79 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 How to cite this article: Enow, S. T. (2023). Volatility persistence in international financial markets in the post covid-19 era. International Journal of Banking and Finance, 18(2), 79-96. https:// doi.org/10.32890/ ijbf2023.18.2.4 VOLATILITY PERSISTENCE IN INTERNATIONAL FINANCIAL MARKETS IN THE POST COVID-19 ERA Samuel Tabot Enow The Independent Institute of Education Vega School, South Africa enowtabot@gmail.com Received: 12/8/2022 Revised: 8/11/2022 Accepted: 30/11/2022 Published: 25/6/2023 ABSTRACT The long-term behaviour of stock markets are of significant importance to asset managers and financial experts due to its direct link with security price valuation. Volatility persistence has a significant impact on the returns of security prices due to its time varying properties. However, there is no real meaningful effect of current volatility on future security prices and returns if the volatility is transitory and not persistent. The aim of this study was to explore conditional volatility properties and determine whether the current volatile environment would persist in the JSE, S&P 500, Nasdaq Index, SSE, CAC 40 and DAX markets. Using a GARCH 1.1 model and a Markov switching model, the findings revealed that volatility would persist in the JSE, S&P 500, Nasdaq Index, SSE, CAC 40, and the DAX from their ARCH and GARCH coefficients, as well as the delay parameters. In addition, the effects of past volatility in the Nasdaq, CAC 40, and DAX would remain in the forecast of variance. A diversified and broader investment approach should be used in the JSE, S&P 500, Nasdaq https://e-journal.uum.edu.my/index.php/ijbf INTERNATIONAL JOURNAL OF BANKING AND FINANCE 80 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 Index, SSE, CAC 40, and DAX indexes to mitigate risk, and portfolio formation should not concentrate on any sector or asset classes. Keywords: Volatility, financial markets, Covid-19, GARCH coefficient, ARCH term, conditional variance. JEL Code: G1, G2, G4. INTRODUCTION Modelling volatility has always been an interesting phenomenon in finance and it is certainly an interesting time to investigate volatility persistence in financial markets because of the recent Covid-19 pandemic, which has had a mirage effect around the globe. Due to the introduction of the vaccine, financial markets are recovering from market disruptions caused by the Covid-19 pandemic (Shang, 2021). The spread of fear and uncertainty has created massive swings in market prices in almost all financial markets around the world. This was evident in stock markets like the S&P 500 index that had plummeted by over 30 percent at the beginning of 2020 with the spread of the pandemic (Li, 2020). The French Stock Market index (CAC 40) and the German blue chip companies trading on the Frankfurt Stock Exchange (DAX) fell by over 12 percent, which contributed to an overall 11 percent decrease in the value of European stocks at the beginning of February 2020, before increasing to a record high (Hathorn, 2021). Prior to the widespread impact of the virus, there was a significant jump in oil prices, breaking through the $1700 per ounce ceiling for the first time since 2012. However, the oil price tumbled to $250 per ounce in March 2020, losing about 60 percent of its value in the first quarter of 2020 due to the mismatch of demand fears and supply concerns (Camp et al., 2020). A similar situation was seen in the bond market, where nearly $4 trillion of municipal bonds experienced unprecedented volatility as investors sold off their positions amid concerns related to Covid-19 (Liang, 2020). The crisis produced unrivalled government responses, which included the introduction of extraordinary stimulus packages. There was a relaxation of banking regulations to ensure capital buffers were not impeding banks from supporting and stabilising economies. These actions were partly to curb financial market volatility and help ensure price stability. 81 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 The concept of volatility persistence refers to how today’s volatility affects the conditional variance of future volatility (Wang & Yang, 2017). Therefore, today’s unconditional volatility variance may be infinite and will continue in the future. Applying this definition in financial markets means that large or small volatility changes will tend to follow the same pattern in the future with unpredictable features and successive disturbances. The study of Karanasos et al. (2014) reveals that volatility has time varying properties with structural breaks. Muguto and Muzindutsi (2022) contend that volatility should not persist, because positive and negative news will induce an equal change. There has always been a compelling need for forecasting volatility persistence in financial markets as it determines the gains and losses from the erratic behaviour of financial markets. It can also be used to determine the level of risk involved in holding a security. The aim of this study was to investigate the conditional volatility properties of the major financial markets around the world post the Covid-19 era. This study is significant to investment practitioners and market participants as it explores the extent to which the current volatile environment will persist, which has important implications for risk management and portfolio management. More specifically, knowledge of volatility persistence is important because the value of financial assets is directly linked to the level of volatility prevailing in the market; it denudes the linkage between some underlying risk factors and security price movements (Christiansen et al., 2012). Moreover, establishing the amount of risk to take should be based on the knowledge of the extent of volatility persistence in financial markets (Engle, 1982). LITERATURE REVIEW Theoretical Perspective Heightened financial market volatility is a direct consequence of macro-economic uncertainty and a lack of liquidity (Kundu & Paul, 2022). These two economic forces are the main drivers of price fluctuations in financial markets. Macro-economic uncertainty makes it very difficult to price an asset, while a lack of liquidity causes fire sales where assets are traded at a lower price (Dow & Han, 2018). A clear distinction between realised and implied volatility should be made in the analysis and discussion of market volatility. Realised 82 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 volatility is associated with technical analysis in which it is concerned with the past, and provides a vivid picture of historical asset price movements within a particular time frame (Paraschiv, 2020). Conversely, implied volatility describes expectations which are often used in option pricing and financial market trading (Mayhew, 1995). There exists a volatility gauge called the VIX index which is used by market participants and investors to access the level of risk and uncertainty in the market. In essence, the VIX is used to assess volatility expectations over a short period. High levels of uncertainty in the VIX index often results in fewer trading activities, a drop in liquidity, and a negative feedback loop. The three largest volatility spikes recorded in the VIX index were in 1987, 2008, and 2020, which was followed by reduced investor holdings. Although volatility is usually analogous to bear markets, heightened financial market volatilities are also experienced in bull markets (Elgammal et al., 2021). This was evident in the 1990 dotcom boom where market volatility rose considerably alongside the tech stocks. The growth expectations that were placed on these untested tech stocks with the accompanying excitement gave rise to uncertainties, which had led to an unsustainable bubble growth. However, market volatility does find stability when market shock subsides and when market participants get a better understanding of the economic environment (Degiannakis et al., 2014). Prior literature (Krichene, 2003; Bobeică & Bojeşteanu, 2008; Oh et al., 2008; Thupayagale, 2012; Gyamfi et al., 2016) contend that volatility tends to have a long-term memory due to recurring macro- economic cycles. These long-term memories are justified by hysteresis and repetitive irrational behaviour, which are contrary to the efficient market hypothesis (EMH). Irrational behaviours are captured in the heteroskedastic variance of the financial market (Maheu & McCurdy, 2000). Accordingly, modelling the heteroscedasticity behaviour and understanding the unconditional mean and variance of a security index is therefore, necessary for asset pricing, risk management, and portfolio optimisation. However, there are three underlying challenges in forecasting conditional volatility: (1) inferring a latent time series from a noisy observation and modelling non-linear temporal dynamics; (2) defining a positive symmetric covariance matrix; and (3) computing maximum likelihood estimations (Bauwens et al., 2006). Due to the aforementioned challenges, successfully quantifying the realised volatility does not necessarily lead to forecasting the implied 83 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 volatility. Hence, the need arises for a more sophisticated model to capture the arbitrage effect between the realised and implied volatility. The Covid-19 pandemic has caused volatility spikes in the global financial markets. This certainly calls for concern as investors, market participants, and the general public are sceptical about the increasing risk as a result of volatility spikes. Although there are some risks worth taking, most risks are detrimental and should be avoided. For example, there is a risk when investing in a market where the economy is experiencing recession. Moreover, the question of the amount of risk to be taken should be analysed in conjunction with the concept of current and future volatilities. Volatility of idiosyncratic moves in stock markets will not be rewarded, as long as volatility persists (Visaltanachoti & Pukthuanthong-Le, 2009). There are mainly two sources of volatility in stock markets, i.e., the amount of new information in the market, and macro-economic uncertainty. New information about certain events has a significant impact on stock prices. However, some information is important while others are not. New important information causes investors to change their expectations, which in turn affect the market price (Bookstaber & Pomerantz, 1989). This new information arrives in clusters and alters the way investors perceive the future. Macro-economic uncertainty is another factor that influences financial market volatility. Concerns about the macro-economic outlook contribute to financial market volatility because it is very difficult to price an asset in an uncertain environment. That is why investors are very interested in the macro- economic events associated with high volatility. According to Engle and Rangel (2006), the macro-economic factors that play an important role in driving financial market volatility include the following: high inflation, slow upward growth, recession and changes in short-term interest rates. Volatility in the financial markets is still staggering because of the Russian-Ukraine crisis. Currently, financial markets are dominated by the Russian-Ukraine crisis, which has increased volatility and uncertainty in the financial system. The spill over effect is a decreased liquidity in certain markets, which prompt investment practitioners to shift their attention to active management. Commodity-based countries and countries with large amounts of United States denominated debt are experiencing tighter financial conditions. Russia is the third largest oil producer, accounting for 11 percent of the world’s total supply (Carpenter, 2022). That is why the disruption caused by the Russian- 84 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 Ukraine war has significant consequences in the global supply of oil, as well as natural gas exports to the European Union. Considering that financial markets are largely driven by market sentiments and are also emotionally structured, fear and panic tend to be reflected in stock prices (Ackert et al., 2003). Speculative and high growth stocks tend to be significantly affected by market volatility because fundamental drivers are driven by market sentiments. Future returns and the growth of speculative stocks take a downward turn with an increase in uncertainty, and investors tend to shift their sentiments toward risk-half. This means that they are less willing to take risks in the long and short term. Volatility in financial markets during Covid-19 has been extensively investigated during the pandemic. A summary of these studies are highlighted. Table 1 Evidence of Market Volatility during the Covid-19 Pandemic Study (Author & year of study) Model Period Country Findings Topcu et al. (2021) Lag augmented vector auto regression 3 January to 15 October, 2020 United States (US) The fear and uncertainty of the pandemic triggered volatility across financial markets. Endri et al. (2021) GARCH model 2 March 2020 to 16 March 2020 Indonesia Evidence of high volatility which had a negative impact on stock price returns on the Indonesian stock exchange. Gherghina et al. (2021) GARCH model January 2020 to April 2021 Romania Notable evidence of market volatility in the Bucharest Exchange Trading index which was similar to that of the 2008-2009 financial crisis. (continued) 85 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 Study (Author & year of study) Model Period Country Findings Ibrahim et al. (2020) Continuous wavelet transformation analysis and plots and GJR-GARCH analysis 15 February to 30 May 2020 Asia-Pacific region Financial markets in China, Japan, South Korea, Malaysia and Philippines experienced high volatilities during the Covid-19 pandemic, although government interventions help curb some of the volatility. Mishra & Mishra (2020) Fixed Effect and GARCH model 2 July 2019 to 12 June 2020 China, Hong Kong, India, Indonesia, Israel, Japan, Malaysia, Philippines, Singapore, South Korea, Thailand, and Taiwan The Covid-19 pandemic amplified market volatility due to the impact of widespread fear. Rahman et al. (2021) Canonical correlation analysis 22 January 2020 to 31 December 2020 NASDAQ 100 options index, the S&P 500 and Dow Jones Industrial Average, DAX, CAC 40, and the EURO Stock 50 index The announcement of new positive and death cases from the pandemic significantly increased market volatility. Table 1 provides evidence of significant volatility during the Covid-19 pandemic. As has been well documented in the study of Muguto and Muzindutsi (2022), volatility can normalise in the future and cease to persist. Therefore, this study fills the gap in the literature by investigating the extent to which the current volatile environment will persist and in so doing, forecasts the implied volatility for selected financial markets. The next section highlights the blueprint used in the data analysis. 86 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 METHODOLOGY To investigate volatility persistence, this study used a generalised autoregressive conditional heterosedasticity (GARCH) (1,1) model as proposed by Bollerslev (1986), and the Markov switching model developed by Hamilton (1989). In the GARCH (1,1) model, conditional volatility at time T is an autoregressive moving average (ARMA) relying on past volatilities and lagged values of the error term (Bauwens et al., 2006). In this respect, the GARCH model is very useful in investigating volatility persistence with lagged shocks together with its momentum. This model provides a parsimonious alternative to the higher autoregressive conditional heteroscedasticity (ARCH) model (Ruilova & Morettin, 2020). The GARCH (1,1) model has the following two important parameters; the ARCH term and GARCH coefficient. The ARCH term captures the extent to which the volatility changes over time due to the previous lag of autoregressive conditions, while the GARCH coefficient reveals the level of volatility symmetry in the market (Bollerslev, 1986). The sum of the ARCH and GARCH coefficients reveals the extent to which volatility will persist in the financial market. Volatility persistence is evident when the sum of the ARCH and GARCH term is closer than, or equal to 1 and vice versa (Nelson, 1990). The GARCH (1,1) model (Bollerslev, 1986) is given by the formula (1): (1) Where Conditional variance α = error term ϕ = ARCH term = Lag value of Conditional variance β = GARCH coefficient = Lag square error term Despite the model’s relevance, Malik et al. (2005) contend that in the absence of a regime shift, the GARCH (1, 1) model may overestimate volatility persistence. In order to have a robust finding, a Markov switching model has also been used to supplement the GARCH (1, 1) model. A Markov switching model is very useful in modelling the time-varying behaviour of volatility in financial markets (Mike et al., 1998). These time-varying behaviours of volatility may be subject to market shocks that create structural breaks (Ndako, 2012). In the context of this study, the Covid-19 pandemic may have caused ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C = ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C 87 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 market shocks that will affect the long memory in financial markets. The switching mechanism in the Markov model captures the complex volatility pattern by dating the breaking points, which is not possible in other models. Failure to incorporate these structural breaks can result in misspecifications. A Markov switching model for a given parameter is given in (2): (2) Where are mean zero random variables (Kuan, 2012). Most importantly the parameters are the delay parameters in which their probabilities will also indicate whether the current volatility will persist, hence complementing the GARCH (1, 1) model. The study used a sample of six major international financial markets, namely the Johannesburg stock exchange (JSE), the Standard and Poor 500 index (S&P 500), the Nasdaq Index, Shanghai Stock exchange (SSE), the French Stock Market index (CAC 40) and the German blue chip companies trading on the Frankfurt Stock Exchange (DAX). This is because these markets are among the largest financial markets in each continent around the world. The sample period was from January 1, 2020 to December 31, 2021, which was the crux of intense volatility in financial markets due to the Covid-19 pandemic. RESULTS Tables 2 and 3 present the results of the descriptive statistics and heteroscedasticity test, respectively. The R square values in Table 2 range from 1 percent to 19 percent in the financial markets under consideration. Most importantly, the Durbin-Watson statistics values are between 1.96 to 2.13, indicating the absence of autocorrelation (Kenton, 2021). The F-statistics p-values for the JSE, S&P 500, Nasdaq, CAC 40, and DAX, are significant at 5 percent, which indicate that the variance of the returns is not constant. This may signal the presence of volatility clustering. From Table 4, it can be seen that all conditions for the stability test are satisfied because the coefficients of the conditional variance are between 0 and 1, and the sum of the ARCH and GARCH coefficients is less than 1 (Bera & Higgins, 1993). The average return of the S&P 500 and Nasdaq index ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C 88 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 is positive and statistically significant at 5 percent. The past value of the Nasdaq is significant at 5 percent, meaning that the past returns in the Nasdaq can be used as a gauge for future returns. These results are also evident using the CAC 40 and DAX index with significant past value returns. The past value returns of the Nasdaq, CAC 40, and DAX have a very strong predictive ability because of the coefficients of their past value returns, and the ARCH term and GARCH term are all significant at 5 percent, which is congruent with the study of Nguyen et al. (2020). Table 2 Descriptive Statistics R Square Adjusted R Square Mean Dependent Variable Durbin-Watson Statistics JSE 0.01 0.01 0.032% 2.02 S&P 500 0.08 0.08 0.025% 2.13 Nasdaq 0.19 0.19 0.029% 2.08 SSE 0.01 0.01 0.025% 1.96 CAC 40 0.04 0.04 0.025% 2.13 DAX 0.02 0.02 0.026% 2.07 Table 3 Heteroscedasticity Test: ARCH F-Statistics p-value (F-Statistics) p-value (Chi square-statistics) JSE 3.05 0.0482* 0.0482* S&P 500 22.95 0.000* 0.000* Nasdaq 57.76 0.000* 0.000* SSE 2.7 0.1006 0.1002 CAC 40 10.85 0.000* 0.000* DAX 4.96 0.0073* 0.0075* The results from Table 4 also indicate that volatility will persist in the JSE, S&P 500, Nasdaq Index, SSE, CAC 40 and the DAX, as the sum 89 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 of their ARCH and GARCH coefficients are significant at 5 percent and close to 1. Table 4 also indicates that the decaying rate of volatility in the JSE, S&P 500, Nasdaq, SSE, CAC 40 and DAX are 0.02, 0.04, 0.04, 0.09, 0.05, and 0.03 respectively. Furthermore, the GARCH coefficients are greater than the ARCH coefficients, confirming that volatility will persist in all financial markets under consideration. Table 4 GARCH (1.1) Results Average return Value of past average return ARCH coefficient GARCH coefficient Sum of ARCH and GARCH coefficient Decaying rate of volatility (1-sum of ARCH & GARCH coefficient) JSE -0.0003 (0.71) -0.028 (0.52) 0.05 (0.0001)* 0.93 (0.000)* 0.98 0.02 S&P 500 0.001 (0.00)* 0.02 (0.41) 0.27 (0.00)* 0.69 (0.00)* 0.96 0.04 Nasdaq 0.002 (0.007)* -0.12 (0.02)* 0.19 (0.00)* 0.77 (0.00)* 0.96 0.04 SSE 0.01 (0.327) 0.01 (0.86) 0.19 (0.00)* 0.72 (0.00)* 0.91 0.09 CAC 40 0.001 (0.088) -0.099 (0.04)* 0.19 (0.00)* 0.76 (0.00)* 0.95 0.05 DAX 0.001 (0.0815) -0.109 (0.05)* 0.16 (0.00)* 0.81 (0.00)* 0.97 0.03 Note. *Significant at 5% Regarding the Markov switching output in Table 5, sigma is significant in all financial markets for both regime 1 and regime 2. Moreover, the p-values of the delay parameters () are all significant at 5 percent, confirming the GARCH (1, 1) output results. The two robust findings confirm the persistence of volatility in the sample financial markets. As already documented in the studies by Ibrahim et al. (2020); Gherghina et al. (2021); Endri et al. (2021); Enow (2023); Rahman et al. (2021); and Topcu et al. (2021), markets will remain volatile with the arrival of positive and negative news. 90 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 Table 5 Markov Switching Model Results JSE Variable Coefficient Standard Error Z-statistics P-value Regime 1 C 0.001487 0.002638 0.563682 0.5730 Log(Sigma) -3.577146 0.115888 -30.86714 0.0000* Regime 2 C -0.000538 0.000772 -0.696638 0.4860 Log(Sigma) -4.439066 0.156048 -28.44678 0.0000* Transition Matrix Parameters 1.441298 0.660442 2.182323 0.0291* -2.498119 1.065777 -2.343942 0.0191* S&P 500 Variable Coefficient Standard Error Z-statistics P-value Regime 1 C 0.001591 0.000444 3.579874 0.0003 Log(Sigma) -4.694589 0.039875 -117.7320 0.0000* Regime 2 C -0.011364 0.008663 -1.311830 0.1896 Log(Sigma) -2.794207 0.106898 -26.13894 0.0000* Transition Matrix Parameters 4.560793 0.532895 8.558519 0.0000* -2.475898 0.594259 -4.166362 0.0000* Nasdaq Variable Coefficient Standard Error Z-statistics P-value Regime 1 C -0.002542 0.004119 -0.617204 0.5371 Log(Sigma) -3.309863 0.090075 -36.74569 0.0000* Regime 2 C 0.001989 0.000535 3.720936 0.0002 Log(Sigma) -4.545140 0.039892 -113.9360 0.0000* Transition Matrix Parameters 2.466291 0.472654 5.217963 0.0000* -4.212685 0.476894 -8.833580 0.0000* ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C (continued) 91 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 SSE Variable Coefficient Standard Error Z-statistics P-value Regime 1 C -0.002288 0.002817 -0.812123 0.4167 Log(Sigma) -3.842527 0.100703 -38.15694 0.0000* Regime 2 C 0.000825 0.000429 1.925522 0.0542 Log(Sigma) -4.793461 0.039309 -121.9440 0.0000* Transition Matrix Parameters 2.258948 0.506996 4.455559 0.0000* -4.201432 0.509416 -8.247548 0.0000* CAC 40 Variable Coefficient Standard Error Z-statistics P-value Regime 1 C -0.001 0.003 -0.359 0.720 Log(Sigma) -3.517 0.074 -47.325 0.000* Regime 2 C 0.001 0.000 1.911 0.056 Log(Sigma) -4.737 0.041 -115.533 0.000* Transition Matrix Parameters 3.106 0.539 5.768 0.000* -4.550 0.562 -8.103 0.000* DAX Variable Coefficient Standard Error Z-statistics P-value Regime 1 C -0.000763 0.002638 -0.289369 0.7723 Log(Sigma) -3.553676 0.072844 -48.78471 0.0000* Regime 2 C 0.000854 0.000463 1.843283 0.0653 Log(Sigma) -4.737073 0.052580 -90.09305 0.0000* Transition Matrix Parameters 2.708401 0.529242 5.117515 0.0000* -3.958786 0.577152 -6.859174 0.0000* Note. *Significant at 5% ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C ℎ𝑡𝑡𝑡𝑡 = α + ϕℎ𝑡𝑡𝑡𝑡−1 + β𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 (1) Bollerslev (1986) ℎ𝑡𝑡𝑡𝑡 α = error term ϕ = ARCH term ℎ𝑡𝑡𝑡𝑡−1 = Lag value of Conditional variance β = GARCH coefficient 𝜇𝜇𝜇𝜇𝑡𝑡𝑡𝑡−12 = Lag square error term r 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 𝑍𝑍𝑍𝑍𝑡𝑡𝑡𝑡 = � 𝛼𝛼𝛼𝛼0 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝛼𝛼𝛼𝛼0 + 𝛼𝛼𝛼𝛼1 + 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡−1 + 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 � 𝛽𝛽𝛽𝛽𝑧𝑧𝑧𝑧𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜀𝜀𝜀𝜀𝑡𝑡𝑡𝑡 𝜌𝜌𝜌𝜌11 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝜌𝜌𝜌𝜌21 𝜌𝜌𝜌𝜌11-C 𝜌𝜌𝜌𝜌21-C 92 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 CONCLUSION Using the GARCH model, the aim of this study was to investigate the extent to which volatility will persist in the international financial world after the Covid-19 pandemic. The results indicate that volatility will persist in the financial markets under consideration. These findings are reliable because they appear to corroborate the results presented in Nguyen et al. (2022), the proposition that financial markets in developed countries have stronger long-term memory than less developed financial markets. In addition, the effects of past volatility in the Nasdaq, CAC 40, and DAX will remain in the forecast of variance due to the significant positive ARCH and GARCH coefficients. Moreover, a small number of market participants in the Nasdaq, CAC 40 and DAX may influence the stock price movements in either direction within a short period. As alluded to by Pereira and Zhang (2010), persistent volatility will affect the market volatility in the bond and equity markets. More specifically, one would expect to see a decrease in demand in the order-driven markets accompanied by wider bid-ask spreads. There might also be a decline in market depth in the sovereign bond markets. The number of corporate credit instruments may not match the corresponding increase in trading volume, which will increase the cost of providing liquidity. Investors and market participants should focus on a diversified investment approach in this market so that risks are not concentrated on any one sector or asset class. Additionally, a broader strategy of executing trade is highly recommended because financial markets will experience an increase in irrational behaviour because of a lack of confidence in the market. In short, financial markets are experiencing metastasis and therefore, market participants and investors should expect agitations, as well as persistent and volatile financial markets (Mohamed, 2022). ACKNOWLEDGMENT This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. REFERENCES Ackert, L. F., Church, B. K., & Deaves, R. (2003). Emotion and financial markets. Economic Review (Federal Reserve Bank of Atlanta), 88(Q2), 33-41. 93 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 Gherghina, S. C., Armeanu, D. S., & Joldes, C. C. (2021). COVID-19 pandemic and Romanian stock market volatility: A GARCH Approach. Journal of Risk and Financial Management 14(341), 1-29. Bauwens, L., Laurent, S., & Rombouts, J. V. K. (2006). Multivariate GARCH models: A survey. Journal of Applied Econometrics, 21(1), 79–109. Bera, A., & Higgins, M. (1993). ARCH models: Properties, estimation and testing. Journal of Economics Survey, 7, 305-366. Bobeică, G., & Bojeşteanu, E. (2008). Long memory in volatility. An investigation on the central and eastern European exchange rates. European Research Studies, 6(4), 8-18. Bollerslev, T. (1986). Generalized autoregressive conditional heterosedasticity. Journal of Econometrics, 31, 307-327. Bookstaber, R. M., & Pomerantz, S. (1989). An information-based model of market volatility. Financial Analysts Journal, 45(6), 37-46. Camp, K. M., Mead, D., Reed, S. B., Sitter, C., & Wasilewski, D. (2020). From the barrel to the pump: The impact of the COVID-19 pandemic on prices for petroleum products. [online] https://www.bls.gov/opub/mlr/2020/article/from-the-barrel-to- the-pump.htm Carpenter, J. W. (2022). The 5 biggest Russian oil companies. [online] https://www.investopedia.com/articles/markets/100515/5- biggest-russian-oil-companies.asp. Christiansen, C., Schmeling, M., & Schrimpf, A. (2012). A comprehensive look at financial volatility prediction by economic variables. Bank of international settlement working paper 374. [online]. https://www.bis.org/publ/work374.pdf Degiannakis, S. A., Filis, G., & Kizys, R. (2014). The effects of oil price shocks on stock market volatility: Evidence from European data. The Energy Journal, 35(1), 35-56. Dow, J., & Han, J. (2018). The paradox of financial fire sales: The role of arbitrage capital in determining liquidity. The Journal of Finance, 73(1), 229-274. http://www.jstor.org/ stable/26653275. Elgammal, M. M., Ahmed, W., & Alshami, A. (2021). Price and volatility spill overs between global equity, gold, and energy markets prior to and during the Covid-19 pandemic. Resources Policy, 74, 102334. https://doi.org/10.1016/j. resourpol.2021.102334. 94 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 Engle, R. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica, 50(1), 987-1007. Endri, E., Aipama, W., Razak, A., Sari, L., & Septiano, R. (2021). Stock price volatility during the COVID-19 pandemic: The GARCH model. Investment Management and Financial Innovations, 18(4), 12-20. https://doi.org/10.21511/imfi.18(4).2021.02. Engle, R., & Rangel, G. (2006). The Spline GARCH model for unconditional volatility and its global macroeconomic causes. Review of Financial Studies, 21(3), 1-42. Enow, S. T. (2023). Forecasting volatility in international financial markets. International Journal of Research in Business and Social Science, 12(2), 197-203. Gherghina, S. C, Armeanu, D. S., & Joldes, C. C. (2021). Covid-19 pandemic and Romanian stock market volatility: A GARCH approach. Journal of Risk and Financial Management, 14(341), 1-29. https://doi.org/10.3390/jrfm14080341. Gyamfi, E. N., Kyei, K. A., & Gill, R. (2016). Long memory in asset returns and volatility: Evidence from West Africa. Investment Management and Financial Innovations, 13(2), 24-28. Hathorn, D. S. (2021). Ranking the biggest factors affecting the performance of the CAC40. [online].https://www.dailyfx.com/ cac-40/factors-affecting-the-performance-of-the-CAC40.html Ibrahim, I., Kamaludin, K., & Sundarasen, S. (2020). Covid-19, Government response, and market volatility: Evidence from the Asia-Pacific developed and developing markets. Economies, 8(105), 1-22. Karanasos, M., Paraskevopoulos, A. G., Ali, F. M., Karoglou, M., & Yanti, S. (2014). Modelling stock volatilities during financial crises: A time varying coefficient approach. Journal of Empirical Finance, 29, 113–28. Kenton, W. (2021). Durbin Watson statistic definition. [online]. https:// www.investopedia .com/terms/d/durbin-watson-statistic.asp Krichene, N. (2003). Modelling stochastic volatility with application to stock returns. IMF Working Papers 03(125). Kuan, C. M. (2002). Lecture on the Markov switching model. Institute of Economics Academia Sinica, 8(15), 1-30. Kundu, S., & Paul, A. (2022). Effect of economic policy uncertainty on stock market return and volatility under heterogeneous market characteristics. International Review of Economics & Finance, 80, 597-612. 95 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 Liang, J. N. (2020). Corporate bond market dysfunction during covid-19 and lessons from the Fed’s response. [online] https://www.brookings.edu/research/corporate-bond-market- dysfunction-during-covid-19-and-lessons-from-the-feds- response/ Li, Y. (2020). This was the fastest 30% sell-off ever, exceeding the pace of declines during the great depression. [online] https:// www.cnbc.com/2020/03/23/this-was-the-fastest-30percent- stock-market-decline-ever.html Maheu, J. M., & McCurdy, T. H. (2000). Identifying bull and bear markets in stock returns. Journal of Business & Economic Statistics, 18(1), 100-112. https://www.jstor.org/stable/1392140 Malik, F., Ewing, B. T., & Payne, J. E. (2005). Measuring volatility persistence in the presence of sudden changes in the variance of Canadian stock returns. The Canadian Journal of Economics/ Revue Canadienne d’Economique, 38(3), 1037-1056. http:// www.jstor.org/stable/3696071. Mayhew, S. (1995). Implied volatility. Financial Analysts Journal, 51(4), 8-20. http://www.jstor.org/stable/4479853. Mike K. P. So., Lam, K., & Li, W. K. (1998). A stochastic volatility model with Markov switching. Journal of Business & Economic Statistics, 16(2), 244-253. https://doi.org/10.2307/1392580. Mohamed, T. (2022). 6 experts warn markets are at breaking point – and the financial system may be starting to crack. [Online] https://www.businessinsider.co.za/stocks-bonds-dollar- market-experts-financial-system-credit-suisse-banks-2022-10. Mishra, P. K., & Mishra, S. K. (2020). Corona pandemic and stock market behaviour: Empirical insights from selected Asian Countries. Millennial Asia, 11(3), 341-365, Muguto, L., & Muzindutsi, P. F. (2022). A comparative analysis of the nature of stock return volatility in BRICS and G7 Markets. Journal of Risk and Financial Management, 15(2), 1-27 https:// doi.org/10.3390/jrfm15020085. Ndako, U. B. (2012). Financial liberalization, structural breaks and stock market volatility: Evidence from South Africa. Applied Financial Economics, 22(15), 1259-1273. Nelson, D. B. (1990). Stationarity and persistence in the GARCH (1, 1) Model. Econometric Theory, 6(3), 318-334. Nguyen, D. B. B., Prokopczuk, M., & Sibbertsen, P. (2020). The memory of stock return volatility: Asset pricing implications. Journal of Financial Markets, 47, 100487. 96 The International Journal of Banking and Finance, Vol. 18, Number 2 (July) 2023, pp: 79–96 Oh, G., Kim, S., & Eom, C. (2008). Long-term memory and volatility clustering in high-frequency price changes. Physica A: Statistical Mechanics and its Applications, 387(5-6), 1247- 1254. Paraschiv, F., Reese, S. M., & Ringkjøb Skjelstad, M. (2020). Portfolio stress testing applied to commodity futures. Computer Management Sciences, 17, 203–240. Pereira, J. P., & Zhang, H. H. (2010). Stock returns and the volatility of liquidity. The Journal of Financial and Quantitative Analysis, 45(4), 1077–1110. Rahman, M. M., Guotai, C., Gupta, A. D., Hossain, M., & Abedin, M. Z. (2021): Impact of early Covid-19 pandemic on the US and European stock markets and volatility forecasting. Economic Research-Ekonomska Istraživanja, 1-19. https://doi.org/10.108 0/1331677X.2021.1997626. Ruilova, J. C., & Morettin, P. A. (2020). Parsimonious heterogeneous ARCH models for high frequency modeling. Journal of Risk and Financial Management, 13(38), 1-19. Shang, Y., Li, H., & Zhang, R. (2021). Effects of pandemic outbreak on economies: Evidence from business history context. Front. Public Health, 9(632043), 1-12. Thupayagale, P. (2012). Long memory in the volatility of local currency bond markets: Evidence from Hong Kong, Mexico and South Africa. In (Ed.), Risk management - Current issues and challenges. IntechOpen. https://doi.org/10.5772/50888. Topcu, M., Yagl, I., & Emirmahmutoglu, F. (2021). COVID-19 and stock market volatility: A time-varying perspective. Economic Bulletin, 41(3), 1-11. Visaltanachoti, N., & Pukthuanthong-Le, K. (2009). Idiosyncratic volatility and stock returns: A cross country analysis. Applied Financial Economics, 19(16), 1269-1281. Wang, J., & Yang, M. (2017). Conditional volatility persistence. [online] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3080693