COVID-19 Pandemic and Romanian Stock Market Volatility: A GARCH Approach

: This paper investigates the volatility of daily returns on the Romanian stock market between January 2020 and April 2021. Volatility is analyzed by means of the representative index for Bucharest Stock Exchange (BSE), namely, the Bucharest Exchange Trading (BET) index, along with twelve companies traded on BSE. The quantitative investigation was performed using GARCH approach. In the survey, the GARCH model (1,1) was applied to explore the volatility of the BET and BSE traded shares. Conditional volatility for the daily return series showed noticeable evidence of volatility that shifts over the explored period. In the ﬁrst quarter of 2020, the Romanian equity market volatility increased to a level very close to that recorded during the global ﬁnancial crisis of 2007–2009. Over the next two quarters, volatility had a downward trend. Besides, after VAR estimation, no causal connection was found among the COVID-19 variables and the BET index.


Introduction
The coronavirus malady (COVID-19) is a sanitary and economic turning point that has harmed the basis of the human condition (Verma and Gustafsson 2020), it being one of the most acute health emergencies in the recent past (Vera-Valdés 2021). The occurrence of the disease hurt the global economies and caused insecurity on worldwide equity markets (Engelhardt et al. 2021). The extensive uncertainty of the plague and its related economic failures has triggered markets to turn extremely volatile and unpredictable . Baker et al. (2020) suggested that no prior contagious virus outburst, including the Spanish Flu, has disturbed the equity market as strongly as the COVID-19 pandemic. Since it was difficult to expect and has never hitherto arose, this slump was described as a "black swan" event (Yarovaya et al. 2021). As compared with the 2008 crash which commenced in the United States and progressively diffused to other nations with a substantial time postponement, the coronavirus disease rapidly brought the worldwide economy to a stoppage by instantaneously hampering demand and supply lines around the globe due to extensive lockdowns (Ozkan 2021). Anser et al. (2021) noticed that COVID-19 contaminated cases are the central element that impedes financial activities and reduces money allocation, but a growing number of recovered cases offer investors' trust to boost stock trade across nations. Agarwalla et al. (2021) documented that the rescue package had limited the extreme tail risks, but the volatility level persisted at a high level. Ghorbel and Jeribi (2021) claimed that equity indices and financial assets rely not only on their earlier volatility, but also on the preceding volatility of the fuel prices. Therefore, in the aftermath of SARS-CoV-2 virus diffusion, the unpredictability in stock exchanges substantially increased, thus causing huge shortfalls for investors (Farid et al. 2021). The decline of the composite indicator of systemic stress among February and April 2020 was equivalent to the failures it recorded at the beginning of the 2008 global financial crisis and the 2011-2012 sovereign debt crisis, whereas the collapse in March 2020 was Group member states (Czech et al. 2020), or several international markets (Al-Najjar et al. 2021;Al-Qudah and Houcine 2021;Anser et al. 2021;Banerjee 2021;Chowdhury et al. 2021;Contessi and Pace 2021;Engelhardt et al. 2021;Höhler and Lansink 2021;Rouatbi et al. 2021;Szczygielski et al. 2021b;Topcu and Gulal 2020;Vera-Valdés 2021;Youssef et al. 2021;Zhang et al. 2020). This paper aims to examine the volatility throughout the Romanian financial market during the COVID-19 pandemic. Investigating volatility is crucial, as an unexpected and substantial rise in instability may cause a financial meltdown (Uddin et al. 2021). We investigate an emerging stock exchange as long as these markets are more exposed to insecurity of pandemics and epidemics than developed markets .
In the present article, it was analyzed how the volatility on the Romanian stock market manifested itself due to the COVID-19 pandemic outbreak. Thus, in order to fulfill the objective of the study, coronavirus daily data were used between January 2020 and April 2021 for the following markets: USA, Italy, and Romania. For the Romanian capital market, we selected the representative index for Bucharest Stock Exchange (BSE), namely, the Bucharest Exchange Trading (BET) index, as well as a number of twelve shares, these being positioned in the top of the most traded on BSE at the time of this research. Regarding the variables used as proxies for COVID-19, they are related to the evolution of the new number of cases of COVID-19 registered in the USA, Italy, and Romania. Italy was the epicenter of the COVID-19 pandemic in Europe, whereas the USA had the highest number of COVID-19 cases. Moreover, the USA has some of the largest stock markets that have a strong impact on other markets. For instance, Celık (2012) proved that emerging markets appear to be the most affected by the contagion consequences from the U.S. Moreover, Le and Tran (2021) found evidence that Vietnamese and the Philippine stock markets are affected by the contagion effect from the US stock market throughout the COVID-19 pandemic.
So far, the evidence for BSE is limited, this being, to the best of our knowledge, among the first studies that address the impact of COVID-19 on the Romanian capital market.
The rest of the paper is organized as follows. Section 2 reviews the related literature. Section 3 describes the dataset and quantitative techniques. Section 4 presents and discusses the empirical outcomes. Section 5 concludes the study.

Literature Review
The occurrence of COVID-19 has harmful effects on worldwide markets , being expected to be the largest economic shock in human history (Insaidoo et al. 2021). Broadly, Xu (2021) noticed an adverse effect of a rise in the COVID-19 cases on the financial market. Chowdhury et al. (2021) claimed that European financial markets were the most terrible victim related to others. In the same vein, Youssef et al. (2021) noticed that European equity markets, excepting Italy, spread more spillovers to the whole other financial markets than they obtained, mainly through the coronavirus outburst. Szczygielski et al. (2021b) showed that pandemic insecurity has affected nearly all territories via smaller returns and heightened market volatility. Hence, the insecurity triggered by the COVID-19 outbreak and the rapidity with which the novel coronavirus dispersed around the world produced a panic in international financial markets (Lo et al. 2021). As such, Zhang and Hamori (2021) noticed that the effect of disease on the volatility of the oil and stock markets surpassed that of the 2008 global financial crisis. Moreover, Szczygielski et al. (2021a) proved that no national energy market was unharmed by COVID-19 insecurity. Hence, coronavirus disease lessened stock market liquidity involving equally the depth and the tightness facets (Mdaghri et al. 2021). For S&P 500 enterprises, Chebbi et al. (2021) documented a negative link among the quotidian increase in the numbers of coronavirus cases and fatalities and stock liquidity. Moreover, for the Shanghai stock market, Ftiti et al. (2021) confirmed the rise of stock market volatility and liquidity risk justified by a ripple effect triggered by the vulnerability of the sanitary sector. However, Curto and Serrasqueiro (2021) argued that coronavirus occurrence did not strike evenly across all the US segments and stock quotes. For instance, Milcheva (2021) noticed that the most affected segments in the US are retail and hotels, but in Asia the most affected segment is the office. Nevertheless, Höhler and Lansink (2021) noted that the food sector was less influenced by the pandemic than other segments.
A first strand of literature was focused on the reaction of stock returns to the existing pandemic. Topcu and Gulal (2020) exhibited that Asian markets were the most influenced by the pandemic, succeeded by South America and the Middle East. As such, Hongsakulvasu et al. (2020) revealed that COVID-19 disquiet negatively influenced every kind of Thailand's stock return. Al-Najjar et al. (2021) claimed that coronavirus incidents exert an adverse impact on equity market indices of G8 countries. Al-Qudah and Houcine (2021) noticed that the surge in established cases of COVID-19 negatively influenced stock returns for the main affected nations in the WHO Regions. For Central, North, and the South American realm, Amin et al. (2021) concluded that COVID-19 cases undesirably influence market indexes. Takyi and Bentum-Ennin (2021) revealed that African financial markets performance lessened between −2.7% and −20 % throughout and subsequently the incidence of the pandemic. Czech et al. (2020) reported a negative association among the Visegrad stock market indices and the COVID-19 diffusion. For the case of emerging markets, Hashmi et al. (2021) advised that the number of coronavirus cases negatively influences stock prices mainly when these financial markets are in a bearish condition. Contrariwise, O'Donnell et al. (2021) found that the everyday amounts of COVID-19 cases did not explain the index price variations in China, Spain, Italy, the United Kingdom, and the United States. Zoungrana et al. (2021) revealed for the West African Economic and Monetary Union's (WAEMU) stock market that weekly validated cases do not influence stock returns, even if the impact of death cases is harmful. However, Brueckner and Vespignani (2021) documented that COVID-19 contaminations had a positive influence on the performance of the Australian equity market.
Another strand of research was oriented on how COVID-19 news influences stock returns and oil prices because terrific fear caused by the mass media is related with growing volatility in the financial markets (Haroon and Rizvi 2020). Chundakkadan and Nedumparambil (2021) provided evidence that emphasis on the pandemic has generated a pessimistic reaction between market players and weakened the stock exchanges. Weng et al. (2021) established that news throughout the coronavirus pandemic has more forecasting information, which is essential for the transient volatility estimating of fuel futures, whereas  reinforced that considering health reports over illnesses boosts stock return foresight. Baek et al. (2020) advised a negativity tendency since adverse announcements concerning the number of fatalities are twice as impactful as optimistic facts with respect to recoveries. Wu et al. (2021) argued that media can stimulate the forecast of oil cost and usage over the COVID-19 contagion. Atri et al. (2021) noticed that the number of casualties and the COVID-19 panic adversely impact petroleum value, but the COVID-19 media coverage positively influences fuel cost in the short run.
Further studies were exploring safe-haven assets throughout ongoing health crises. Huang et al. (2021) suggested that Bitcoin can promote efficient diversification and risk alleviation, whereas Mariana et al. (2021) strengthened that Ethereum is a superior safehaven than Bitcoin. Similarly, Disli et al. (2021) advised that gold, oil, and Bitcoin offer diversification benefits at extended investment perspectives. Ji et al. (2020) underlined that gold and soybean futures may uphold the worth of an investment. Contrariwise, Będowska-Sójka and Kliber (2021) claimed that cryptocurrencies rarely performed as weak safe-haven assets during several market disorders, whilst Conlon and McGee (2020) argued that Bitcoin does not behave as a safe haven over the bear market stemming from the coronavirus disease. For financial markets of Africa, Omane-Adjepong and Alagidede (2021) concluded that the safe-haven potential of precious metals, particularly gold, has diminished. In the same vein, Umar et al. (2021) contradicted the safe-haven feature of precious metals over the coronavirus plague, apart from silver. For the case of Chinese portfolios, Pho et al. (2021) found that Bitcoin is appropriate to risk-prone investors, whereas gold is adequate to prudent investors.
A summary of prior literature exploring equity market volatility due to the coronavirus pandemic is exhibited in Table 1.

Sample Selection
For our study, we selected the most traded companies on the Bucharest Stock Exchange (BSE)-ALR, BRD, BVB, COTE, EL, FP, SNG, SNP, TEL, TLV, TRP and WINE-for the period January 2020-April 2021. To capture the types of causality between the variables regarding COVID-19 and the Romanian stock exchange, we decided to select the latest number of cases of COVID-19 registered in the USA, Italy, and Romania. The selected measures are presented in Table 2.

Variables regarding Romanian financial market
BET Bucharest Exchange Trading is the first index developed by BSE and signifies the reference index for the Romanian equity market. BET indicates the performance of the most traded corporations on BSE's regulated market, apart from financial investment enterprises. It is a free float market capitalization weighted index, with the highest weight of its components of 20%.

ALR
Alro is affiliate of Vimetco N.V., a worldwide, vertically-integrated primary and processed aluminium manufacturer. Field of activity: Aluminium production.
BRD BRD Groupe Societe Generale is the second leading bank in Romania and the fourth market capitalization on the BSE. Field of activity: Other monetary intermediation.

BVB
The Bucharest Stock Exchange is the most significant organization of the local capital market. It coordinates and operates the regulated markets of financial instruments under European guidelines. Field of activity: Administration of financial markets.

COTE
CONPET delivers specific gas transport services via tubes and by rail, ensuring the supply of the factories with domestic and imported crude oil and derivatives. It manages a 3800 km pipeline grid encompassing 24 Romanian counties. Field of activity: Transport via pipeline.

EL
Societatea Energetica Electrica is a major participant in the energy sharing and supply market in Romania. Field of activity: Business and other management consultancy activities.

FP
Fondul Proprietatea is a joint stock company running as a closed-end investment company (Alternative Investment Fund) short of a set period, integrated in Romania, trading on the BSE since January 2011, and on the London Stock Exchange since April 2015. Field of activity: Trusts, funds and similar financial entities.

SNG
Romgaz is the leading natural gas manufacturer and the key provider in Romania. It is a joint stock corporation whose majority stockholder is the Romanian State owning a 70% share. Field of activity: Extraction of natural gas.

SNP
OMV Petrom is the leading energy corporation in Southeastern Europe. The firm is involved along the whole energy value chain: from exploration and fabrication of oil and gas, to processing and fuels supply, and further on to power production and advertising of gas and power. Field of activity: Extraction of crude petroleum.

TEL
Transelectrica is the Romanian Transmission and System Operator which performs a vital position in the Romanian electricity market. It operates and runs the energy spread system and delivers the electricity connections among the Central and Eastern European nations as a member of European Network of Transmission and System Operators for Electricity. Field of activity: Transmission of electricity.
TLV Banca Transilvania is the first largest bank in Romania in terms of total assets. Field of activity: Other monetary intermediation.

TRP
TeraPlast SA is the parent corporation of the TeraPlast Group, respectively the major Romanian manufacturer of construction materials. Field of activity: Manufacture of plastic plates, sheets, tubes and profiles.

WINE
Purcari Wineries Group is a prominent participant in the wine and brandy sectors in the Central and Eastern Europe area, handling around 1,300 hectares of vineyards and 4 wineries placed in Romania and the Republic of Moldova. Field of activity: wineries. The data consist of daily observations. For the variables regarding the Romanian financial market, the data source was Thomson Reuters Datastream, whereas for the variables regarding the COVID-19 pandemic, the data source was Our World in Data.

RO_COVID
The formula for daily yields is in line with Banerjee (2021) where R i,l is the yield of the index/asset i in period l, P i,l is the asset price/stock market index value i in period l and P i,l−1 is the price of the asset/stock market index value in the l-1 period. Logarithmic yields were used because they are expected to have a normal distribution.

Quantitative Methods
To explore the selected financial time series, we will employ ARCH/GARCH models as in prior studies (Salisu and  These models simultaneously evaluate and test processes of yields and volatility processes. ARCH models were introduced by Engle (1982) and Generalized (GARCH) by Bollerslev (1986). A GARCH model allows conditional variation to be dependent on its previous lags. GARCH models transform the AR process from the ARCH model into an ARMA process by adding an MA process. The GARCH model (p, q) has the following form: where ω > 0 and α i ≥ 0, β i ≥ 0. From Equations (2) and (3), it can be seen that the conditioned variance of random perturbations depends both on the historical values of the shocks and on the values of the variance in the past. The coefficients of σ 2 t−p represent persistence of volatility, whereas the coefficients of ε 2 t−q signify the rate of reaction of volatility to shocks in the financial market. Parameter p is the order of the terms GARCH and q is the order of the ARCH terms.
According to Baybogan (2013), the core issue with an ARCH specification is that it involves a substantial number of lags to seize the type of the volatility, whereas the GARCH framework is generally much more parsimonious for the reason that it integrates much of the evidence that a larger ARCH model with considerable lags would cover.
In order to analyze the causality between the BET index and the number of new COVID-19 cases, we will estimate in the first instance three vector autoregression (VAR) models, much like those found in Anser et al. (2021), Chen et al. (2021), Chowdhury et al. (2021), and Youssef et al. (2021), incorporating the stock market index and each COVID-19 pandemic measure, as described below: where δ 1 and δ 2 are the intercepts, β, γ, ψ, and ϕ are the endogenous variables coefficients, whereas u are the residual terms. Further, for each estimated VAR model, we will employ the Granger causality test, as in earlier literature (Bourghelle et al. 2021;Chen et al. 2021;Liu et al. 2021b). So as to perform the causality test, the data series must be stationary and zero average (Granger 1969). The null hypothesis is that b does not cause Granger on c and that c does not cause Granger on b. The following bivariate regressions are given: Thus, a first step in the present study is the examination of the stationarity of the selected variables, which is an important stage in any econometric study. This will be verified by the ADF test, much like Bai et al. (2021) and Chen et al. (2021). Next, we intend to model BSE volatility through the GARCH model and identify the types of causality that are established between BSE and COVID-19 variables through the Granger causality test after VAR estimation.
The research hypotheses are formulated as follows: Hypothesis 1 (H1). The COVID-19 pandemic negatively influences the return of the Bucharest Exchange Trading Index.
Hypothesis 2 (H2). The COVID-19 pandemic adversely impacts the return of the companies traded on the Bucharest Stock Exchange. Table 3 shows descriptive statistics for the daily logarithmic returns of the shares traded on BSE, as well as for the BET stock market index, whereas Figure 1 reveals the density graphs. The selected shares have a negative skewness (except for the TRP share) in line with Agarwalla et al. (2021), Banerjee (2021), Malik et al. (2021), Yousaf (2021), and Zhang and Hamori (2021). As a common condition of financial markets, skewness is negative, suggesting an asymmetry to the left.   Table 2.

Preliminary Statistics
The Kurtosis indicator measures the magnitude of the extreme values. Accordingly, in the current investigation, all the explored variables register a value of kurtosis greater than three. This fact indicates that the return series has fatter tails than the normal distribution, similar to Banerjee (2021) Yu et al. (2021), and Zhang and Hamori (2021). This feature is referred to as leptokurtosis, which could be caused by volatility clustering.  Yu et al. (2021), and Zhang and Hamori (2021). This feature is referred to as leptokurtosis, which could be caused by volatility clustering.
Additionally, through the Jarque-Bera test, we can decide the distribution of variables. Consistent with the empirical results presented in Table 2, the probability accompanying the test is 0%. Hence, the test values are quite different from those of the normal distribution, proving that the series are not normally distributed, much like Curto and Serrasqueiro (2021), Liu et al. (2021a), Malik et al. (2021), Yousfi et al. (2021), and Zhang and Hamori (2021). Figure 2 shows the Q-Q (quantile-quantile) plots. The quantiles-quantiles graph is a straightforward method used to compare two distributions. Therewith, it signifies the graph of an empirical distribution versus a theoretical distribution (normal distribution). If the empirical distribution is normal, the subsequent Q-Q graph should be the first bisector. However, in current investigation, the distribution is very different from the normal one. Zhang and Hamori (2021). Figure 2 shows the Q-Q (quantile-quantile) plots. The quantiles-quantiles graph is a straightforward method used to compare two distributions. Therewith, it signifies the graph of an empirical distribution versus a theoretical distribution (normal distribution). If the empirical distribution is normal, the subsequent Q-Q graph should be the first bisector. However, in current investigation, the distribution is very different from the normal one.  Table 2. Figure 2. Q-Q plots for daily logarithmic returns. Source: Authors' own work. Notes: Variables' descriptions are provided in Table 2.
The density graph (see Figure 1) and Q-Q plot (see Figure 2) against the normal distribution show that the returns distribution also exhibits fat tails confirming the results in Table 2.
Further, we studied the stationarity of stocks and the stock market index using the ADF (Augmented Dickey-Fuller) test, much like  Zhang and Hamori (2021). ADF test is a very common method of assessing stationarity. The null hypothesis of the test is that the analyzed data series is not stationary and has a root unit. The outcomes of ADF test are revealed in Table 4. According to the results presented by the ADF stationarity test in Table 4, the null hypothesis of a unit root can be rejected, indicating that the daily logarithmic returns are significant at the 1% level, hence stationary, similar to Bai et al. (2021) and Yu et al. (2021). Thus, taking into account the empirical results of the ADF stationarity test, the examined variables are stationary and have an integration order I (0). Likewise, the stationarity of the series can be seen in Figure 3, where the daily yields of the analyzed series are represented.   Table 2. Figure 3 shows the evolution of the selected returns. Hence, there is acknowledged a phenomenon of "volatility clustering" and an alternation between periods of low volatility and those with high volatility, similar to Abuzayed et al. (2021), Insaidoo et al. (2021), Malik et al. (2021), and Yousfi et al. (2021). Moreover, "volatility clustering" implies a strong autocorrelation of returns. Figures 4-6 reveal the evolution of the BET index against the new cases of COVID-19. The relationship between the evolution of the BET index and the number of new SARS-CoV-2 cases (USA, Italy, and Romania) is an indirect one. Thus, the increase in the number of infections (USA, Italy, and Romania) determined a decrease in the local stock market index and its return.  Table 2.  Table 2.   Table 2.  Table 2.

GARCH Outcomes
Before applying GARCH models, it is advisable to perform preliminary tests to detect the effects of ARCH. Heteroskedasticity was investigated by determining autocorrelation (AC), partial autocorrelation (PAC), and Q test. The number of offsets used for all the time series was 20. The outcomes of AC, PAC and Q-Stat are revealed in Table 5.  Table 2.

GARCH Outcomes
Before applying GARCH models, it is advisable to perform preliminary tests to detect the effects of ARCH. Heteroskedasticity was investigated by determining autocorrelation (AC), partial autocorrelation (PAC), and Q test. The number of offsets used for all the time series was 20. The outcomes of AC, PAC and Q-Stat are revealed in Table 5. According to the results of the Q test, in most of the cases, the existence of the serial correlation, heteroscedasticity (p-value less than 5%), is confirmed. However, in the case of TRP, the probability is greater than 5% and the null hypothesis of the absence of the serial correlation up to lag 20 cannot be rejected. Therefore, the data series shows heteroscedasticity that can be modelled by GARCH models (except TRP, because heteroskedasticity is a pre-condition for applying GARCH models for financial time series, where we may not be able to match GARCH models). Table 6 shows the outcomes of GARCH approach. The model used was GARCH (1,1), restriction-Variance target, error distribution: Student's t being selected to register among the smallest AICs among the other available variants, similar to Czech et al. (2020) and Xu (2021). Only valid models, whose coefficients are statistically significant and different from 0, have been selected.
Following the application of a GARCH model (1,1), we estimated the conditioned volatilities which are plotted in Figure 7.  Table 2.
Further, Figure 8 exhibits the daily evolution of selected shares' yields and the BET index for the period 2007-2021 in order to highlight the fact that the volatility in the period 2007-2009 was much more significant than that during the COVID-19 pandemic. This fact supports Le and Tran (2021), which pointed out for the case of Vietnam that the contagion effect throughout the coronavirus period was lesser than that over the global financial crisis.  Table 2. Thus, the first two quarters of 2020 were marked by an increase in volatility on international financial markets, more pronounced in March and April, and the companies   Table 2.
We notice that in the first quarter of 2020, the volatility of the Romanian capital market increased to a level very close to that recorded during the global financial crisis of 2007-2009. Similarly, Curto and Serrasqueiro (2021 noticed an intensification of volatility following February 2020. Hence, this outcome is in line with both of the proposed hypotheses H1 and H2. Our results are consistent with Czech et al. (2020), which noted that Visegrad Group member countries were hit by the COVID-19 disease at the outset of March 2020 when the first case was registered.
Besides, in the next two quarters, volatility had a downward trend, argued by the fact that COVID-19 vaccine findings were declared ). In the same vein, Rouatbi et al. (2021) reinforced that the launch and expansion of the vaccinations reduce stock market volatility.
Further, Figure 8 exhibits the daily evolution of selected shares' yields and the BET index for the period 2007-2021 in order to highlight the fact that the volatility in the period 2007-2009 was much more significant than that during the COVID-19 pandemic. This fact supports Le and Tran (2021), which pointed out for the case of Vietnam that the contagion effect throughout the coronavirus period was lesser than that over the global financial crisis.
Thus, the first two quarters of 2020 were marked by an increase in volatility on international financial markets, more pronounced in March and April, and the companies FP, ALR, SNP, and BVB had the highest volatilities during this period. August, September, and October show moderate volatility, being higher than before the outbreak of the COVID-19 pandemic. Further, Figure 8 exhibits the daily evolution of selected shares' yields and the BET index for the period 2007-2021 in order to highlight the fact that the volatility in the period 2007-2009 was much more significant than that during the COVID-19 pandemic. This fact supports Le and Tran (2021), which pointed out for the case of Vietnam that the contagion effect throughout the coronavirus period was lesser than that over the global financial crisis.  Table 2.
Thus, the first two quarters of 2020 were marked by an increase in volatility on international financial markets, more pronounced in March and April, and the companies FP, ALR, SNP, and BVB had the highest volatilities during this period. August, September,

Causality Analysis
Given that the health crisis has a significant impact on the global economy, we also aimed to explore the causal relationships that are established between the variables regarding COVID-19 and the BET stock market index. Primarily, it was checked if the stock market index and each COVID-19 pandemic measure were cointegrated. In this regard, Table 7 exhibits the outcomes of the Phillips-Ouliaris cointegration test. Accordingly, we reject the null hypothesis and decide that the series are cointegrated. Further, the lag selection criterion is explored. Table 8 reveals the related lag order selection criteria. Hence, the Schwarz information criterion suggests five and seven lags. After estimating the VAR model for the stock market index and each COVID-19 variable (see Tables A1-A3), we proceed to explore the Granger causality relationships. According to Freeman (1983), a variable, X, which evolves over time, causes another variable in evolution, Y, if the predictions of the value Y based on its own past values and on the previous values of X are better than the predictions of Y based only on Y's own past values. Table 9 shows the empirical results of the Granger causality test after VAR estimation.
Thus, for the analyzed period January 2020-April 2021, no causal relationship was identified between the COVID-19 variables and the BET index. This outcome is not consistent with Liu et al. (2021b), who found that fear sentiment causes stock market crash risk. Therefore, Yu et al. (2021) cannot be maintained either since it was found that the COVID-19 Anxiety Index causes stock market returns.

Concluding Remarks
The COVID-19 virus has spread very rapidly around the globe, negatively impacting the economy, and according to the latest information, it undergoes various mutations, with new variants of COVID-19 always appearing. The study of volatility has always been a hotly debated topic by experts, especially now in these times of uncertainty. The impact of COVID-19 on the capital markets did not take long to appear, so it initially manifested itself on the largest stock markets in the world, then, due to the contagion effect, it was transmitted to the other smaller markets. To our knowledge, the studies conducted on the Romanian capital market related to the research of volatility during the pandemic are extremely limited, which led us to focus on analyzing the volatility of the BSE indices.
Our main goal of the article was to analyze the BSE volatilities during the COVID-19 pandemic, selecting indices and a group of traded shares (these being among the most traded on BSE, which are also found in the BET stock index). To study volatility, we used the GARCH model (1,1), and the graphical outputs capture the episodes of volatility. Finally, through the Granger causality test, after VAR estimation, we were able to identify the relationships to be established between BSE stock index, respectively, the shares traded on BSE and variables that capture the evolution of the COVID-19 pandemic in the USA, Italy, and Romania.
This research contributes to the existing literature, which is the reason that we studied the volatility of the main companies traded on the Bucharest Stock Exchange, between January 2020 and April 2021, a period subject to a major change due to the COVID-19 pandemic, using GARCH models. We found that the distribution of the daily return series for the Romanian stock market is leptokurtic, it is not normally distributed, and has significant time dependencies. The GARCH (1,1) model was used to model volatility on the Romanian stock market.
The study revealed strong evidence of volatility that lasts over time, a trend of high and low volatility periods, and a high persistence of volatility on the Bucharest Stock Exchange. In the first quarter of 2020, capital market volatility in Romania increased to a level very close to that recorded during the global financial crisis of 2007-2009. In the next two quarters, volatility had a downward trend. Nevertheless, no causal association was noticed between the COVID-19 variables and the BET index.
The empirical outcomes could help investors and asset managers to adjust their trading strategies. Moreover, the government should consider economic relief packages and formulate policies to lessen severe falls in prices (Hashmi et al. 2021).