The Correlation Between Social Media and Consumer Behaviour: A

Research on Movie Industry in China

Enjun Long

Hwa Chong International School, 269783 Bukit Timah, Singapore

Keywords: Social Media, Movie Box Office, Consumer Behavior, Movie Industry in China

Abstract: The large domestic coverage of social media makes it significantly influential from a wide perspective. In the

recent decade, social media's role has become increasingly important in the commercial world. Over time,

social media platforms have transformed into persuasive tools that significantly influence how consumers

explore, evaluate, and purchase products. This paper embarks on an exploration of the intricate interplay

between social media and the movie box office, employing quantitative research methodologies to examine

the impact of social media on the movie box office. In this study, the impact of social media on movie sales

performance will be analyzed from two perspectives: The correlation between social media (as alternative

evaluation sources) and the total movie box office; and the correlation between social media’s engagement

indexes (reflected by different social media indexes) and movie box office daily. A significant positive

relationship has been found in both two perspectives. The research provides new findings for marketers in the

movie industry to better understand the factors influencing movie sales performance.

1 INTRODUCTION

1.1 The Contemporary Social Media

Usage and Influence

According to the DataReportal, at the beginning of

2023, there were 1.05 billion internet users in China,

accounting for 73.7% of the total domestic population.

Among the internet users, 98.1% of them were social

media users (Kemp 2023). In addition, a growing

trend of social media usage was foreseen in the next

few years. The large domestic coverage of social

media makes it significantly influential from a wide

perspective. In the recent decade, social media's role

has become increasingly important in the commercial

world. Unlike the traditional way of one-way

advertising, two-way communication on social media

allows consumers to have peer interactions (Shareef et

al. 2019). Social media functions such as like,

comment, share, and subscribe enable consumers a

much wider communication and stronger connection

with each other. Early in 1999, it was proven that an

informational society could effectively affect product

evaluation in the buying process (Kozinets 1999).

Over the years, social media has been advocated to be

effective in influencing consumer’s perceptions and

behavior (Chopra et al. 2021, Nash 2019). With the

expansion of social media networks and the maturity

of individual opinion dissemination, social media

marketing has been taken as a strategic approach by

merchants to achieve marketing objectives and boost

sales performance (Wibowo et al. 2020).

1.2 The Movie Industry and Social

Media

As one of the most popular entertainment options,

movies have become a great portion of the media

products consumed by people (Kubrak 2020). It

provides a two-hour stress-free time and environment

for audiences to sit back, and relax, which benefits

one’s mental and emotional health (Molaie et al.

2010). According to China Movie Database, by 13

November 2023, the movie industry sales

performance has topped 50 billion yuan for 2023

(China Film Data Information Network 2023). The

industry witnessed a year-on-year growth rate of

52.88% by the first half of 2023, demonstrating a

good recovery from the pandemic impact and a

vibrant market demand (Chen 2023).

Marketing, specifically social media marketing,

has been playing an important role in driving movie

sales. Industry studies proved that movie sales

performance is significantly affected by the Internet

Word of Mouth (IWOM) (Zhu & Ma 2021). Asur and

260

Long, E.

The Correlation Between Social Media and Consumer Behaviour: A Research on Movie Industry in China.

DOI: 10.5220/0012820000004547

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Data Science and Engineering (ICDSE 2024), pages 260-266

ISBN: 978-989-758-690-3

Huberman also showed evidence that box-office

revenues for movies could be predicted with social

media via a specific analysis of Twitter comments

(Asur & Huberman 2021). Similarly, the movie box

office was proven to be predictive based on the

Facebook comments from the research done by

Matthias, Michel, Drik, and Asil (Bogaert et al.

2021). Nowadays, as the industry scale grows in

China, it becomes a burgeoning interest to unlock the

relationship between social media and movie sales

performance from different perspectives. However,

relevant research in the Chinese market is

comparatively little.

1.3 Research Background

Inspired by the increasing impact of social media on

consumer purchase decisions and realizing most of

the previous research was done based in Europe and

America, or on a single social media platform, this

paper aims to study the relationship between social

media and consumer behavior in the movie industry

in China. The consumer behavior model adopted in

this paper is the Engel-Blackwell-Kollat Model, of

which consumer behavior could be explained in five

steps: 1) problem recognition; 2) information search;

3) evaluation of alternatives; 4) purchase; 5) post-

purchase evaluation (Engel et al. 1995). Two main

perspectives have been explored through the

following angles:

The correlation between social media (as

alternative evaluation sources) and the total movie

box office.

The correlation between social media engagement

(reflected by different social media indexes) and

movie box office daily.

Through studying the two perspectives, this paper

enhances the understanding of the relationship

between social media and movie sales performance at

the same time brings new insights for the marketer in

the Chinese movie industry.

2 METHODOLOGY AND

RESEARCH DESIGN

2.1 The Correlation between Social

Media (as Alternative Evaluation

Sources) and the Total Movie Box

Office

From the Engel-Blackwell-Kollat model, it is

understood that the purchase decision is affected by

the evaluation of alternatives (Nash 2019). Likewise,

the movie box office would be affected by the

IWOM, peer reviews, the level of popularity, and

favourability on social media. This is because the

movie, as an experienced product, would also be

through the step of evaluation of alternatives in the

consumer behavior process. To avoid a bad movie

experience, consumers tend to take mass reviews

such as movie ratings and audience comments on

social media as alternative evaluation factors (Chen

et al. 2011). To find out the correlation between the

several common-used evaluation factors and movie

box office and have a direct visualization of the

correlation, Pearson Correlation Coefficient

heatmaps are applied. Four alternative evaluation

sources have been proposed: 1) professional rating, 2)

public rating, 3) social media popularity, and 4) social

media favourability.

2.1.1 Data Interpretation

Thirteen movies released in the year 2023 have been

chosen for the first perspective. To improve the

reliability of the research, four principles were

applied when selecting sample movies (Table 1). To

accommodate those control variables, the movie box

office of the following thirteen movies is chosen in

this study: 1) No more bets, 2) Never Say Never, 3)

Dust to Dust, 4) Creation of the Gods l: Kingdom of

Storms, 5) Just for meeting you, 6) Lost in the stars,

7) One and only, 8) Post Truth, 9) Papa, 10) Ping

Pong: The Triumph, 11) Manifesto, 12) Too Beautiful

to Lie, and 13) Heart’s Motive.

Table 1. Sample Selection Principles.

Principles Significance of controlling

General

Releasing

Date

Movies released in the following three

peak seasons have been excluded from the

sample choice: the Chinese New Year

period, the first week of May (Labor Day

Golden Week), and the first week of

October (National Day Golden Week).

This is to ensure the movie box office is

contributed by a reasonable normal traffic

flow instead of seasonal traffic flow.

Similar

Cast and

Crew

To avoid the celebrity effect that

potentially boosts the movie box office.

Series

Movies

Exclusive

To avoid fans or detractors carrying

forward from the previous movies,

resulting in a potential increase or decrease

in the movie box office.

Same

Genre of

Movie

Similar to the celebrity effect, choosing

movies sharing a similar genre is to avoid

high movie box sales driven by high-tech

The Correlation Between Social Media and Consumer Behaviour: A Research on Movie Industry in China

261

animation production. In this sample pool,

the movies are all real-life shooting with

real actors.

Measurement for social media as evaluation

sources is as below.

Professional Rating = 0.5 (Weibo Rating + IMDb

Rating)

75% of the Weibo rating is contributed by

authority media, industry experts, and top influencers

in the movie industry, and thus it has been considered

a professional rating source. In the meanwhile, Weibo

Rating is retrieved from Weibo. cn and IMDb Rating

is retrieved from imdb.com.

Public Rating = 1/3 × (Maoyan Rating + Douban

Rating + Taopiaopiao Rating)

Three popular public rating benchmarks are

adopted. Maoyan Rating is retrieved from the

application Maoyan PRO; Douban Rating is retrieved

from m.douban.com; Taopiaopiao Rating is retrieved

from the application Taopiaopiao.

Social Media Popularity = 0.68 / (0.68 + 0.87)

Accumulated Hottest Topics on Weibo + 0.87 /

(0.68+0.87) × Accumulated Hottest Topics on

Douyin.

Social Media Popularity is estimated by two

sources, the Accumulated Hottest Topics and

Accumulated Hottest Topics on Douyin. Both data

sources are retrieved from the application Maoyan

PRO. Proportion is allocated based on the Social

Media Usage Report (Ipsos 2023).

Social Media Favorability = 0.68 / (0.68 + 0.87) ×

(0.2 × number of like on Weibo + 0.3 × number of

comments on Weibo + 0.5 × number of Weibo repost)

+ 0.87 / (0.68 + 0.87) × number of like on Douyin

Social Media Favourability is estimated by two

sources: Weibo engagement in 3 perspectives

(retrieved from Weibo.cn.), and several like on

Douyin (retrieved from the application Douyin).

Proportion is allocated based on the Social Media

Usage Report (Ipsos 2023).

Movie box office = Total Movie box office of the

movie

Data is retrieved from China Movie Database (*

The measurements of popularity and social media

favourability are based on movies' official accounts

on Weibo and Douyin only.)

2.1.2 Result and Observation

The heatmap is outputted based on the Pearson

Correlation Coefficient, which indicates the

measurement of the linear association strength

between the two variables. According to Sedgwick, a

small sample needs a correlation coefficient in larger

values so that the linear association could be

significant, which is closer to − 1 or 1 (Sedgwick

2012). In terms of the heatmap result obtained, the

correlation coefficient value is observed to range

from 0.2 to 1. A higher value in the result suggests a

stronger positive correlation between the two

variables, and the cell would be in a lighter color

accordingly. The correlation between social media (as

an alternative evaluation source) and the total movie

box office in Figure 1.

Figure 1: Pearson Correlation Coefficient heatmaps (Picture credit: Original).

ICDSE 2024 - International Conference on Data Science and Engineering

262

Surprisingly, being the most obvious factor when

people are looking for reviews on social media, both

professional ratings and public ratings don’t seem to

have a strong correlation with the movie box office.

They have only a noticeable relationship of

correlation coefficient of 0.5 and 0.46 separately with

the movie box office. This indicates when there is an

increase in professional rating and public rating there

is a tendency for the movie box office to increase

while the degree of accuracy would need to be

moderated.

Social media popularity has a substantial

association with the movie box office than the ratings

of 0.69. It indicates that the movie box office is likely

to change according to the social media popularity

although it is not perfect. The correlation coefficient

of social media favourability has the strongest

relationship with the movie box office of 0.91 which

is close to a perfect correlation. It suggests that when

social media popularity increases, the movie box

office is very likely to increase similarly in a high

degree of confidence. In addition to the main factors

being boxed in red, the correlation coefficients are

interesting to be found in individual factors. Social

media as the factor most related to the movie box

office, has a strong positive correlation with the

number of likes on Douyin. This implies favourability

reflected by Douyin has a strong influence on one’s

movie consumption. On the other hand, the lowest

correlation coefficient of 0.17 is between Weibo

repost and the movie box office. All correlation

coefficients about Weibo are in the range from 0.17

to 0.66. The influence of Weibo is hence shown to be

decreasing.

2.2 The Correlation Between Social

Media Engagement (Reflected by

Different Social Media Indexes)

and Movie Box Office

The data sourse: We Are Social; DataReportal;

Meltwater. From the data, the top six social media

platforms with the highest share of internet users as

of Q3 2023 are WeChat, Douyin, QQ, Baidu Tieba,

Xiaohongshu, and Weibo (Fig. 2). Due to the absent

of QQ and Xiaohongshu index daily, as well as the

temporary maintenance on the Weibo index website,

out of the top six social media platforms, only Baidu

Index, Douyin Index and Wechat Index have been

adopted as independent variables. In addition,

Toutiao is a core and significant application of

ByteDance China, whose relationship with movie

sales is also worth exploring.

Figure 2: Top 6 social media with the highest share of

internet users in China as of Q3 2022 (Original).

To figure out the correlation between social media

engagement and movie box office daily, and how the

social media platform affects movie sales

performance, the Ordinary Least Squares (OLS)

Regression model is adopted.

2.2.1 Data Interpretation

Based on the sample in the first prospective research,

the following four movies have been chosen for

further study in the regression model for movie box

office prediction: 1) No more bets, 2) Creation of The

Gods I: Kingdom of Storms, 3) Dust to dust”, and 4)

Papa. The selection of sample movies in the second

perspective research has been one step further

complied with the principle of the normal release

date. The realizing time frame is narrowed from July

to September 2023.

Douyin index is extracted from oceanengine.com.

And it is calculated based on content score, spread

score, search score in three parts.

Baidu index is from index.baidu.com, which

calculated based on Baidu's intelligent distribution

and recommendation content data, the information

index is obtained by weighted summation of social

media users’ behavior such as reading, comments,

sharing, likes, and dislikes.

WeChat index is from the original Tencent

WeChat source, reflecting the popularity and

importance of keywords in the related content.

Toutiao Keyword Index is extracted from

oceanengine.com, and it is calculated based on the

sample principle as that of the Douyin index.

Variable measurement:

Dependent variable (Y): Daily movie box office

of the four movies.

Independent variable (X1): Daily Douyin Index

based on exact movie name.

Independent variable (X2): Daily Baidu Index

based on exact movie name.

Independent variable (X3): Daily WeChat Index

based on exact movie name.

82%

72%

62%

58%

50%

49%

10%

20%

30%

40%

50%

60%

70%

80%

90%

The Correlation Between Social Media and Consumer Behaviour: A Research on Movie Industry in China

263

Independent variable (X4): Daily Toutiao

Keyword Index based on exact movie name.

Sample size (n) is 80, which is made up of the first

twenty days index of each variable for the four

movies.

The four variables multiple linear regression

model is estimated as:

𝑌=𝛽



+𝛽



𝑋



+𝛽



𝑋



+⋯+𝛽



𝑋



2.2.2 Hypothesis

H1: Douyin Index has a positive influence on movie

box office,

𝛽



≠0

H2: Baidu Index has a positive influence on movie

box office, 𝛽



≠0

H3: WeChat Index has a positive influence on

movie box office, 𝛽



≠0

H4: Toutiao Keyword Index has a positive

influence on movie box office,

𝛽



≠0

2.2.3 OLS Regression Result and

Observation

From the OLS Regression result in 1 (Table 2, 3, 4),

the P value of both the Douyin Index and the Baidu

Index is less than 0.05, which shows a statistically

significant relationship between the two independent

variables and the dependent variable of the movie box

office. The P value of the Weixin Index and Toutiao

Keyword Index is larger than 0.05, indicating both

indexes are not statistically significant in predicting

movie box office. Therefore, H3 and H4 are rejected.

Table 2: OLS Regression result 1(a).

Dep. Variable: Movie

Box

Office

R-squared: 0.835

Model: OLS Adj. R-

uared:

0.826

Method: Least

uares

F-statistic: 94.62

No.

Observations:

80 Prob (F-

statistic):

1.60e-

Df Residuals: 75 Log-

Likelihood:

1474.7

Df Model: 4 AIC: 2959.

Covariance

Type:

nonrobust BIC: 2971.

Table 3. OLS Regression result 1(b).

coef std er

tP>

[0.025 0.975]

const 1.529e+07 4.14e+06 3.695 0.000 7.05e+06 2.35e+07

Dou

in Index 11.1489 2.082 5.355 0.000 7.001 15.296

Baidu Index 7.6118 1.325 5.746 0.000 4.973 10.251

Weixin Index 0.0234 0.042 0.558 0.579 -0.060 0.107

Toutiao Keyword

Index

-45.5014 58.701 -

0.775

0.441 -162.440 71.437

Table 4. OLS regression result 1(c)

Omnibus: 4.436 Durbin-Watson: 1.431

Prob (Omnibus): 0.109 Jarque-Bera (JB): 4.188

Skew: 0.308 Prob (JB): 0.123

Kurtosis: 3.936 Cond. No. 2.34e+08

Table 5. OLS regression result 2(a).

Dep. Variable: Movie Box Office R-squared: 0.833

Model: OLS Adj. R-squared: 0.829

Method: Least Squares F-statistic: 192.4

No. Observations: 80 Prob (F-statistic): 1.13e-30

Df Residuals: 77 Log-Likelihood: -1475.1

Df Model: 2 AIC: 2956.

Covariance Type: nonrobust BIC: 2963.

Table 6. OLS Regression result 2(b).

coef std err t P>|t| [0.025 0.975]

const 1.459e+07 3.96e+06 3.680 0.000 6.69e+06 2.25e+07

Douyin Index 10.2052 1.190 8.576 0.000 7.836 12.575

Baidu Index 8.1049 1.156 7.010 0.000 5.803 10.407

ICDSE 2024 - International Conference on Data Science and Engineering

264

Table 7. OLS Regression result 2(c).

Omnibus: 4.141 Durbin-Watson: 1.405

Prob (Omnibus): 0.126 Jarque-Bera (JB): 3.587

Skew: 0.337 Prob (JB): 0.166

Kurtosis: 3.789 Cond. No. 8.45e+06

With only the Douyin Index and Baidu Index

retained for the second round of OLS regression

modeling, the result is output as OLS Regression

result 2 (Table 5, 6, 7).

P value for both Douyin Index is less than 0.05,

indicating a statistically significant relationship

between Douyin Index and the Movie Box Office. A

coefficient value of 11.1489 suggests every unit

increase in Douyin Index will result in an increase in

the movie box office by 11.1489 units.

P value for Baidu Index is 0.05, indicating a

statistically significant relationship between Baidu

Index and the Movie Box Office. The coefficient

value of 8.1049 suggests every unit increase in the

Baidu Index will result in an increase in the movie

box office by 8.1049 units.

R- square value is 0.833 (Adj R- square = 0.829),

suggesting 83.3% of the total variation could be

measured by the model. Both the R – square value and

Adjusted R – square value at 80% + level, reporting

the total amount of variation that can be accounted for

linear regression model are high.

F-statistic is observed at 192.4. Together with the

low P value (0.000), it suggests the model is

statistically significant, and the linear regression

between Y and X1, and X2 is strong.

Given the coefficient estimators from the OLS

Regression result, the two-variable multiple linear

regression model is expressed as:

𝑌 = 1.459𝑒 + 07 + 10.2052 𝑋1 + 8.1049 𝑋2

Overall, out of the four investigated social media

platforms, Douyin and Baidu are the two platforms

that appear to be significant in terms of predicting

movie sales box. At the same time, that reflect social

media has a significant impact on consumer behavior

of purchase in the movie market.

3 CONCLUSION

In this paper, the correlation relationship between

social media and the movie box was studied from two

perspectives. Results did show that social media plays

a role in influencing movie sales performance.

Specifically, from the two perspectives research

results, social media platforms, such as Douyin and

WeChat, serving as a general information channel,

appear to have a stronger positive impact the movie

sales performance. Surprisingly, social media index

from specific movie informational platforms, such as

Maoyan and IMDb has comparable week correlation

with movie sales. The marketer may want to shift

their marketing effort to the general social media

platform instead to achieve more effective marketing

objectives.

In the meanwhile, limitations of the research

model are also observed as below. There will be time

between information search and purchase. While in

the modeling, the factor has not been taken into

consideration, mainly because of very limited

research on the time buffer between the two steps,

especially in experience consumption. Also,

considering the convenience of online purchases

nowadays, the effect on the time buffer should be

minimized. However, its potential influence on

consumer behavior remains an area not thoroughly

examined in this research.

Due to the researcher’s technical expertise and

data availability online, this study only considers the

impact brought by some social media platforms in

China. Other social media platforms such as Youku,

IQIYI, Bilibili, as well as news channels, are not in

consideration. Thus, this study might not fully capture

the diverse impacts different platforms can have on

movie box office performances.

The heatmap and multiple regression model only

show the correlation between various factors and the

movie box office, they do not show the causation

caution. The movie box office may be affected due to

other external reasons such as political stance and

societal zeitgeist which are not accounted for in this

study.

Looking forward, it is valuable for researchers to

deep dive into the relationship between movie box

office performance and the social media impact in a

more precise manner. A larger sample size and

diversified social media platforms will continue to

have a wider outlook.

The Correlation Between Social Media and Consumer Behaviour: A Research on Movie Industry in China

265

REFERENCES

S. Kemp, Digital 2023: China. DataReportal - Global

Digital Insights,

https://datareportal.com/reports/digital-2023-china.

M. A. Shareef, B. Mukerji, Y. K. Dwivedi, N. P. Rana, R.

Islam, J Retail Consum Serv 46, 58-69(2019)

R. V. Kozinets, Eur Manag J 3, 252-64(1999)

A. Chopra, V. Avhad, A. S. Jaju, Business Perspectives R

1,77-91(2021)

J. Nash, J Fashion Mark Mana 23, 82-103(2019)

A. Wibowo, S. C. Chen, U. Wiangin, Y. Ma, A.

Ruangkanjanases, Sustainability 13, 189(2020)

T. Kubrak, Behav Sci Law 5, 86(2020)

A. Molaie, A. Abedin, M. Heidari, Proc Social Behav Sci

5, 832-837(2010)

China Film Data Information Network. The 2023 film box

office surpasses 500 billion yuan.

https://www.zgdypw.cn/sc/scfx/202311/20/t20231120

_7373236.shtml

Y. Chen, Film Mark Indu Report 45(2023)

R. Zhu, Y. Ma, J Chaohu Univ 23, 71-79(2021)

S. Asur, B. A. Huberman, IEEE 1, 492-499(2021)

M. Bogaert, M. Ballings, D. Van den Poel, A. Aztecan,

Decis Support Syst 147, 113517(2021)

J. F. Engel, R. D. Blackwell, P. W. Miniard, J Consum

Behav 8(1995)

Y. Chen, S. Fay, Q. Wang, J Interact Mark 25, 85-94(2011)

Ipsos. 2023 Social Media Usage Report. Retrieved from

https://www.199it.com/archives/1607624.html.

P. Sedgwick, Pea correlation coefficient 345(2012)

ICDSE 2024 - International Conference on Data Science and Engineering

266