic and political crisis ae ff cted the stock market. However, it also has two drawbacks to consider in order to interpret the data correctly. First, the evolution of the pandemic over time cannot be observed as it summarises the whole period in a single coeci ffi ent and second, the coeci ffi ent obtained is directly related to daily growth data. This is not to say that it is not relevant data in itself, but in order to know the real daily infection growth it is also necessary to know the average daily growth of cases. Table 5 shows the results for the mean model just including a constant. The mean is the cross-sectional average of the γ coefficient, the standard deviation of these coeci ffi ents is also shown as well as the same relevant figures presented in the traditional event study: negative cases, negative and signic fi ant cases, the average adjusted R-square and the F-test for global signic fi ance. e s Th ample is the same as in the event study. fi ffi [29] Table 5. Sensitivity to new cases South Eastern America North World Europe Asia MENA Africa Oceania Europe and the America Caribbean Mean –0.887 –0.934 –1.287 –1.436 –0.393 –0.664 –0.492 –0.766 –2.294 SD 0.864 0.684 0.599 0.866 0.320 0.620 0.638 1.233 1.186 Negative (%) 88.8 94.7 100.0 100.0 66.7 88.9 75.0 80.0 100.0 Negative & Signicant (%) 70.0 94.7 100.0 66.7 66.7 55.6 41.7 60.0 100.0 Avg Adj R2 0.061 0.071 0.101 0.087 0.025 0.034 0.067 0.040 0.070 F-test 25.075*** CS t-test –9.179*** –5.953*** –5.685*** –4.974*** –2.126 –4.543*** –2.669** –1.964* –2.736 Notes: Coecients multiplied by 100. See Table 3. Source: Own elaboration. 30 Economics and Business Review, Vol. 8 (22), No. 4, 2022 e g Th lobal sensitivity to new cases is about –0.89% which means that for an increase of 1% in daily cases the average world returns declined by 0.0089%. Considering the average increase in cases during the pandemic (19.98%) the average daily decline of world stock indices is 0.1778%. This number is impres - sive considering that it is a daily g fi ure and that the local minimum number of months in the sample is about three. This high incidence can also be seen in almost the entire sample with only nine countries obtaining a positive co- eci ffi ent and 56 obtaining a negative and signic fi ant coeci ffi ent. It should also be borne in mind that individual signic fi ance takes into account the volatil - ity of each stock index and it is very high in a large proportion of the sample. As an example, it is 2.98% in Brazil and 0.78% in Malta and it is signic fi ant in both countries. As in the event study there are large die ff rences across regions. As the whole event is concentrated in one data point there are no positive coeci ffi ents here, but North America, Asia, the MENA region and Africa are above average in all indicators, especially North America. Previously, Oceania was one of the best performing regions in terms of abnormal performance, but now it is the one with the highest negative sensitivity. It should not be forgotten that it contains only two countries and although a country may have obtained a very high sen- sitivity to new COVID cases if few new cases occur the overall ee ff ct (which is observable through the ARs) is still small. Compared to the previous tables the first thing that stands out is the low standard deviations in relation to their mean. e Th coeci ffi ent of variation of the global sensitivity coeci ffi ent is –0.97 while for the AR (0) it was –2.30, or –1.20 for the CAR (0,11). The level of negative cases which exceeds 80% in six regions is also a notable die ff rence as is the increase in the number of these cases being statistically signic fi ant in some instances as much as double the number of ab - normal returns. The explanatory power of this model is small compared to the g fi ures of over 30% in the event study where at most 10% (Eastern Europe) is explained. i Th s was to be expected taking into consideration that the model only includes the constant and the growth variable and leaves out the market index. In summary, this all fits with the fact that the entire pandemic is reduced to one indicator per country; it produces more uniform results and makes it easier to determine the negative global incidence of COVID-19. Although the model is not very explanatory its statistical signic fi ance is robust and economic sig - nic fi ance is of large ee ff ct for any of the coeci ffi ents. Overall, these data support those found by Ashraf (2020) or Seven and Yilmaz (2020) but without forget- ting that there are large die ff rences across regions which is also oe ft n detected in the stock market ae ft r extreme events (Davies & Studnicka, 2018; Shahzad et al., 2019; Angosto-Fernández & Ferrández-Serrano, 2020). Table 6 presents the results of the regression incorporating the extended market model. While the above model proves that most capital markets reacted to new information about the pandemic it is also informative to test whether fi ffi [31] Table 6. Sensitivity to new cases (extended market model) South Eastern America North World Europe Asia MENA Africa Oceania Europe and the America Caribbean Mean –0.477 –0.460 –0.458 –0.663 0.197 –0.526 –0.335 –0.461 –1.911 SD 0.618 0.500 0.808 0.562 0.337 0.477 0.560 0.775 –0.809 Negative (%) 82.5 100.0 85.7 88.9 66.7 83.3 58.3 70.0 100.0 Negative & Signicant (%) 51.3 73.7 57.1 55.6 0.0 44.4 33.3 40.0 100.0 Avg Adj R2 0.404 0.592 0.468 0.349 0.778 0.290 0.287 0.268 0.482 F-test 11.127*** CS t-test –7.862*** –4.622*** –1.967* –5.762*** 0.924 –4.156*** –2.408** –2.162* –3.553 Notes: Coecients multiplied by 100. See Table 3. Source: Own elaboration. 32 Economics and Business Review, Vol. 8 (22), No. 4, 2022 the variables incorporated in this model (related to the global market and to domestic performance) are able to absorb the shock. e m Th ean sensitivity coeci ffi ent increases considerably across the sample with the exception of Oceania, which is quite similar. For example, the second lowest coeci ffi ent (-0.66) falls short of the global average above. Above all, North America led by Mexico has a positive performance. The levels of negative coef - ci fi ents in the sample remain very close to the previous model, but the propor - tion of significant data does fall considerably. This makes sense because some of this variability is now explained by the relationship of firms in that country with the world market. Despite this fact which may suggest a certain predict- ability of returns the coeci ffi ents are still of great statistical and economic sig - nic fi ance for most of the sample. The world average increase in cases was pe - nalised with a returns’ decline of 0.095 daily. e Th se data which are more comparable with the event study do not present major changes with respect to what was previously discussed with the excep- tion of the substantial change in the coeci ffi ents of determination. They are even lower than those obtained in the abnormal returns model (the world av- erage is 40.37% vs. 43.52% and is only higher for the MENA region). Therefore, providing a model that incorporates weekly abnormal returns and allowing for temporal evolution is more descriptive. This makes sense, but neither are these coeci ffi ents very high for the amount of additional information they incorpo - rate. On average the results obtained show that sensitivity to the increase in cases explains 93% of what the traditional event study is capable of explaining. e Th results regarding the 3-factor model are not reported because they do not alter the results shown in this table except for slight increases in the coeci ffi ents and positive cases and slight decreases in the coeci ffi ents of determination. 3.3. Winners and losers: Model comparison: e n Th ext two tables present the countries most ae ff cted during the first wave of the pandemic. The first table shows the countries with the lowest and the high - est cumulative returns. To do so the CAR (0,9) is taken because it is the last g fi ure that contains the full sample (80 countries) and only those CARs that are statistically signic fi ant at 10% are reported. Both Venezuela and Zimbabwe have CARs higher than 30%, but they are excluded since they reported infla - tion rates higher than 500%. Table 7 shows the overall negative ee ff ct with only three countries showing signic fi ant positive data (five if the high-ina fl tion coun - tries are included) and interestingly, one of them is the country where the first outbreak of the virus was detected. Observable is also the dispersion of Asian data which tops the list of gainers and losers with figures exceeding –40%, an abnormal loss of close to half the value of the index in just ten weeks. In Table 8 again winners and losers are compared, but according to the size of the sensitivity coeci ffi ent obtained in the mean model. In addition, panel B P. L. Angosto-Fernández, V. Ferrández-Serrano, World capital markets facing the first wave 33 Table 7. Largest cumulative abnormal returns (extended market model) Lowest Highest Sri Lanka –45.211 ai Th land 20.209 Cambodia –40.185 China 18.233 Brazil –37.520 Argentina 15.674 Colombia –34.541 Jordan –8.092 Greece –33.505 Tunisia –8.355 Notes: Using CAR (0,9) multiplied by 100. Only signic fi ant abnormal returns. Source: Own elaboration. considers the economic ee ff ct and corrects these coeci ffi ents for the mean of the independent variable data. Among the countries most sensitive to the in- crease in caseloads are Namibia and Australia which disappear in panel B. This is due to a much lower economic ee ff ct because the increases in caseloads were very low with relatively low total caseloads. However, the column of winners remains almost unchanged. This is consistent with the explanation that cul - tural factors, such as fear, have a signic fi ant inu fl ence on the financial market (Ashraf, 2021; Fernández-Pérez et al., 2021). Some countries are highly sensitive to the increase in cases despite having few cases compared to other countries. With respect to the above table there is no comparison whatsoever with countries with higher values. This can be explained by the delay in registering the first case in each country as the results of this second table depend directly on this fact and it is not the same to have cases concentrated in three weeks as in ten weeks. Moreover, it should be remembered that data that are statistically signic fi ant in one model do not necessarily have to be signic fi ant in another. The Table 8. Most sensitive countries (mean model) Panel A: Largest coefficients Panel B: Mean adjusted coefficients Lowest Highest Lowest Highest Namibia –3.758 Netherlands –0.048 Brazil –0.808 Netherlands –0.018 Australia –3.480 Italy –0.082 Colombia –0.799 Italy –0.052 Brazil –2.886 Cote d’Ivore –0.274 Greece –0.541 Cote d’Ivore –0.064 Colombia –2.838 Germany –0.318 Morocco –0.493 Germany –0.065 Greece –2.760 Egypt –0.375 South Africa –0.485 Sri Lanka –0.074 Notes: Only signic fi ant coeci ffi ents. Largest coeci ffi ents are the γi from the mean model equation. Mean adjusted coeci ffi ents are the γi from the mean model equation multiplied by the average daily growth in cases. All coeci ffi ents multiplied by 100. Source: Own elaboration. 34 Economics and Business Review, Vol. 8 (22), No. 4, 2022 disappearance of Sri Lanka and Cambodia is explained by the fact that their stock indices fell signic fi antly in the first ten weeks, despite having a very low incidence of cases (1633 and 125 respectively as of 1 June); hence the sensitiv- ity coeci ffi ent is very small or not signic fi ant. In the view of the authors the ex - cessive incidence in these countries in relation to the few reported cases may be due to an interdependent relationship which would be consistent with the contagion ee ff ct found ae ft r the Japan earthquake (Valizadeh et al., 2017), for example, with more ae ff cted countries and/or to investors estimating a higher number of infections than the authorities. Furthermore, this limits the ability of the coec ffi ients associated with the cases as estimators of COVID-19 impact, as they are used in the research of Alkhatib and others (2022). e n Th ext two figures (maps) are presented below to allow an appreciation of the difference between markets most ae ff cted in terms of capitalisation loss (Figure 2) and those most case-sensitive (Figure 3). The first corresponds to the abnormal performance during the four weeks around the pandemic dec- laration and the second shows the coeci ffi ents of sensitivity to the model of mean cases. In both maps black represents non-signic fi ant coeci ffi ents and grey countries are out of the sample. e v Th ast majority of markets are sensitive to the increase in cases while in the case of CARs (–1,2) a higher amount of non-signic fi ant data is observable including important markets such as the USA, the UK, China or India. It is par- ticularly striking that China and part of Southeast Asia are not signic fi ant in either case. This could be because the estimation period for these countries is too long (from the first case to 1 June); however, this is not the case as changes to the event period were implemented reducing it to 1 April and virtually all No significative Higher than –9.91% –16.56% to –9.91% –23.21% to –16.56% Lower than –23.21% Figure 2. Cumulative Abnormal Returns from week –1 to 2 (CAR (–1, 2)). Pandemic week. Full sample Source: Own elaboration based on regression data. Thanks to mapchart.net. P. L. Angosto-Fernández, V. Ferrández-Serrano, World capital markets facing the first wave 35 No significative Higher than –0.79% –1.53% to –0.79% –2.27% to –1.53% Lower than –2.27% Figure 3. Returns sensitivity coefficient to new cases (Mean model). Full sample Source: Own elaboration based on regression data. Thanks to mapchart.net. results prevail. In summary, investor fear of the pandemic is widespread glob- ally with a sample covering more than 98% of global capitalisation, but there are large regional die ff rences and notable country exceptions. Conclusions e c Th urrent epidemic is unheard of for most of us and is not only causing a glob - al health crisis, but also a political and economic one which is why it needs to be investigated in all disciplines including finance. In this regard, this paper tries to contribute to the research about the ee ff ct of this crisis on financial markets. e t Th wo experiments presented here show evidence of the negative ee ff ct of COVID-19 on the global stock market but find notable die ff rences between countries and regions. The study of the event highlights the particularly nega - tive inu fl ence in the regions of Europe, Eastern Europe and South America and the Caribbean with countries having negative abnormal returns of more than 30%. It is also notable that the largest proportion of losses were concentrated in the weeks around the WHO announcement. The analysis of the growth in cases illustrates the negative and signic fi ant relationship with returns almost everywhere in the world and it is robust to the introduction of die ff rent return models. The economic ee ff ct of the growth in cases is enormous. Comparing the different experiments two findings stand out: the striking differences between the countries with the worst abnormal returns and those with the highest nega- tive sensitivity to growth in cases; and the absence of statistical signic fi ance in countries that would have been preliminarily included among the most ae ff cted. 36 Economics and Business Review, Vol. 8 (22), No. 4, 2022 iTh s research complements other studies in the same direction and could be followed by further research into the underlying causes of these die ff rences. For example, different investment cultures or interdependence between mar - kets which could explain why some indices are so ae ff cted despite the low lo - cal incidence of the virus. Likewise, the comparison made here may help other academics to know which method to choose depending on the objective of the research. Finally, the competent authorities may also benet f fi rom some of these results, especially those markets that, despite not having a notable increase in cases did sue ff r a strong negative stock market ee ff ct since they should evalu - ate their interdependence with other stock markets. It is also interesting to re- e fl ct on the WHO pandemic announcement since it is striking that the bulk of the losses are concentrated around that date and not around the date of the r fi st local case. This fact has implications on how important it is for public in - formation to be truthful and published in a timely manner. References Alkhatib, K., Almahmood, M., Elayan, O., & Abualigah, L. (2022). Regional analytics and forecasting for most ae ff cted stock markets: The case of GCC stock markets during COVID-19 pandemic. International Journal of System Assurance Engineering and Management, 13(3), 1298‒1308. Angosto-Fernández, P. L., & Ferrández-Serrano, V. (2020). Independence Day: Political risk and cross-sectional determinants of firm exposure ae ft r the Catalan crisis. International Journal of Finance & Economics, 27(4), 4318‒4335. Ashraf, B. N. (2020). Economic impact of government interventions during the COVID-19 pandemic: International evidence from financial markets. Journal of Behavioral and Experimental Finance, 27. Ashraf, B. N. (2021). Stock markets’ reaction to COVID-19: Moderating role of national culture. Finance Research Letters, 41, 101857. Baker, S. R., Bloom, N., & Davis, S. J. (2016). Measuring economic policy uncertainty. e Th Quarterly Journal of Economics , 131(4), 1593‒1616. Baker, S. R., Bloom, N., Davis, S. J., Kost, K., Sammon, M., & Viratyosin, T. (2020). The unprecedented stock market reaction to COVID-19. e R Th eview of Asset Pricing Studies, 10, 742‒758. Barro, R. J. (2006). Rare disasters and asset markets in the twentieth century. The Quarterly Journal of Economics, 121(3), 823‒866. Bourdeau-Brien, M., & Kryzanowski, L. (2017). The impact of natural disasters on the stock returns and volatilities of local firms. e Q Th uarterly Review of Economics and Finance, 63, 259‒270. Brooks, R. M., Patel, A., & Su, T. (2003). How the equity market responds to unantici- pated events. e Th Journal of Business , 76(1), 109‒133. Brown, K. C., Harlow, W. V., & Tinic, S. (1988). Risk aversion, uncertain information, and market efficiency. Journal of Financial Economics, 22(2), 355‒385. Chiang, T. C. (2019). Economic policy uncertainty, risk and stock returns: Evidence from G7 stock markets. Finance Research Letters, 29, 41‒49. P. L. Angosto-Fernández, V. Ferrández-Serrano, World capital markets facing the first wave 37 Contessi, S., & De Pace, P. (2021). The international spread of COVID-19 stock market collapses. Finance Research Letters, 42, 101894. Davies, R. B., & Studnicka, S. (2018). e Th heterogeneous impact of Brexit. Early indica - tions from the FTSE. European Economic Review, 110, 1‒17. Donadelli, M., Kizys, R., & Riedel, M. (2017). Dangerous infectious diseases: Bad news for Main Street, good news for Wall Street?. Journal of Financial Markets, 35, 84‒103. Economic Policy Uncertainty. (n.d.). Monthly Global Economic Policy Uncertainty Index. Retrieved September 6, 2022 from http://www.policyuncertainty.com/index.html European Union Open Data Portal. (2020). European Centre for Disease Prevention and Control. COVID-19 Coronavirus data daily (up to 14 December 2020). Retrieved December 27, 2020 from https://data.europa.eu/data/datasets/covid-19-coronavi- rus-data?locale=en Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33, 3–56. Fernández-Pérez, A., Gilbert, A., Indriawan, I., & Nguyen, N. H. (2021). COVID-19 pandemic and stock market response: A culture ee ff ct. Journal of Behavioral and Experimental Finance, 29. Goodell, J. W., & Vähämaa, S. (2013). US presidential elections and implied volatil- ity: The role of political uncertainty. Journal of Banking & Finance, 37, 1108‒1117. Gormsen, N. J., & Koijen, R. S. J. (2020). Coronavirus: Impact on stock prices and growth expectations. e Th Review of Asset Pricing Studies , 10, 574‒597. He, Y., Nielsson, U., & Wang, Y. (2017). Hurting without hitting: The economic cost of political tension. Journal of International Financial Markets, Institutions & Money, 51, 106‒124. Heyden, K. J., & Heyden, T. (2021). Market reactions to the arrival and containment of COVID-19: An event study. Finance Research Letters, 38. Hillier, D., & Loncan, T. (2019). Political uncertainty and stock returns: Evidence from the Brazilian Political Crisis. Pacic-B fi asin Finance Journal , 54, 1‒12. Iheonu, C. H., & Ichoku, H. E. (2022) Terrorism and investment in Africa: Exploring the role of military expenditure. Economics and Business Review, 8(2), 92‒112. Investing. (2022). World and sector indices. Retrieved September 22, 2022 from https:// uk.investing.com/indices/world-indices Kaplanski, G., & Levy, H. (2010). Sentiment and stock prices: The case of aviation di - sasters. Journal of Financial Economics, 95, 174‒201. Karafiath, I. (1988). Using dummy variables in the event methodology. The Financial Review, 23(3), 351‒357. Kenneth R. French. (2022). Kenneth R. French—data library. Retrieved December 27, 2020 from http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/index.html Li, W., Chien, F., Kamran, H. W., Aldeehani, T. M., Sadiq, M., Nguyen, V. C., & Taghizadeh- Hesary, F. (2022). The nexus between COVID-19 fear and stock market volatility. Economic Research-Ekonomska Istraživanja, 35(1), 1765‒1785. Liu, L. X., Shu, H., & Wei, K. C. J. (2017). e i Th mpacts of political uncertainty on as - set prices: Evidence from the Bo scandal in China. Journal of Financial Economics, 125, 286‒310. Liu, Y., Wei, Y., Wang, Q., & Liu, Y. (2022). International stock market risk contagion during the COVID-19 pandemic. Finance Research Letter, 45, 102145. 38 Economics and Business Review, Vol. 8 (22), No. 4, 2022 Narayan, P. K., Phan, D. H. B., & Liu, G. (2021). COVID-19 lockdowns, stimulus pack- ages, travel bans, and stock returns. Finance Research Letters, 38. O’Donnell, N., Shannon, D., & Sheehan, B. (2021). Immune or at-risk? Stock mar- kets and the signic fi ance of the COVID-19 pandemic. Journal of Behavioral and Experimental Finance, 30, 100477. Pandey, D. K., & Kumari, V. (2021). Event study on the reaction of the developed and emerging stock markets to the 2019-nCoV outbreak. International Review of Economics and Finance, 71, 467‒483. Papakyriakou, P., Sakkas, A., & Taoushianis, Z. (2019) The impact of terrorist attacks in G7 countries on international stock markets and the role of investor sentiment. Journal of International Financial Markets, Institutions & Money, 61, 143‒160. Ramelli, S., & Wagner, A. F. (2020). Feverish stock price reactions to COVID-19. The Review of Corporate Finance Studies, 9, 622‒655. Rizwan, M. S., Ahmad, G., & Ashraf, D. (2020). Systemic risk: The impact of COVID-19. Finance Research Letters, 36. Samitas, A., Kampouris, E., & Polyzos, S. (2022). COVID-19 pandemic and spillover ee ff cts in stock markets: A financial network approach. International Review of Financial Analysis, 80, 102005. Schiereck. D., Kiesel, F., & Kolaric, S. (2016). Brexit: (Not) another Lehman moment for banks?. Finance Research Letters, 19, 291‒297. Seven, Ü., & Yilmaz, F. (2020). World equity markets and COVID-19: Immediate re- sponse and recovery prospects. Research in International Business and Finance, 56. Shahzad, K., Rubbaniy, G., Lensvelt, M. A. P., & Bhatti, T. (2019). UKs stock market reaction to Brexit process: A tale of two halves. Economic Modelling, 80, 275‒283. Smales, L. A. (2016). The role of political uncertainty in Australian financial markets. Accounting & Finance, 56, 545‒575. Spatt, C. S. (2020). A tale of two crises: The 2008 mortgage meltdown and the 2020 COVID-19 crisis. e Th Review of Asset Pricing Studies , 10, 759‒790. Uddin, M., Chowdhury, A., Anderson, K., & Chaudhuri, K. (2021). The ee ff ct of COVID–19 pandemic on global stock market volatility: Can economic strength help to manage the uncertainty?. Journal of Business Research, 128, 31–44. Valizadeh, P., Karali, B., & Ferreira, S. (2017). Ripple ee ff cts of the 2011 Japan earthquake on international stock markets. Research in International Business and Finance, 41, 556‒576. Wagner, A. F., Zeckhauser, R. J., & Ziegler, A. (2018). Company stock price reactions to the 2016 election shock: Trump, taxes, and trade. Journal of Financial Economics, 130, 428‒451. Yu, X., Xiao, K., & Liu, J. (2022). Dynamic co-movements of COVID-19 pandemic anxi- eties and stock market returns. Finance Research Letters, 46, 102219. Zaremba, A., Kizys, R., Aharon, D. Y., & Demir, E. (2020). Infected markets: Novel coro- navirus, government interventions, and stock return volatility around the globe. Finance Research Letters, 35. Zellner, A. (1962). An eci ffi ent method of estimating seemingly unrelated regres - sions and tests for aggregation bias. Journal of the American Statistical Association, 57(298), 348‒368. Zhang, D., Hu, M., & Ji, Q. (2020). Financial markets under the global pandemic of COVID-19. Finance Research Letters, 36.
Journal
Economics and Business Review – de Gruyter
Published: Dec 1, 2022
Keywords: financial markets; event study; COVID-19; coronavirus; stock returns; G01; G14; G15; F65; C32