The Effect of Marketing Activities on Web Search Volume: An Empirical Analysis of Chinese Film Industry Data

Abstract

Prior research on consumers’ web searches primarily examined the effect of web searches on product sales or the characteristics of the web searchers. Differing from prior research, we investigate the effect of marketing activities on web search volume. We selected 314 movies released in China whose box office revenues were greater than RMB 10,000. Then, we collected data points on web search volume and marketing activities from the Baidu, Sina Weibo, and Douban platforms from the 3 weeks prior to the release of each movie. Marketing activity data points were derived from three sources: news media, social network marketing, and film stars. Our data analysis of 6594 observations revealed two major findings. First, news media, social network marketing, and the effect of film stars increased the web search volumes of the films. In particular, social network marketing had the strongest impact on the web search volume. Second, the previous-day web search volume increased the present-day web search volume without marketing activities, suggesting a spillover effect. We discuss the academic contributions and managerial implications of our findings in the context of online marketing and new product launches.

1. Introduction

A search engine is an effective online marketing instrument for new products, such as newly released movies [1]. When a movie is newly released, for instance, consumers often use search engines to obtain information on its quality. Marketers should understand how to entice consumers to effectively use search engines, particularly how the consumers’ web search volume is influenced by other large-budget marketing activities.

During this past decade, researchers have paid much attention to consumers’ web searches. For example, some researchers have investigated how web search behavior influences sales and demand [2,3]. Other researchers have explored the distinctive characteristics of the people who conduct web searches [4]. Although a vast amount of research investigates the impact of web searches on sales or the characteristics of web searchers, little is known about whether marketing activities influence consumers’ web searching [5].

This study aims to close this gap by empirically examining whether the web search volume is determined by three popular marketing activities: news media, social network marketing, and film stars [4,5]. To achieve this goal, we collected a comprehensive data set regarding the Chinese film industry from 2013 to 2015 from three reliable sources: the largest search engine (Baidu.com), the most popular microblog (Weixin), and a leading movie review site (Douban.com). Note that the Chinese film industry is a useful research source because China represents the largest amount of web search traffic in the world. It has more than 566 million people representing 82% of internet users, and about 80% of its internet users perform web searches when they search for entertainment information (CIW) (Source: https://www.chinainternetwatch.com/statistics/china-internet-users) (accessed on 30 December 2021).

Our data analysis reveals several interesting findings. First, the three marketing activities increased the web search volume. Out of the three marketing activities, two activities (news media and social network marketing) increased the web search volume specifically related to the movie. The last activity, the effect of the film stars, increased the web search volume, but this was not due to the acting quality of the stars specific to the film and was rather due to their popularity. Furthermore, social network marketing was a more effective marketing activity than news media and the effect of film stars in increasing the web search volumes of films. The elasticity of social network marketing (1.01) was exceptionally larger than that of the news media (0.01) and that of the film stars (0.02). Second, there was a spillover effect of web searching; about 30% of the web search volume was generated without marketing activities.

Our findings contribute to academic communities and add insights into the online marketing literature by expanding our understanding of the consumers’ web search generation process and verifying its relationship with marketing activities frequently initiated for new product launches. Our conclusions urgently call for further theoretical studies to examine web search generation, which has been minimally discussed among marketing researchers [5].

In addition, our findings provide fresh insights into marketing with clear managerial implications. The findings provide a better understanding of the role of stars in the film industry. Whether the stars increase the sales of a movie has been discussed for several decades, while stars of a movie have been unanimously considered to be a critical attribute of a movie [6]. Our findings that film stars led consumers to search for information about the film suggest that stars are effective marketing tools. More broadly, marketers will benefit more when investing in social network marketing than investing in public relation outputs such as news media or film star promotion.

The rest of the paper is organized as follows. We provide a summary of the existing literature in the next section, describe our data set, and introduce our empirical model in Section 3, provide the estimation results and then the ensuing discussions in Section 4, and finally conclude the paper with discussions, implications, and suggestions for future research.

2. Literature Review

Web search activity is a topic of great interest in several academic disciplines. It is an indicator of public interest in the epidemiology field. For instance, the volume of the Google web search queries is an efficient, real-time public interest indicator compared with traditional surveillance measurements [7]. It is also an economic indicator, as economists use web search volume to forecast unemployment rates, household expenditures, and housing prices [8,9,10]. Furthermore, the web search volume also predicts the stock price in the financial market [11].

Not surprisingly, web search activity is an important topic among marketing researchers. On one hand, some researchers focus on the role of web searches as a leading indicator of sales or the demand for new products. Web search trends predict new car sales in the US automobile industry [5]. Web search volume predicts the demand for tour and retail products [12]. A cinema admission forecasting model is more accurately forecasted when cinema-related Google Trends search data are included in the model [3]. Similarly, the predictive power of forecasting models for the box office revenue of a new movie improves when web search data are employed [2]. A recent analysis of Google Trends search data in Finland showed that the peaks in prime time for streaming media services have replaced television primetime [13]. On the other hand, other researchers investigated the distinctive characteristics of web searchers. Younger and more educated consumers prefer the internet over traditional research vehicles because the internet reduces reaching costs and provides information more efficiently than interpersonal sources or print information [14]. Web searchers prefer to use short queries and only see one page of search results [4].

While the prior marketing literature focused on web search effects or web searchers’ characteristics, how a web search is generated has been minimally discussed. This question is important for marketers, because when marketers launch new products to the market, they often spend a large amount of money on marketing activities to compel consumers to search for information about the products. To implement these marketing activities successfully, they should understand the relative effectiveness of marketing activities, particularly in the contemporary environment, where online services and mobile devices are major sources of information.

Therefore, this paper investigates how marketing activities generate consumers’ web search behavior. Specifically, we test our hypotheses in the film industry, where the quality of each product is difficult to convey to consumers through other marketing activities [15]. Note that movie marketers perform various marketing activities to attract consumers’ attention and deliver detailed information about the movies before they are officially released. In this paper, we focus on three popular marketing activities: news media, social network marketing, and film star promotions.

2.1. News Media

News media provides information about an upcoming entertainment product to stimulate curiosity among consumers [16]. Such curiosity, together with the cognitive pleasure generated by acquiring new information, triggers consumer interest in the product [17]. Therefore, news related to a new movie can stimulate consumers’ interest and lead to web searches for more detailed information. Furthermore, the news is viewed as more credible than advertising [18]. News that appears in a non-advertising context can exert a stronger influence on consumers’ attitudes toward a product, because consumers use credible information sources to assess the trustworthiness of information in the marketplace [19].

2.2. Social Network Marketing

Introducing new products to social networks such as Twitter or Facebook, the most recent and effective conversation instruments, is recommended. We predict that exposure to a social network increases web search volume through two paths. First, it stimulates receivers’ interest in new products. In addition, it entices people who are connected to the primary receivers through the WOM effect. Indeed, social networks are effective tools to generate WOM in the film industry [15,20]. Characterized by ease of use, speed, and reachability, social networks have successfully transformed public discourse in society and led to trends in public topics such as the environment, politics, technology, and entertainment [20]. In particular, microblogs such as Twitter are effective at generating WOM promotion for films [15].

2.3. Film Stars

Film stars are another trigger of public interest in movies. Prior research suggests that stars in a movie can capture their expected economic rent [21]. There are two reasons why stars increase web search volume. First, the fact that the actors or actresses are famous often signals a movie’s quality [21]. In particular, moviegoers favor movies with award-winning stars because they guarantee the quality of the movies. Stars are important marketing tools in the film industry because they are symbols of a movie’s identity and attract consumers with corresponding tastes to the cinema [6]. Second, stars increase web search volume regardless of their acting qualities. They are communal points and act as a medium of communication among consumers regardless of their performances [22]. Star power is particularly important for loyal consumers or fans. Fans go to the cinema mainly out of loyalty to the stars themselves [23]. Similarly, consumers who are loyal to stars are more likely to be loyal to their movies [24]. Thus, the presence of stars will boost consumers’ web searches.

2.4. The Dynamic Effect of Web Search Generation

The web search volume is not only influenced by these marketing activities in a static manner. It may also have its own dynamic impact on the generation process caused by a spillover effect. For example, consumers who search for information about a movie may speak to other consumers about the obtained information and incite more consumers to search for the information about the same movie. Specifically, the expanded image and information about a movie can remain over time and drive further research of a movie. If this is the case, marketers should consider the dynamic impact of consumers’ web searches; otherwise, they underestimate the impact of other marketing activities on the web search volume. Thus, we investigate the dynamic effect of the web search generation process.

Accordingly, we have three research questions to address in this paper. First, we test whether and how three marketing activities (news media, social network marketing, and the effect of film stars) generate web searches for a new movie. Second, we investigate how effectively these activities generate web searches. Finally, we explore whether a dynamic impact on the generation process caused by a spillover effect exists.

3. Data and Model

3.1. Data

We collected data from movies released between 2013 and 2015 in China. In particular, we collected data from 314 movies whose box office revenues were greater than RMB 10,000. For each movie, we collected the web search volume from Baidu and the data from the three marketing activities from these reliable sources: (1) news media from the Baidu Media Index, (2) social network marketing from Sina Weibo, and (3) film information from Douban, a leading online community about movies. Since we collected data points from the 3 weeks before each movie was released, our data set had a total of 6594 observations.

Along with the aforementioned panel data set, we employed a dynamic panel model to address our research questions properly [25,26,27]. To explore the dynamic effect of the web search generation process, we incorporated a lagged dependent variable in the linear panel model that included a time-invariant individual effect for each movie. We considered the individual effect as a fixed effect to capture the specific characteristic related to each movie unobserved by a researcher. Additionally, the key variables for generating the web search volume would be specified.

3.2. Dependent Variable

Our dependent variable was the web search volume (WSV) for a movie prior to its release. This variable is often measured by a search engine index (e.g., Google Trends in [9,10]). The Baidu Index is a more qualified proxy in the context of Chinese-based searches than Google Trends, because Google is blocked in China due to government regulations, and Baidu is the search engine most widely used in China. According to the Report on Searching Behavior of Chinese Internet Users released in 2014 by the China Internet Network Information Center (CINIC), 97.4% of the Chinese internet users have experience searching via Baidu, and 88.7% use it as their primarily selected search engine (CINIC). Baidu computes the Baidu Index based on the actual search volume of queries and provides search volume trends daily. Therefore, we employed the Baidu Index as a measurement of web search volume.

3.3. Independent Variable

We included three key variables—news media, social network marketing, and film stars—to capture marketing activities. First, we measured news media (NWM) by employing the Baidu Media Index. This index shows how many headlines of the major online media and web portals a movie title appears in. It indicates the online news coverage of each movie. Media generally handles “newsworthy” issues that are considered important information for moviegoers [28,29,30]. Thus, it plays a critical role in inducing consumers to search for information about the related movie. In addition, news headlines are frequently used to measure the media effect on the prior literature [31,32,33,34].

Second, we measured social network marketing (SNM) by collecting the daily number of “mentions” of a movie title on Sina Weibo, the most popular microblog in China [35]. It measures the exposure of a movie through social network services. The microblog has been heavily adopted as a representative SNM tool in the electronic word-of-mouth (eWOM) literature [35,36,37]. Word of mouth is considered a more persuasive source for a marketer to provide information on the relevant product to their consumers than other marketing activities [38,39]. Additionally, the number of “mentions” is representatively employed to measure the effect of SNM [40,41].

Third, we measured the power of stars by employing two types of variables: performance-based star power (PSP) and non-performance-based star power (NSP). (Note that Star Power is related to the intrinsic feature of a movie and therefore could be viewed as the quality of casting. We appreciate anonymous reviewers who suggested we see the variables from a different perspective.) Performance-based star power was measured by the total number of awards won by the main actors and actresses in a movie. This captured the power of the actors and actresses in implying the quality of the casting. It is well known that actors and actresses cast in a movie are considered an important indicator of the quality of a movie for moviegoers [6]. The awards won by actors or actresses are frequently used as a measurement of the quality of the casting which implies the quality of a movie in the marketing literature. On the other hand, there exists the non-performance power of stars that induces people to watch a movie rather than the quality of the acting they provide [21]. This is related to the popularity of the stars based on their fandoms. It is also well known as an important factor that influences moviegoers to decide to watch a movie. For non-performance-based star power, we measured the total number of fans of the main actors and actresses in a movie.

3.4. A Lagged Dependent Variable and Control Variables

In addition to the dependent and independent variables, we included a lagged dependent variable or the dependent variable of one previous period (WSV_t−1) as an independent variable. This variable captured the effect of the volume of the previous day’s web searches on the volume of the present-day web searches without considering marketing activities. We did this to test whether a dynamic effect existed in the web search volume.

We also included two sets of control variables in our model. First, we identified the characteristics of the movies. We included the genre (eight different genres), format (whether they were regular movies or three-dimensional stereoscopic (3D) ones), and IP (whether they were adapted from intellectual properties (IPs) which had been popular in recent years). Specifically, IP movies can be described in the following way: “IP is a pre-existing property that sees a film studio hire screenwriters, actors, and producers to adapt into a film. This means huge franchises, comic book characters such as Spider-Man, and all remakes of other movies and shows are IP” (source: https://www.thefocus.news/culture/ip-movies-meaning/) (accessed on 30 December 2021). We considered these variables because public reactions toward movies are known to be influenced by such movie characteristics [42]. We also included a monthly dummy to control the seasonality effect.

3.5. Empirical Model Specification

Along with the variables introduced in the previous section, we specify our empirical model as follows:

$$\begin{array}{l}WS{V}_{it}={\beta }_0+{\beta }_1\ast WS{V}_{it-1}+{\beta }_2\ast NW{M}_{it}+{\beta }_3\ast SN{M}_{it}+{\beta }_4\ast PS{P}_i+{\beta }_5\ast NS{P}_i\\ +GenreDumm{y}^{\prime }\mathsf{\gamma }+3DDumm{y}^{\prime }\mathsf{\theta }+IPDumm{y}^{\prime }\mathsf{\lambda }+MonthDumm{y}^{\prime }\mathsf{\eta }+e_{it} ,\end{array}$$

where $e_{it}$ = ${\alpha }_i+{\epsilon }_{it}$, i equals ‘movie’, and t equals ‘day’. To capture the unobserved individual effect, we included ${\alpha }_i$.

Table 1. Descriptive statistics of key variables.

Variable	Min	Mean	s.d.	Max	Obs
Web Search Volume (WSV)	0	20,173	77,426	2,743,351	6594
News Media (NWM)	0	52	137	3432	6594
Social Network (SNM)	0	124,340	1,898,231	7.81 × 107	6594
Number of Awards won by Actor or Actress (PSP)	0	1.44	2.50	21	6594
Number of Fans of Actor or Actress (NSP)	0	31,061	40,366	202,868	6594

reports the descriptive statistics of the key variables, and

Table 2. Correlation coefficient analysis between key variables.

	WSV	NWM	SNM	PSP	NSP
WSV	1
NWM	0.417	1
SNM	0.013	0.011	1
PSP	0.006	0.077	−0.011	1
NSP	0.245	0.274	−0.032	0.161	1

reports the correlation analysis between a dependent variable and four independent variables. Multicollinearity between variables was not observed.

4. Results

We employed the Arellano–Bond (AB) dynamic panel estimation approach to address our research questions. The results of our estimation are provided in

Table 3. Estimation results of AB dynamic panel estimation.

	Coefficient Estimates	Standard Error	p-Value
Constant	−9978.64 ***	2234.36	0.000
$WS{V}_{it-1}$	0.343 ***	0.0001	0.000
$NW{M}_{it}$	10.346 ***	0.0297	0.000
$SN{M}_{it}$	0.565 ***	0.0006	0.000
$PS{P}_i$	−146.90	684.75	0.830
$NS{P}_i$	0.041 ***	0.016	0.011
$\mathrm{Genre} \left(\mathrm{Drama}\right)$	347.24	981.74	0.724
$\mathrm{Genre} \left(\mathrm{Comedy}\right)$	4564.11 ***	923.80	0.000
$\mathrm{Genre} \left(\mathrm{Action}\right)$	4844.28 *	2534.04	0.056
$\mathrm{Genre} \left(\mathrm{Romance}\right)$	2242.35 *	1157.86	0.053
$\mathrm{Genre} \left(\mathrm{Fiction}\right)$	119.85	1501.43	0.936
$\mathrm{Genre} \left(\mathrm{Animation}\right)$	1843.57	1627.04	0.257
$\mathrm{Genre} \left(\mathrm{Thriller}\right)$	245.56	1298.53	0.850
Genre (Documentary)	361.42	1066.78	0.735
3D	2708.74	1827.19	0.138
IP	5670.06 ***	1349.61	0.000
January	8119.13 ***	1368.28	0.000
March	3930.59 *	2001.42	0.050
April	716.79 ***	1442.36	0.619
May	6172.33 ***	2116.58	0.004
June	4175.15	2585.75	0.106
July	2834.88 *	1624.05	0.081
August	3361.38 *	1900.21	0.077
September	3058.37 ***	1142.65	0.007
October	1460.31	1836.80	0.427
November	0.054	0.071	0.444
December	0.087	0.223	0.696

(We can easily adjust this statement if it is not appropriate; p-value: * <0.1, *** <0.01.)

.

First, the three marketing activities influenced the web search volume. As expected, news media (NWM) and social network marketing (SNM) significantly increased the web search volume (p-value = 0.000 for both NWM and SNM). While performance-based star power (PSP) had no impact, non-performance-based star power (NSP) increased the web search volume (p-value = 0.83 and p-value = 0.011 for PSP and NSP, respectively). These results imply that news media, social network marketing, and star power, particularly non-performance-based star power, could play a significant role in the web volume generation process. It is noteworthy that the popularity of star actors and actresses can be helpful for increasing web searches of the movie they are cast in. However, the quality of casting may not induce people to search more for movie information, but it might help a moviegoer to determine their decision to watch a movie [6,16,21].

In addition, we found that the genre of the movie played an important role in influencing the web search volume. This indicates that the web search volume was significantly higher for the comedy genre movies (p-value = 0.000), and we found marginal evidence for the action and romance genres that might have been helpful for increasing the web search volume (p-value = 0.056 and 0.053, respectively). Furthermore, we found that IP movies could be beneficial in increasing the web search behaviors of people. The results show a significant increase in web search behavior for IP movies (p-value = 0.000). Additionally, we found a significant seasonal effect on web search behavior for several months, namely January, March, May, and September (p-value = 0.000, 0.05, 0.004, and 0.007). This indicates that web search behavior was relatively higher for certain months due to seasonal events such as holidays, vacations, weather, school breaks, and so on [6].

To address the second research question, we computed the elasticity of each activity. This showed that social network marketing (SNM) generated the web search volume most strongly, followed by non-performance-based star power (NSP) and news media (NWM) (

Table 4. Elasticity of marketing activities for web search generation.

	NWM	SNM	NSP
Elasticity of Web Search Generation	0.008	1.018	0.018

). Specifically, a one percent change in the three variables (NWM, SNM, and NSP) increased the web search volume by 0.008%, 1.018%, and 0.018%, respectively. This implies that social network marketing is relatively more effective at generating the web search volume compared with the other two variables. Thus, it provides important managerial implications for a marketer to promote their newly introduced movie so it should be considered during a consumer’s buying process. Social network marketing can be a more effective tool for inducing potential consumers to search for information on their movie rather than other marketing activities.

Finally, we found that the previous web search volume ($WS{V}_{it-1}$) significantly increased the present web search volume (p-value = 0.000). This finding implies that a dynamic effect exists on web search behavior; that is, when people search the web for information on a movie, the generated volume can help further increase web searches for the movie afterward without any additional marketing activities. This may be because a movie that is highly searched on the web can be viewed by consumers as a popular movie or perceived as a more valuable movie to search. These findings provide an important implication to marketing managers, because the effect of marketing activities on the web search generation process would be expanded for a long time period. Thus, if they do not consider the dynamic effect of the web search generation process, they are likely to underestimate the effect of their key marketing activities.

5. Conclusions

This study investigated web search generation. Unlike prior research studying web search effects or web searchers’ characteristics, we investigated the effect of marketing activities on the web search volume and examined its dynamics. Our analysis of the data collected from the Chinese film industry revealed two major findings. First, the web search volume was influenced by three marketing activities (news media, social network marketing, and the effect of the stars of the film), and social network marketing had the strongest influence on the web search volume compared with the other two activities. Second, we found that web search generation was dynamic; the previous day’s web search volume increased the present-day web search volume [6]. Our two findings contribute to the academic discussion by providing fresh empirical findings related to web search generation. They also provide critical managerial implications to marketers who design marketing activities to launch new products.

More specifically, our first findings suggest that understanding the effect of each marketing activity on the volume of web searches enables marketers to improve their marketing activities. Investing their resources in social network marketing rather than in news media or film star promotion is an effective strategy in the film industry. Next, the star’s performance, such as his or her acting capabilities, has no impact on web searches, whereas their fans increase the web search volume. This supplements the prior research discussing whether the stars of a movie increase movie demand [6]. Our findings explain that stars do increase movie demand by inducing consumers to search for information about movies, discovering the new role of a film star. All in all, our first findings imply that generating a web search is a more complex task than merely increasing consumers’ awareness of a new product [2]. Consumers may search for information not merely to obtain information from others but to share information with others.

Our second findings suggest that a web search generated at one time leads to web searching at another time. This spillover effect indicates that consumers who searched for and obtained information about a movie spread that information to others, which leads new consumers to search for information about the same movie. Therefore, the word-of-mouth effect still exists even when marketing activities are not incorporated. Note that this will provide meaningful implications for international multimedia industries such as TV show production. According to previous analysis of the real-world search data from the Bing search engine, TV show preferences vary systematically based on whether the query is observed before, during, or after a TV show is aired [42]. Research also found that consumers’ search and purchase behaviors for experience goods were heavily influenced by the product reviews from other consumers and multimedia [43]. Our obtained spillover effect could be either amplified in this specific media context or nullified.

While we provided important insights for marketers, four potential future research projects are needed to overcome the limitations of this study. First, the commercial impact of marketing activities needs to be further studied beyond the web search volume. We certainly believe that not only information searches, but also evaluation, choice, and even online review generation will be determined by marketing activities [44,45,46]. However, this question was out of the scope of our current research, because we were specifically interested in the web search volume. Second, this topic can be further explored in a non-film industry, such as in an electronic device shopping context. Social network marketing is known to decrease uncertainty about a newly developed product, which increases favorable attitudes toward it [47]. Similarly, news media and stars may enhance people’s preferences for new products. Third, there are alternative explanations for our obtained findings. For instance, the searched movie might remain in consumers’ memories and lead them to search for it again in the future, which can form a more positive image to the consumers due to the familiarity or increased knowledge of the movie. Indeed, consumers tend to evaluate new product bundles highly when they identify that the bundles are composed of familiar products [47]. It would be meaningful if future research investigates this issue further. Fourth, our obtained findings on the impact of marketing activities on movie goers’ web search behavior need to be tested outside of China to obtain external validity. This is particularly important because different countries are in different stages of motion picture industry development and thus are characterized by different behaviors of moviegoers [3,6]. Despite these limitations, our research provides sound data that marketers should focus on social network marketing when looking to increase the web search volume for a new product.