BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

191 

The Impact of COVID-19 Pandemic on the Social Media 
Industry: A Long-term Perspective 

Zijing Chen* 

Department of Statistical Science, University College London, London, WC1E 6BT, United 
Kingdom 

*Corresponding author: zijing.chen.20@ucl.ac.uk 

Abstract. With the emergence of the coronavirus pandemic around the world, social media has 
rapidly become an effective tool for information generation and broadcast, entertainment, marketing, 
and consumption. Depending on some studies, social distancing rules, lockdown measures, 
business closures, self-quarantines, and the fear of infection during the COVID-19 pandemic limited 
people's physical social interaction; as a result, individual and organizational activities on social 
media platforms grow explosively. This degree of social media popularity has a double-edged impact 
on the expansion and financial performance of social media companies. In this article, a well-known 
American corporation, Twitter, lnc. was chosen and daily stock data from January 2020 to April 2022 
is extracted, as well as corresponding daily new confirmed cases in the US and around the world. 
The VAR model was applied in the study to evaluate the link between variables, and the ARMA-
GARCH model was used to determine and analyze both earning ability and stock volatility throughout 
the pandemic. Interestingly, Twitter's financial performance is merely a microcosm of the entire stock 
market, in which investors can quickly rebound from pessimism and reinvest in the market. The study 
forecasts the possible future of the social media industry and provides managerial and investment 
recommendations to its stakeholders. 

Keywords: COVID-19 pandemic; Social media industry; Stock returns; Stock return volatility; Twitter. 

1. Introduction 

The National Science Foundation Network (NSFNET), the basis of modern internet and social 

media, debuted in 1987. This marked the beginning of the social media network's growth [1]. Since 

the internet's development has expedited the introduction of digital communication and social media 

networks, there have been some significant milestones in the internet's rapid expansion and rise. 

Social media global penetration has reached 60% in 2022, with global social media audiences totaling 

4.59 billion users, and it is expected to rise steadily over the next five years [2]. Based on decades of 

research, social media sites and apps have become some of the most essential instruments in this 

digital era, and they may be regarded a part of people's lives. People use platforms like as YouTube, 

Facebook, Instagram, Twitter, and Sina Weibo to receive information, stay in touch with friends, fill 

spare time, and even do business. 

Because of the rapid spread of COVID-19, the globe has been in a state of panic and turmoil since 

the initial report in December 2019 became public and then brought unprecedented issues globally. 

The social media industry is not an exception; the advent of the pandemic interferes with it.  

According to the WHO's report on Social Media & COVID-19: A Global Study of Digital Crisis 

Interaction among Gen Z and Millennials, approximately 44% of respondents would prefer to post 

scientific content about COVID-19 on social media during the pandemic, rather than entertaining or 

emotional content [3]. More specifically, between April 9 and April 15, 2020, around 530000 tweets 

about CORONAVIRUS and COVID-19 were retrieved from Twitter [4]. During the early stages of 

the pandemic in China, the average number of daily posts about the pandemic on Sina Weibo was 

around 343000 [5]. These figures indicate that COVID-19 pandemic discourses have spread fast via 

social media to compensate for restrictions on physical interactions and social contact among people. 

Previous research has found close association between social media and the COVID-19, which is 

involved in politics, health care, education, and marketing. According to Nulty et al., every political 

party uses social media platforms for politics, and these channels allow those with less budgets to 



BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

192 

convey their messages inexpensively [6]. Since a list of COVID-19 pandemic hashtags appeared as 

hot topics on Twitter, 64.8% of UN member states had a leader who tweeted about COVID-19, and 

there was a considerable growth rate of Twitter followers for leading politicians who used their 

accounts frequently in comparison to pre-pandemic months [7]. President Donald Trump of the 

United States, Prime Minister Boris Johnson of the United Kingdom, WHO Director, and numerous 

European Union presidents are examples of people who use social media to update information and 

policies on vital problems [8]. In order to promote people's compliance with COVID-19 legislation, 

governments use such a new and stylish strategy to maximize the possibility of the public reading 

and learning the latest politics and information from key government sources [7]. 

Additionally, social networking networks are drawing more medical or health care organizations. 

Gaini et al., for example, claimed that the majority of US neurology residency programs started 

Twitter and Instagram profiles after March 1st, 2020 [9]. Research groups and large national and 

international studies, such as RECOVERY and the EuroELSO survey, were also aggressively 

recruiting and reporting on Twitter, Facebook, and similar social media platforms [10]. The inclusion 

of health care studies and programs causes social media to cover not just entertainment but also 

academic discoveries, thereby expanding key audience demographics. Although social media has a 

positive impact on raising domestic awareness about preventive measures [11], the amplification, 

inaccuracies, and misinformation about health [10, 12], as well as the public's relatively negative 

perspectives on aspects of social media [13], indeed contribute to trustworthiness concerns [8]. In 

other words, social media usage and activity vary according to the material presented. 

Moreover, social media products play important roles in education; for example, some schools 

decided to adapt e-learning via YouTube [14], to gain digital social support and use collaboration 

functions in Instagram [15], and to allow radiology students to share videos and make the study more 

interactive via Facebook Live [16]. These all show that powerful and multifunctional tools spark 

public participation on social media sites. 

Furthermore, many companies attempt to use social media marketing and advertising to raise brand 

awareness, attract potential consumers, meet customer demands, and achieve targeted conversion 

rates. According to Statista, companies seeking additional consumers had to pay much more to have 

their adverts seen by a thousand potential customers from 2020 to 2021 due to an increase in 

advertising space demand [17]. For example, Nike, a well-known sportswear retailer, offers online 

exercises through its own channels and social media platforms, providing athletes all over the world 

with tools, motivation, and affirmation to help fuel physical and mental health [18]. Under the strain 

of the restaurant shutdown and self-quarantine policy, some takeaway companies, including Hungry 

Panda and Deliveroo, sought to advertise on Instagram, Twitter, and YouTube. Other companies' 

exposure on social media may also motivate people to visit these channels more regularly. In terms 

of more attention and demands for advertising spaces during the pandemic, social media firms appear 

to be lucrative. 

These previous studies highlighted how COVID-19 altered people's and organizations’ social 

media platform usage behaviors and habits. The increased utilization might be attributed to more new 

contracts, the availability of new products, and a rise in stock price. In that way, it may stimulate 

stock return volatility. However, only a few studies discussed how exactly the stock volatility of the 

social media company reacts to changes in the COVID-19 pandemic. To fill the void, this paper 

examines not only how the increase in confirmed cases in the US affects a specific social media 

company Twitter in terms of stock volatility, but also how COVID-19 confirmed cases around the 

world influence a specific social media company Twitter in regard to stock volatility, given that more 

than 80% of active Twitter users are outside of the US. The specific reason for analyzing the 

relationship is to understand the fluctuation pattern of a social media firm's stock return after the 

spread of the COVID-19 pandemic, to determine the differences and similarities with the broad US 

stock market, and to provide relevant advice to investors and organizations. As a result, some 

irrational investments or strategies can be avoided. 



BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

193 

The remainder of this paper is organized as follows: Section 2 covers information about the data 

source, data stability, and the models used in this paper. Section 3 follows, with a full discussion of 

the results from the VAR model and the ARMA-GARCH model, as well as additional analysis on 

the stock return, stock volatility, and market participants' behavior. Following that, there is a 

discussion on the study's focus, objective, and importance. Finally, Section 5 reiterates the final 

conclusion briefly. 

2. Research Design  

2.1 Data Source 

Twitter is one of the most prominent social media platforms with large masses of available data; 

it can be representative and serve as an example for comprehending the social media industry's 

pandemic influence.  The study extracted daily closing stock prices of Twitter and daily total 

confirmed cases in the US and around the world from January 23, 2020 to April 13, 2022, from the 

Choice financial terminal. The rationale for halting data collection till now is to avoid the impact of 

the issue that Elon Musk made an offer to buy all of Twitter's shares [19]. Data processing is required 

to figure out the impact of increased new cases in the United States and throughout the world on 

Twitter stock return fluctuations. To elaborate, today's total confirmed cases in the US and around 

the world are subtracted from the previous day's total confirmed cases to obtain the daily growth in 

US and global new cases, and Twitter stock returns are obtained by dividing the difference between 

two days' closing stock prices by the one on the previous day, and data is transformed by the formula 

𝑙𝑛(1 + 𝑥), continuing analysis in the logarithmic scale. Because there are some total confirmed cases 

lost on specific days, the missing data are omitted during the study. With the updated edited data, 

Stata was used to analyse the data and construct models for further exploration. 

2.2 Augmented Dickey–Fuller (ADF) Unit Root Test 

Testing whether or not the data are stationary is the first step before proceeding. Based on the ADF 

testing conducted in Stata, the p-values in Table 1 for Twitter stock return and daily new confirmed 

cases in the US and the globe all equal 0, which is considered statistically significant. Due to these 

findings, there is enough evidence to reject the null hypothesis that the variable contains a unit root. 

In other words, the model built on the data is feasible and the data are stationary. 

Table 1. ADF test 

Variables t-statistic p-value 

Twitter -15.1340 0.0000*** 

New confirmed cases   

US -10.2810 0.0000*** 

Global -11.2650 0.0000*** 

   

2.3 Vector Autoregression (VAR) Model 

The use of the VAR model predates the rise of the VAR approach [20], and it can be traced back 

to a study On the Statistical Treatment of Linear Stochastic Difference Equations [21] and A Study 

of the Autoregressive Nature of the Time Series Used for Tinbergen’s Model of the Economic System 

of the United States [22] before reaching Sim's well-known contribution to VAR model application 

[23-25]. The VAR model can be employed to capture the relationship between multiple variables 

while avoiding the challenge of model construction based on rigorous economic theory [26]. 

Although a few analyses suggested that the COVID-19 announcement had an impact on the social 

media company, little research suggests that there is a strong intercorrelation between stock returns 

and rises in daily confirmed cases. In this context, the VAR model was chosen, and there are three 

separate time series variables, denoted by 𝑥𝑡,1, 𝑥𝑡,2, 𝑥𝑡,3, resulting in a trivariate VAR(p) model. 



BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

194 

𝑥𝑡,1 =  𝛼1 + ∅11𝑥𝑡−1,1 + ⋯ + ∅1𝑝𝑥𝑡−𝑝,1 + 𝛽11𝑥𝑡−1,2 + ⋯ + 𝛽1𝑝𝑥𝑡−𝑝,2 + 𝛿11𝑥𝑡−1,3 + 

… + 𝛿1𝑝𝑥𝑡−𝑝,3 + 𝑒1𝑡 .                             (1) 

𝑥𝑡,2 =  𝛼2 + ∅21𝑥𝑡−1,1 + ⋯ + ∅2𝑝𝑥𝑡−𝑝,1 + 𝛽21𝑥𝑡−1,2 + ⋯ + 𝛽2𝑝𝑥𝑡−𝑝,2 + 𝛿21𝑥𝑡−1,3 + 

… + 𝛿2𝑝𝑥𝑡−𝑝,3 + 𝑒2𝑡 .                              (2) 

𝑥𝑡,3 =  𝛼3 + ∅31𝑥𝑡−1,1 + ⋯ + ∅3𝑝𝑥𝑡−𝑝,1 + 𝛽31𝑥𝑡−1,2 + ⋯ + 𝛽3𝑝𝑥𝑡−𝑝,2 + 𝛿31𝑥𝑡−1,3 + 

… + 𝛿3𝑝𝑥𝑡−𝑝,3 + 𝑒3𝑡 .                             (3) 

[

𝑥𝑡,1

𝑥𝑡,2

𝑥𝑡,3

] =  [

𝛼1

𝛼2

𝛼3

] +  [

∅11 ⋯ ∅1𝑝

∅21 ⋯ ∅2𝑝

∅31 ⋯ ∅3𝑝

] [

𝑥𝑡−1,1

⋮
𝑥𝑡−𝑝,1

] + [

𝛽11 ⋯ 𝛽1𝑝

𝛽21 ⋯ 𝛽2𝑝

𝛽31 ⋯ 𝛽3𝑝

] [

𝑥𝑡−1,2

⋮
𝑥𝑡−𝑝,2

] + 

[

𝛿11 ⋯ 𝛿1𝑝

𝛿21 ⋯ 𝛿2𝑝

𝛿31 ⋯ 𝛿3𝑝

] [

𝑥𝑡−1,3

⋮
𝑥𝑡−𝑝,3

] + [

𝑒1𝑡

𝑒2𝑡

𝑒3𝑡

].                       (4) 

The equation (1), (2), (3) above are for Twitter stock return, daily new cases in the US, and daily 

new cases worldwide, respectively, while equation (4) is in matrix form. To clarify, in equation (1), 

𝛼1 + ∅11𝑥𝑡−1,1 + ⋯ + ∅1𝑝𝑥𝑡−𝑝,1 represents a linear function of past lags of Twitter stock return, 

while 𝛽11𝑥𝑡−1,2 + ⋯ + 𝛽1𝑝𝑥𝑡−𝑝,2 and 𝛿11𝑥𝑡−1,3 + ⋯ + 𝛿1𝑝𝑥𝑡−𝑝,3 represent past lags of daily new 

cases in the US and daily new cases in the world, 𝑒1𝑡 is the error term. As a result, the variable 

Twitter stock return is modelled using historical values for the variable and the other two variables. 

Similarly, the structures of the equations for the remaining two variables are the same, but the variable 

on the left of the equation and coefficients are changed. 

2.4 ARMA-GARCH Model 

The ARMA-GARCH model can evaluate both the return and volatility of the Twitter stock. This 

model is broken down into two sections in this paper: ARMA and GARCH. 

2.4.1 ARMA 

𝑦𝑡 =  ∅0 + ∑ ∅𝑖
𝑝
𝑖=1 𝑦𝑡−𝑖 + 𝛼𝑖 − ∑ ∅𝑖

𝑞
𝑖=1 𝛼𝑡−𝑖 .                  (5) 

The general expression of the ARMA model is displayed in equation (5). The AR(p) is represented 

by the component ∅0 + ∑ ∅𝑖
𝑝
𝑖=1 𝑦𝑡−𝑖, whereas the rest of the equation is MA(q). AR(p) estimates 

future value applying past Twitter stock returns from January 2020 to April 2022, whilst MA(q) 

forecasts using an error term. 

2.4.2 GARCH 

Then comes another section GARCH. GARCH is derived from ARCH fundamentally, and both 

treat volatility as a variance to be modelled [27]. In this paper, GARCH (1,1) was taken into account, 

where the first 1 denotes one autoregressive lag and the second 1 denotes one moving average lag. 

The reasons for selecting GARCH (1,1) are as follows: it has fewer parameters than ARCH(p); it fits 

many time series well [28], it is sufficient to capture the volatility clustering in the data mentioned in 

Introductory Econometrics for Finance ed Third [29], and it only requires one lag for analysis [29]. 

As a result, the model GARCH (1,1) was picked. 

𝜎𝑡
2 =  𝛼0,1 + 𝛼1휀𝑡−1

2 + 𝛽𝑡𝑚𝑡 + 𝛾1𝜎𝑡−1
2 .                      (6) 

In the equation (6), term 𝛼1휀𝑡−1
2  is ARCH part and 𝛾1𝜎𝑡−1

2  represents GARCH part, and the 

additional term 𝛽𝑡𝑚𝑡 other than the generalised formula represents confirmed cases that acted as an 

extra explanatory variable in the model. 



BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

195 

3. Empirical Results and Analysis 

3.1 Order of VAR Model 

To find out the optimal lag order for a VAR model, the LR statistic and other information criterion 

of each lag should be assessed. An asterisk sign (*) appears after the data to signify the desired lag 

order. 

Table 2. VAR model identification 

Lag LL LR df p FPE AIC HQIC SBIC 

0 -1199.5600    0.07185 5.8804 5.8921 5.9099 

1 -1152.3000 94.5160 9 0.0000 0.0595 5.6934 5.7399 5.8111 

2 -1138.0000 28.6020 9 0.0010 0.0580 5.6674 5.7490 5.8735 

3 -1119.3200 37.3640 9 0.0000 0.0553 5.6201 5.7366 5.91454 

4 -1097.6900 43.2480 9 0.0000 0.0520 5.5584 5.7098 5.9411 

5 -843.7530 507.8800 9 0.0000 0.0157 4.3606 4.5470 4.8317* 

6 -826.3720 34.7630 9 0.0000 0.0150 4.3196 4.5409* 4.8790 

7 -812.7910 27.1620 9 0.0010 0.0147 4.2972 4.5535 4.9449 

8 -803.5110 18.5600 9 0.0290 0.0147 4.2959 4.5871 5.0319 

9 -785.9930 35.0360* 9 0.0000 0.0141* 4.2542* 4.5804 5.0785 

10 -778.4890 15.0080 9 0.0910 0.0142 4.2615 4.6226 5.1742 

11 -777.7230 1.5312 9 0.9970 0.0148 4.3018 4.6978 5.3028 

12 -773.1160 9.2147 9 0.4180 0.0151 4.3233 4.7543 5.4126 

 

Table 2 reveals that lags 9, 6, and 5 all have that sign. A comparison of AIC differences is necessary 

to determine the optimal lag order. The difference in AICs between lags 9 and 10 is around 0.1, while 

the difference between lags 5 and 6 is approximately 0.4. As a result, lag 9 is the better option to start 

with. Furthermore, despite the fact that both FPE and AIC with an asterisk sign are smallest for lag 

9, HQIC and SBIC suggesting other alternative orders, the present optimal lag order for VAR model 

should be 9. 

Following the specification of the VAR model's order, it is crucial to examine whether or not the 

VAR model is stationary. If the VAR model is non-stationary, the impulse-response function will not 

converge to zero, indicating that daily new cases have long lasting effects on the Twitter stock return. 

After that, the model's applicability was determined by carrying out the unit root test and sketching a 

unit circle with the roots. All of the roots are clearly within the circle in Fig. 1, implying that there is 

no need to revaluate the lag order and that trivariate VAR(9) is a stable model. 

 

Fig. 1 Unit root test 



BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

196 

3.2 Impulse Response 

The impulse response function quantifies the time profile of a shock's effect on the behaviour of a 

series [30].  In this paper, investigation on how Twitter stock returns respond to new case impulses 

in the United States and around the world was done. Previous research found that a high level of 

policy uncertainty had a detrimental impact on stock market performance because investors or market 

participants were pessimistic about the expected discount rate, resulting in a reduction in stock price 

[31]. Likewise, people were more concerned about the future performance of the stock market, 

wherever they were, after a short period of the spread of COVID-19 policy, such as physical 

restrictions and travel bans, and a surge in death. These investors were terrified of the substantial 

chance of benefit loss. As a consequence, massive capital flowed out of the financial market, causing 

in a cliff drop in the financial index at the beginning of the outbreak. For example, stocks in the United 

States, Europe, and Asia fell by approximately 40%, 40%, 20% to 25%, respectively, in March 2020 

compared to January 5, 2020 [32]. The stock has been cut by nearly half on March 20 2020, as 

reflected in the US stock market index US30 (Fig. 2), provoking the panic until March 25. In such a 

short period of time, the stock recovered brilliantly, soaring over 20% above the previous low. 

Notwithstanding the deeper understanding of the COVID-19 pandemic, whether or not the changing 

pattern will continue to be like a roller coaster ride, investors' behaviour will return to normalcy, or 

the impact will be long-lasting, requires more evidence in long-term perspective. 

 

Fig. 2 US stock market (US30) fluctuation 

Note: United States Stock Market Index (US30) changes from January 24 2020 to August 31 

2020, extracted from TRADING ECONOMICS, https://tradingeconomics.com/united-

states/stock-market 

As shown in the impulse response diagram (Fig. 3), when the US daily new cases climb by 1% at 

t = 0, between t = 0 and t = 4, the Twitter stock return has a clear negative impact. However, it 

suddenly exerts the greatest positive influence on stock returns at t = 7, with a figure of around 0.3%, 

and the effects continue to be positive in general beyond t = 7. The global daily new cases have the 

same effect on the stock return. Both shocks in daily new cases have an adverse impact at first but 

revert to zero after 20 steps, rather than staying in a long-period pessimistic situation. 

Having said that, it appears that investors failed to learn from previous experiences, as the severe 

impact to the social media business had not deterred their investment. This supports Hsu and Tang's 

conclusion that the size of the effect of investor sentiment on unanticipated volatility declined later 

in the pandemic [33]. Twitter’s stock performance is almost a microcosm of the world's stock market. 

https://tradingeconomics.com/united-states/stock-market
https://tradingeconomics.com/united-states/stock-market


BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

197 

It is history repeating as the entire stock market performed in 2020, but shorter time intervals have 

been used for the recurrence. 

One thing to note is that, in the context of pandemic normalisation, investors' concern has faded 

gradually, and there are fewer repercussions on the extent of Twitter stock return. 

US 

 

Global

 

Fig. 3 Impulse and response 

3.3 ARMA Specification 

Regarding to the order of the stock return in logarithm, PACF and ACF can be helpful to derive 

the lag orders for AR(p) and MA(q). In the Fig. 4, the first part beyond the critical values is 2 for both 

PACF and ACF plots, demonstrating that AR(p) and MA(q) both have order 2 and value of p and q 

are equal to 2. 

PACF ACF 

  

Fig. 4 PACF and ACF 

3.4 ARMA-GARCH Estimation Eesults and Variance Equation 

As described in Section 2.4.2, implementing GARCH(1,1) is adequate to capture volatility 

clustering in time series, hence the ARMA-GARCH model was built. 

Fig. 5 depicts the characteristics of Twitter stock return volatility. This time series data clearly 

exhibits overt volatility clustering at the beginning, middle, and end of the periods, revealing 

conditional heteroskedasticity. However, drawing conclusions solely from the plot is invalid; it would 

be more logical and reasonable to examine other information about the statistical significance of 

conditional heteroskedasticity in the model estimation results. 



BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

198 

 

Fig. 5 Yield trend 

Table 3 contains the model estimation findings as well as the variance equation. The ARCH and 

GARCH terms in the variance equation both have p-values less than 0.05, showing that they are 

significant. The presence of conditional heteroskedasticity fulfils the principal requirement of 

GARCH model building, implying that the Twitter stock return has significant conditional 

heteroskedasticity. Both coefficients for the AR model and the MA model have p-values greater than 

0.05, meaning that they are insignificant and that there is insufficient evidence to reject the null 

hypothesis that the coefficients should be 0. The growth of daily new instances in the United States 

or around the world, in particular, will have no substantial impact on the volatility of Twitter stock 

returns. 

Table 3. ARMA-GARCH estimation results 

 (1) (2) 

 Coefficient Std. err p>|Z| Coefficient Std. err p>|Z| 

Mean equation 

AR, L2 0.0010 0.3891 0.9980 0.0052 0.3930 0.9890 

MA, L2 0.1293 0.3867 0.7380 0.1257 0.3900 0.7470 

Constant 0.0013 0.0017 0.4330 0.0013 0.0017 0.4570 

Variance equation 

New confirmed cases       

US -0.0304 0.0159 0.0560    

Global    -0.0031 0.0149 0.8350 

GARCH (1, 1)       

ARCH, L1 0.2393 0.0433 0.0000 0.2361 0.0441 0.0000 

GARCH, L1 0.3870 0.1549 0.0130 0.3439 0.1613 0.0330 

Constant -7.3199 0.3367 0.0000 -7.4666 0.3199 0.0000 

4. Discussion 

In comparison to other studies, this paper focuses on how COVID-19 US and global new cases 

affect the stock return and stock return volatility of an American social media company, whereas 

existing articles discuss the impact of the COVID-19 pandemic on mental health, living habits, 

interactive behaviors, information dissemination or infodemic [34] on social media. Although some 

articles investigate stock market performance during the pandemic, their conclusions and topics are 

more related to big industries such as communication services, technology, healthcare, energy, and 

so on. However, one similarity between us in analyzing stock market volatility is that many authors 



BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

199 

adopted methodology or related transformations such as VAR model, GARCH model, TGARCH 

model [35], EGARCH model [33], or others in their research to determine stock market movements. 

Through this paper, managers of social media companies could pay more attention to designing 

contingency strategies for the emergence of potential black swan events in order to avoid the creation 

of a bad atmosphere, as well as taking a broad view to judge the likely duration or possible 

normalization of events, so that public sentiment on social media is under control and can generate 

much more value than just misinformation exaggeration. Further, supervise stakeholders and market 

participants with social media accounts, such as policymakers or organizations, taking responsibility 

to regulate and assure the direction of the discussion is vital, especially in the case of an accident. 

In the digital age, social media, as a multitasking tool, typically moves swiftly. The COVID-19 

pandemic is only a temporary setback for this industry, and it serves as a reminder for social media 

companies to improve the current functions and characteristics of their products. According to the 

findings of this study, it is important for investors who are concerned about social media stocks to 

forecast the industry's stability and the durability of event impacts on the social media business. 

Therefore, investors need keep an eye on the latest news in order to modify their short-term 

investment strategies. 

5. Conclusion 

The objective of this study is to look into how new cases of the COVID-19 pandemic correlate to 

the social media industry in terms of usage, stock return, and stock volatility, with an emphasis on 

the American social media company Twitter. VAR and ARMA-GARCH models are introduced for 

this purpose, with the VAR model exploring impulse response and the ARMA-GARCH model 

assessing stock returns and conditional variances. The study leads to a conclusion after conducting 

empirical investigation. 

Finally, this article demonstrates that the public's unexpected boost in use of social media sites 

during the pandemic does not indicate a continual development or fall of an industry, but rather a 

short-term shock. Despite the fact that social media penetration and user engagement are increasing 

every single year, the detrimental impact of the COVID-19 pandemic was only apparent at the 

beginning. The impact on the long-term future of the social media stock return and its volatility will 

eventually diminish and return to normal, following the general trend of the stock market. 

References 

[1] Harris S R, Gerich E. Retiring the NSFNET backbone service: Chronicling the end of an era. ConneXions, 

1996, 10(4): 2-11. 

[2] S. Dixon. Global social network penetration 2018-2027. (2022-6-15) [2022-9-4]. 

https://www.statista.com/statistics/260811/social-network-penetration-worldwide/. 

[3] Volkmer I. Social media and COVID-19: A global study of digital crisis interaction among Gen Z and 

millennials. World health organisation, 2021. 

[4] Manguri K H, Ramadhan R N, Amin P R M. Twitter sentiment analysis on worldwide COVID-19 

outbreaks. Kurdistan Journal of Applied Research, 2020: 54-65. 

[5] Guo S, Fang F, Zhou T, et al. Improving Google flu trends for COVID-19 estimates using Weibo posts. 

Data Science and Management, 2021, 3: 13-21. 

[6] Nulty P, Theocharis Y, Popa S A, et al. Social media and political communication in the 2014 elections 

to the European Parliament. Electoral studies, 2016, 44: 429-444. 

[7] Haman M. The use of Twitter by state leaders and its impact on the public during the COVID-19 pandemic. 

Heliyon, 2020, 6(11): e05540. 

[8] Kaya T. The changes in the effects of social media use of Cypriots due to COVID-19 pandemic. 

Technology in society, 2020, 63: 101380. 

https://www.statista.com/statistics/260811/social-network-penetration-worldwide/


BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

200 

[9] Gaini R R, Patel K M, Khan S A, et al. A rise in social media utilization by US neurology residency 

programs in the era of COVID-19. Clinical Neurology and Neurosurgery, 2021, 207: 106717. 

[10] Wong A, Ho S, Olusanya O, et al. The use of social media and online communications in times of 

pandemic COVID-19. Journal of the Intensive Care Society, 2021, 22(3): 255-260. 

[11] Sharif N, Opu R R, Alzahrani K J, et al. The positive impact of social media on health behavior towards 

the COVID-19 pandemic in Bangladesh: A web-based cross-sectional study. Diabetes & Metabolic 

Syndrome: Clinical Research & Reviews, 2021, 15(5): 102206. 

[12] Wang Y, McKee M, Torbica A, et al. Systematic literature review on the spread of health-related 

misinformation on social media. Social science & medicine, 2019, 240: 112552. 

[13] Auxier B, Anderson M. Social media use in 2021. Pew Research Center, 2021, 1: 1-4. 

[14] Elareshi M, Habes M, Youssef E, et al. SEM-ANN-based approach to understanding students’ academic-

performance adoption of YouTube for learning during Covid. Heliyon, 2022, 8(4): e09236. 

[15] Richter E, Carpenter J P, Meyer A, et al. Instagram as a platform for teacher collaboration and digital 

social support. 2022. 

[16] Kauffman L, Weisberg E M, Fishman E K. Using Facebook Live and Zoom as tools for online radiology 

education: A practical guide. Current Problems in Diagnostic Radiology, 2022. 

[17] Statista Research Department. Social media ads CPM worldwide 2020-2021. (2021-10-27)[2022-9-5]. 

https://www.statista.com/statistics/873631/social-media-advertising-cpm/. 

[18] NIKE, Inc. COVID-19 Response Efforts. (2021-5-14) [2022-9-5]. 

https://about.nike.com/en/newsroom/statements/covid-19-response-efforts. 

[19] CNN. Elon Musk's bumpy road to possibly owning Twitter: A timeline. (2022-8-19)[2022-9-6]. 

https://edition.cnn.com/2022/05/17/tech/twitter-elon-musk-timeline/index.html. 

[20] Qin D. Rise of VAR modelling approach. Journal of Economic Surveys, 2011, 25(1): 156-174. 

[21] Mann H B, Wald A. On the statistical treatment of linear stochastic difference equations. Econometrica, 

Journal of the Econometric Society, 1943: 173-220. 

[22] Orcutt G H, Irwin J O. A study of the autoregressive nature of the time series used for Tinbergen's model 

of the economic system of the United States, 1919-1932. Journal of the Royal Statistical Society. Series 

B (Methodological), 1948, 10(1): 1-53. 

[23] Sims C A. The role of approximate prior restrictions in distributed lag estimation. Journal of the American 

Statistical Association, 1972, 67(337): 169-175. 

[24] Sims, C. A. Macroeconomics and Reality. Econometrica, 1980, 48(1), 1. 

[25] Sims C A. Empirical analysis of macroeconomic time series: VAR and structural models: by Michael P. 

Clements and Grayham E. Mizon. European Economic Review, 1991, 35(4): 922-932. 

[26] Freeman J R, Williams J T, Lin T. Vector autoregression and the study of politics. American Journal of 

Political Science, 1989: 842-877. 

[27] Engle R. GARCH 101: The use of ARCH/GARCH models in applied econometrics. Journal of economic 

perspectives, 2001, 15(4): 157-168. 

[28] Hill R C, Griffiths W E, Lim G C. Principles of econometrics. John Wiley & Sons, 2018. 

[29] Ghani I M M, Rahim H A. Modeling and forecasting of volatility using arma-garch: Case study on 

malaysia natural rubber prices. IOP Conference Series: Materials Science and Engineering. IOP 

Publishing, 2019, 548(1): 012023. 

[30] Koop G, Pesaran M H, Potter S M. Impulse response analysis in nonlinear multivariate models. Journal 

of econometrics, 1996, 74(1): 119-147. 

[31] Tiwari A K, Abakah E J A, Karikari N K, et al. The outbreak of COVID-19 and stock market liquidity: 

Evidence from emerging and developed equity markets. The North American Journal of Economics and 

Finance, 2022, 62: 101735. 

[32] Statista Research Department. Effect of coronavirus on major global stock indices 2020-2021. (2022-1-

11) [2022-9-5]. https://www.statista.com/statistics/1251618/effect-coronavirus-major-global-stock-

indices/. 

https://www.statista.com/statistics/873631/social-media-advertising-cpm/
https://about.nike.com/en/newsroom/statements/covid-19-response-efforts
https://edition.cnn.com/2022/05/17/tech/twitter-elon-musk-timeline/index.html
https://www.statista.com/statistics/1251618/effect-coronavirus-major-global-stock-indices/
https://www.statista.com/statistics/1251618/effect-coronavirus-major-global-stock-indices/


BCP Education & Psychology CEPC 2022 

Volume 8 (2023)  

 

201 

[33] Hsu Y L, Tang L. Effects of investor sentiment and country governance on unexpected conditional 

volatility during the COVID-19 pandemic: Evidence from global stock markets. International Review of 

Financial Analysis, 2022, 82: 102186. 

[34] World Health Organization. Infodemic. [2022-9-6].  https://www.who.int/health-

topics/infodemic#tab=tab_1. 

[35] Lúcio F, Caiado J. COVID-19 and Stock Market Volatility: A Clustering Approach for S&P 500 Industry 

Indices. Finance Research Letters, 2022, 49: 103141. 

 

https://www.who.int/health-topics/infodemic#tab=tab_1
https://www.who.int/health-topics/infodemic#tab=tab_1