This study aims to explore the interdependent impacts of online word-of-mouth (WOM) and online ads on digital product adoptions, as well as their dynamic changes throughout the product life cycle.
This study adopted an empirical approach by using a unique data set of five mobile games launched between 2012 and 2014 provided by Renren Games Ltd. in China.
The results indicated that advertising generally has a positive impact on WOM. During the product life cycle, the influence on volume and variance gradually decreases, whereas the impact on valence increases over time. WOM (including WOM volume and WOM valence) and advertising both have positive impacts on game adoptions. They complement each other to shape adoptions throughout the product life cycle: advertising is more effective in encouraging adoptions in the early and later stages of the demand evolution process, whereas WOM has a greater impact on adoptions in the mid-stage.
This study provided detailed managerial recommendations on how to effectively integrate different types of marketing communication and optimize the investment strategy of online ads and online WOM in different stages of the product life cycle.
First, the study enriched the theory of digital marketing communication by studying the relationship between mass media (online ads), interpersonal media (online WOM) and product adoptions in the network context. Second, it provided an empirical basis for the inference of the dynamic development of media effect in the new product diffusion theory. Third, the results will be helpful to end the debate in current theoretical literature on whether there is a complementary or alternative relationship between the two effects. Last but not least, it enriched research on the antecedents and dynamic effects of online WOM.
Gong, S., Wang, W. and Li, Q. (2019), "Marketing communication in the digital age: online ads, online WOM and mobile game adoptions", Nankai Business Review International, Vol. 10 No. 3, pp. 382-407. https://doi.org/10.1108/NBRI-12-2018-0073
Emerald Publishing Limited
Copyright © 2019, Emerald Publishing Limited
Consumers usually turn to two types of media to obtain information of products or services when making purchasing decisions: the mass media and the interpersonal media. The internet is playing an increasingly significant role against the backdrop of China’s “Internet plus initiative.” With the emerging of more digitalized and networked products or services, online ads and online word-of-mouth (WOM) have gradually become two most important marketing communication tools representing the two abovementioned media.
Online ads have been preferred by many companies right after the internet came into being. Compared with the traditional offline advertising, the biggest advantage of online advertising is that it can use the network technology to obtain rich consumer information, and thus has a more accurate access to target consumers (Evans, 2008). According to a recent global advertising survey carried out by eMarketer, global advertising spending reached a total amount of $615bn in 2016. Among different types of ad spending, online advertisement has maintained a high growth rate of more than 20 per cent since 2012, reaching a total of $198bn in 2016, which accounted for 32 per cent of global advertising spending.
As social networks and e-commerce develop, the role of online WOM has been attached greater importance by business managers. According to a statistical report, about 20 per cent of the information in social media contains consumers’ WOM comments on brands or products (Jansen et al., 2009). In addition, many e-commerce companies (such as Taobao, Jingdong, Amazon, etc.) and third-party consumer review websites (such as Dianping.com and Douban) have also established online comment systems to encourage consumers to post and share WOM information. It is found in a survey that more than 70 per cent of Chinese consumers take their initiative in looking up the WOM information on these websites when making online shopping decisions (Hu, 2008). Many consumers even refer to the WOM information to reduce risks when shopping offline (Derbaix and Vanhamme, 2003).
In the academia, online ads and online WOM are also interesting topics for researchers (Ambler and Vakratsas, 1999; Godes and Mayzlin, 2004). However, few studies have explored from the perspective of integrated marketing communication, especially the interaction among online ads, online WOM and digital product adoptions, and the dynamic changes of these relationships throughout the product life cycle. Therefore, this study aims to answer the following important questions: how online ads influence online WOM, how online ads and online WOM influence mobile game adoptions and how these two impacts change dynamically over time throughout the product life cycle.
This study adopted a unique data set of five mobile games launched between 2012 and 2014 by Renren Games Ltd, and adopted an empirical approach to study two main questions: the impact of online ads on online WOM and the joint impact of online ads and online WOM on product adoptions. Considering this special research goal, this study established a simultaneous equation model to systematically analyze these two impacts. On this basis, in accordance with the life cycle of mobile game products, the development process of products was divided into three stages in this study, early stage, mid-stage and later stage, and thus the dynamic changes of the two abovementioned impacts over time were further analyzed.
2. Literature review and hypothesis
2.1 The impact of online marketing communication on product adoptions
According to previous marketing research, the new product diffusion theory first emphasized the important influence of marketing communication and media effect on product adoptions (Bass, 1969; Mahajan et al., 1991; Rogers, 2003). Scholars believe that the diffusion of new products is largely attributed to the transmission of new product information through the mass media (e.g. advertising) and the interpersonal media (e.g. WOM). This theory paved the way for a series of marketing communication research.
In recent years, with the development of the internet, piles of marketing communication research focus on online WOM. Compared with the traditional offline WOM, online WOM has two prominent advantages: on the one hand, consumers can easily generate and look up WOM information online, relieving WOM from limitations of time and space, and greatly enhancing the promotion effect and influence of traditional WOM. On the other hand, it is relatively cheaper to produce and manage WOM once it is put online. Therefore, online WOM can effectively reduce companies’ costs in marketing communication (Chen and Xie, 2008).
Many studies have discussed the relationship between online WOM and product adoptions or corporate performance. In summary, prior research mainly used three dimensions of WOM when studying the impact of online WOM – volume, valence and variance – the volume of WOM measures how widely the WOM information has been spread, usually measured by the total number of online reviews; WOM valence shows whether the information is positive or negative, which is usually measured by the mean value of online ratings or the proportion of positive or negative comments; and the variance of WOM information measures the differences in WOM information, usually measured by online consumer rating variance. Given these three dimensions, relevant studies explored the relationship between online WOM and product adoptions or corporate performance in various industries (Table I), such as television ratings (Godes and Mayzlin, 2004), movie box office (Liu, 2006; Dellarocas et al., 2007; Duan et al., 2008; Chintagunta et al., 2010; Rui et al., 2013; Gopinath et al., 2013), book sales (Chevalier and Mayzlin, 2006; Sun, 2012; Gong et al., 2012, 2013), beer sales (Clemons et al., 2006), stock prices (Luo, 2009), restaurant turnover (Lu and Feng, 2009), game adoptions (Zhu and Zhang, 2010), online retail sales (Moe and Trusov, 2011) and video player sales (Ho-Dac et al., 2013). Recently, Baker et al. (Baker et al., 2016) took a large sample of 804 brands from 55 US industries as the research object to comprehensively analyze the impact of online WOM on consumer purchasing behaviors.
However, existing literature paid little attention to the joint effect of online WOM and other online communication tools (e.g. online advertising) on product sales. In fact, according to Mahajan et al. (1995), consumers have different preferences for different communication media, and this tendency changes with time. For example, Villanueva et al. (2008) used vector autoregressive models to analyze customer value brought about by WOM effects and other traditional marketing communication methods (such as radio, e-mail, etc.). They found that the customer value obtained through different marketing communication methods varies greatly in different stages. Onishi and Manchanda (2012) used a simultaneous equation model to study the impact of blogging activities and advertising on movie box office and found a positive synergy between the two.
This study focused on exploring the joint effect of online WOM and online ads on product sales. According to the summary of Mahajan et al. (1995) on the media effect and new product diffusion theory, different types of communication media (for example, mass media vs interpersonal media) may have different impacts on sales in different stages of the product life cycle. The follow-up question is: Does the dynamic effect of online ads and online WOM, the most common means of mass media and interpersonal media communication under the internet environment, on product sales works the same way? And how does the enterprise-led mass media (online ads) dynamically influence the consumer-led interpersonal media (online WOM)? This study assumed that there is a certain degree of interaction and dynamic linkage among online ads, online WOM and product sales. Figure 1 summarizes the theoretical framework of this study, and shows that the study focused on theoretical deduction in two aspects:
How does the investment in online ads affect the development of online WOM (volume, valence and variance), and how does online advertising affect online WOM dynamically in the product life cycle (early stage, mid-stage and later stage); and
How do online ads and online WOM affect product sales, and how do online ads and online WOM affect sales in the product lifecycle (early stage, mid-stage and later stage)?
2.2 The dynamic impact of online ads on online WOM
Existing research paid little attention to the impact of online ads on online WOM. Gopinath et al. (2013) considered the relationship between the overall expenditure on advertising in various media and online WOM. They found that advertising expenditure in a certain month was positively related to the volume of online WOM next month. This conclusion provided preliminary evidence for the relationship between online ads and online WOM. Recently, Fossen and Schweidel (2016) studied the impact of TV commercials on the WOM volume on social media. They found that television ads helped increase WOM volume of brands and TV programs simultaneously.
This study first analyzed the potential impact of online ads on online WOM from three dimensions (WOM volume, valence and variance). First, Berger and Schwartz (2011) and Lovett et al. (2013) both found that the major drivers for consumers to generate WOM are accessibility and visibility (i.e. whether consumers can easily access and know the products). Many literature on advertising have shown that advertising plays an important role in attracting consumers’ attention and improving product awareness (Aravindakshan and Naik, 2011). Therefore, this study can intuitively infer that online ads have a significant and positive impact on the volume of online WOM. Second, a lot of studies have pointed out that advertising can shape and change consumers’ emotions and attitudes toward products or brands (Miniard et al., 1990; Olney et al., 1991; Brown and Stayman, 1992). Under the influence of advertisements, consumers’ evaluations of products or brands tend to be more positive. Therefore, this study concluded that online ads have a positive effect on the valence of online WOM. Third, Sun (2012) pointed out that the main reason for WOM variance is that consumers in different segmentations have different preferences for products, and thus have different evaluations of products. As a result, product information will reach more consumers in different segmentations when companies’ online ads increase, thus increasing WOM variance. Therefore, this study believed that online ads also have a positive impact on online WOM variance, and put forward the following hypothesis:
Online ads have a positive effect on volume, valence and variance of online word-of-mouth.
The study then analyzed the dynamic changes of online ads with volume, valence and variance of online WOM. First, the impact of online ads on the volume of online WOM gradually weakens over time during the product life cycle. The reason is that, in the early stage of product release, consumers can learn the product information from advertisements and share and discuss with other consumers, so advertising has a strong positive impact on WOM volume in this stage. However, as time goes by, it becomes more difficult for consumers to get new information from advertising, so the impact of ads on WOM volume gradually weakens (Naik et al., 1998). Second, the impact of advertising on WOM valence gradually increases throughout the product life cycle, because it takes time for consumers to learn about the product and have positive comments on it. In the early stage of product release, consumers do not thoroughly understand the product, and the main role of advertising is to let consumers know and understand the product. With the passage of time, most consumers have already been familiar with the product, so the persuasive effect of advertising begins to work, and its impact on WOM valence is gradually emerging. Third, the impact of ads on WOM variance gradually weakens throughout the product life cycle. In the beginning, consumers do not know much about the product. When they first reach the product due to advertisement, their judgments of the product quality are not accurate enough, resulting in a greater variance in WOM. With the passage of time, consumers will have a better understanding of the product, and their judgments of product quality tend to be consistent, so the impact of ads on WOM variance is gradually weakened.
Based on the analysis above, the study put forward the following hypotheses:
The impact of online ads on the online word-of-mouth volume is gradually weakened throughout the product life cycle.
The impact of online ads on online word-of-mouth valence is increasing throughout the product life cycle.
The impact of online ads on online word-of-mouth variance is gradually weakened throughout the product life cycle.
2.3 The dynamically changing impacts of online ads and online word-of-mouth on product adoptions
Online ads and online WOM, as two important marketing communication tools, have impacts on product adoptions.
First, as for online ads, Evans (2008) believed that compared with traditional ads, better results can be achieved by using more personalized data to identify potential customers. In recent years, some empirical studies also began to explore the impact of online ads on customer purchasing behaviors. For example, Chatterjee et al. (2003) found that online ads mainly increase clicks of new visitors. Manchanda et al. (2006) found that online advertising also has a significant impact on increasing repeat purchases of existing customers. Hoban and Bucklin (2015) found that online ads can increase the purchases of new, registered and existing customers at the same time, but have no effect on unregistered visitors. All of these studies indicated that online ads can improve product sales directly or indirectly. Recently, Lewis and Reiley (2014) conducted a large-scale field experiment on Yahoo’s website involving 1.6 million customers to test the causal relationship between the online ads and product sales. The results showed that Yahoo advertising increased product sales by 5 per cent, with 78 per cent of the increase coming from new visitors. Thus, this study assumed that online ads have a similar impact on the adoptions of digital products (e.g. mobile games), and proposed the following hypothesis:
Online ads have a significant and positive impact on digital product adoptions.
When it comes to the dynamic relationship between online ads and sales or adoptions, Kolsarici and Vakratsas (2010), Aravindakshan and Naik (2011) and Bruce (2008) found that the advertising effect in the product life cycle appears in a systematically increasing or decreasing trend. There are mainly two views explaining the dynamic changes of the advertising effect. The first view is that, the main reason is fluctuations in ad investment throughout the product life cycle. Bruce (2008) pointed out that in the movie, music and games industries, companies’ ad investment usually starts before the new product enters the market, peaks during the product launch and gradually declines afterwards. And then, in the middle and later stages of the product life cycle, this investment trend will be repeated. Different levels of advertising investment create different levels of consumer awareness in different stages of the product life cycle, and thus release different product quality signals to consumers, causing dynamic changes in the effect of advertising. The second view is that in the product life cycle, changes in consumers’ psychological effects also cause dynamic changes in the advertising effect (Naik et al., 1998). The first psychological effect is wear-in effect – after advertisements are broadcasted, consumers gradually become more aware of and interested in the product, and the effect of advertising also increases (Campbell and Keller, 2003). The second is wear-out effect – as the product promotion process enters a later stage, consumers no longer get new information from the advertisement, so they gradually get weary and forget the content of the advertisement, weakening the effect of the advertisement (Erdem et al., 2008). The third is restoration effect: as consumers tend to forget the content of ads when it comes to the middle and later stage of the product life cycle, they will regard reappeared ads as new ones and be affected again, restoring the effect of ads (Pechmann and Stewart, 1988). Based on the above analysis, this study speculated that the impact of online advertising on digital product adoptions in the product life cycle appears in a positive U-shaped curve over time. In the early stage, online ads have a strong positive impact on sales or adoptions, which can be attributed to the importance companies attached to advertising input and the wear-in effect of advertising. In the middle stage, the impact of advertising declines because of the decreasing advertising input and the wear-out effect. In the later stage, the advertising effect is reinforced as companies increase advertising input again and because of the restoration effect. Thus, the study put forward the following hypothesis:
The impact of online ads investment on product adoptions appears in a positive U curve throughout the product life cycle.
Then, this study discussed the impacts of online WOM volume, valence and variance on digital product adoptions.
First, it is generally believed that the impact of online WOM volume on adoptions or sales comes from the awareness effect – the more the WOM is, the more the consumers discussing about the product will be, which make other potential consumers more likely to know the product, and thus improve future product sales. Many empirical studies proved that the volume of online WOM has a positive impact on sales. Taking the movie industry as an example, Liu (2006) and Duan et al. (2008) explored the impact of Yahoo’s movie reviews on box office revenue, and found that the more the comments are, the higher the box office revenue of movies will be. Rui et al. (2013) used Twitter’s tweets to measure WOM and found that WOM volume was positively correlated with box office revenue. Dellarocas et al. (2007) added the volume of WOM to the box office prediction model and found that the prediction accuracy of the model significantly improved. In addition, Godes and Mayzlin (2004) studied the television program industry, Chevalier and Mayzlin (2006), Gong et al. (2012, 2013) studied the book industry, Lu and Feng (2009) studied the restaurant industry, Zhu and Zhang (2010) studied the digital games industry, Moe and Trusov (2011) studied online retailers and Ho-Dac et al. (2013) studied the Blue-ray DVD players. All of these studies supported the positive impact of online WOM volume on adoptions.
Second, for WOM valence, it is generally believed that its impact on adoptions mainly comes from persuasive effect – the more the positive comments toward the product are, the easier the consumers will be persuaded to buy. There were also many empirical studies on the relationship between WOM valence and product adoptions. For example, Duan et al. (2009) found that more than 20 per cent of consumers rank products according to other consumers’ ratings when browsing product information on CNET, which highly indicated that WOM is an important factor influencing consumers’ decision-making. Clemons et al. (2006) analyzed the role of WOM in the beer industry and found WOM valence positively correlated with beer sales, and the positive (negative) influence of the first-shown 25 per cent of WOM could best interpret the sales volume. Chintagunta et al. (2010) extended the WOM study in the film industry to multiple regional markets and pointed out that WOM valence is the most important factor affecting box office revenue. Ho-Dac et al. (2013) found that WOM valence only had an impact on products with lower brand equity, instead of products with higher brand equity. Baker et al. (2016) compared the influence of online and offline WOM valence on purchase predisposition and found that offline WOM valence had a greater impact on adoptions than the online one.
Third, there are two opposite views toward the impact of WOM variance on product adoptions. The first view is that WOM variance has a positive impact on sales. Martin et al. (2007), for example, held that high WOM variance would arouse consumers’ curiosity and drive them to purchase. Sun (2012) pointed out that products with high WOM variance are more likely to be preferred by minority consumers. The other view is, for products with higher WOM variance, consumers will perceive greater purchase risks and be less driven to buy the products. Findings of Meyer (1981) supported this view. He found that consumers will lower their evaluation of goods to adapt to inconsistent WOM. Zhang (2006) made an empirical analysis of 128 films released in 2004, and found that in the early stage of the film release, the higher the WOM variance is, the faster the box office revenue decreases.
Based on the above analysis and taking the research scenario into consideration, this study put forward the following hypothesis:
Word-of-mouth volume and valence have a significant positive impact on digital product adoptions, but WOM variance does not necessarily have a significant impact on digital product adoptions.
For the dynamic relationship between online WOM and adoptions or sales, Liu (2006), Trusov et al. (2009) and Moe and Trusov (2011) found that the effect of online WOM also appears in a dynamically upward or downward trend throughout the product life cycle. Relevant research has put forward various viewpoints to explain this phenomenon. Mahajan et al. (1991) argued that most consumers are “innovators” or “experts” with extensive experience in the early stage of a product launch, whereas “followers” or “green hands” that are not familiar with the product tend to buy after a certain period of time. Among these four types of people, “followers” or “green hands” are far less good at distinguishing good products from bad ones compared with “innovators” or “experts,” and thus rely more on WOM. As the proportion of “followers” or “green hands” in consumers increases, the impact of WOM will change. Li and Hitt (53) held that consumers generating WOM are subject to strong self-selection bias. As consumers may have different preferences for products during different periods, WOM volume and valence will increase or decrease dynamically, which will further influence consumers’ purchasing behaviors. Moe and Trusov (2011) found that the dynamic changes in WOM effect in the product life cycle mainly come from the social influence of WOM generators. WOM produced at the beginning of a product launch will have a significant impact on WOM produced later, and have further impact on sales.
Given the above findings, this study assumed that the impact of online WOM on adoptions in the product life cycle appears in an inverted U-shaped curve over time. The impact of online WOM on adoptions is relatively weak in the beginning, increases and peaks in the mid-stage, and declines in the later stage. The explanation to this phenomenon is that, in the earlier product life cycle, there is only little WOM information with low volume, unstable valence and high variance. Early consumers are thus more likely to obtain product information from ads and make purchasing decisions based on their own judgments. Entering the mid-stage, WOM begins to enrich, and is gradually regarded as a more reliable source of information than advertising by consumers. New consumers also usually follow the earlier consumers’ WOM when making purchase decisions. In the later stage, WOM information has been well known to consumers, so its impact will decline. Thus, this study put forward the following hypothesis:
The impact of online word-of-mouth (the overall effect of volume, valence and variance) on adoptions appears in an inverted U-shaped curve throughout the product life cycle.
3.1 Data collection
Data in this paper were provided by Renren Game Company. Founded in 2006, Renren Game is a leading company in R&D and operation of mobile games in China.
First, the study obtained the data of five mobile game products released by Renren Game Company from 2012 to 2014, including the release time of the game, daily downloads, daily online advertising cost and so on. The five products belong to two of the most popular game types in the market at that time – three are role-playing games (RPG) and two are strategy simulation games (SLG). The games were launched in Taiwan and Southeast Asia (including the Philippines, Malaysia, Thailand, Singapore and Vietnam). There are great differences in the development of the mobile game industry between the two regions, and Taiwan has a higher coverage rate of smartphones than Southeast Asia. By choosing these two regions as research objects, both developed and underdeveloped markets can be studied to make the conclusions more universally applicable. In addition, considering the different purchasing powers in these two regions, the study converted the advertising investment cost according to the purchasing power parity index to adapt to different situations.
Second, the online WOM data of the five games were collected by Renren Game Company through self-developed crawler software. Main sources of the WOM data were consumer reviews in Apple Store and Google Play. Promotion and adoptions channels of mobile games are relatively concentrated in Taiwan and Southeast Asia. As Apple Store and Google Play are the two most popular channels, most WOM information is from these two app stores. Taking one game in the study as an example, it had 5,376 online reviews in Apple Store and Google Play, and only 36 and 21 online reviews on Facebook’s homepage and Renren Game Forum during the same time period. Therefore, reviews from the Apple Store and Google Play are very representative. Reviews in these two app stores consist of two parts:
star rating (1 to 5 stars). More stars represent higher evaluation of the consumer; and
comment (a short text review commonly includes user experience and opinions).
When the consumer completes the review, the page will update in real time and show the statistics of reviews, including the number of reviews, average ratings, the ratio of 1 to 5 stars, and the specific ratings and text reviews of every consumer.
The entire data came from 2,497 samples, and contained numbers of game adoptions, advertising investment and WOM collected on a daily basis. The descriptive statistics is shown in Table II. Daily adoptions of each game varied in different periods and reach an average of 872. There was a huge gap between the minimum 3 and the maximum 22,396, and between the mean 872 and the median 201, which indicated that certain games received a large number of adoptions only on certain days. Second, advertising investment and WOM volume also had the same characteristics. The average daily advertising investment and WOM volume were 14,381 Yuan and 4.5 pieces (of WOM) respectively, and there was a big gap between the minimum and maximum values of the two. This phenomenon indicated that there might be a strong correlation between advertising investment, WOM volume and game adoptions. Third, the study used star rating to refer to WOM valence, and calculated the variance of the data to refer to WOM variance. The average WOM valence was 3.64, indicating that consumers had a relatively high evaluation of the game, which was consistent with the findings of previous studies (Chevalier and Mayzlin, 2006). The average WOM variance was 1.78, not much different from the standard deviation 2.14, indicating that the overall scoring of consumers on the game was relatively consistent.
In addition, descriptive statistics were also made by region and game type based on the data collected, and the results were shown in Table III. First, the research found that adoptions in Taiwan were lower than those in Southeast Asia, but the user acquisition input was significantly higher than that in Southeast Asia, which was reflected in the ratio of advertising investment to game adoptions. This phenomenon showed that in Taiwan, with a more developed mobile game industry, the competition between game companies and users was more intense, so the cost of advertising was higher. Other dimensions of WOM data did not vary much between Taiwan and Southeast Asia. From the perspective of game type, SLG were obviously more popular with game companies and consumers than RPG. On the one hand, Renren Game spent twice as much on SLG as RPG. On the other hand, adoptions, WOM volume and the proportion of high-scored WOM of SLG were also significantly higher than those of RPG.
The difference in game adoptions between Taiwan and Southeast Asia was likely to be caused by the marketing and communication activities of companies, or by local cultural, technological, economic and other macro factors. Therefore, the research further collected a series of macro factors (Table IV) that might have an impact on game adoptions and tried to find out the relationship between macro factors and game adoptions. The research found that internet penetration (Internet user ratio), economic development (GDP per capita) and education development degree (literacy rate of population aged above 15) were inversely related to game adoptions in Taiwan and Southeast Asia, and the number of internet users (Internet broadband access number) and the population of adolescents (aged 10-24) were consistent with the number of downloads, but the difference between these numbers was much greater than that of the two adoptions numbers in the two regions. Therefore, from the numerical and logical perspectives, these macro factors were unlikely to become main factors that affect game adoptions.
3.2 Key variables
According to the raw data provided by Renren Game Company, we collected variables for the empirical analysis. Table V summarized and illustrated the variables. It is noteworthy that the linear transformation of natural logarithms was performed for continuous variables with values greater than 0. By doing this, the study can reduce heteroskedasticity and control the influence of outliers.
Dependent variable: In this study, adoptions of mobile games were measured by game downloads. Therefore, we performed natural logarithmic transformation for the dependent variable, namely downloads per day per game.
Independent variables: Independent variables of this paper included online ads and online WOM. First, advertising investment was measured by the actual advertising cost per day per game. As mentioned earlier, as there were two regions involved in the sample, Taiwan and Southeast Asia, the research adjusted their advertising investment according to the local PPP index, and then took the natural logarithm. Then, the research calculated variables related to online WOM based on consumer reviews from Apple Store and the Google Play. Among them, WOM volume was the sum of the number of reviews received in the two stores per day per game; WOM valence was the average scores per day per game in the two stores; and WOM variance was the variance of the scores per day per game in two stores. Natural logarithm transformation was taken for these three WOM variables.
Figure 2 showed the relationship between part of independent variables and the dependent variable. The top-down curves represented advertising investment, game adoptions and WOM volume, respectively. In Figure 2, the three curves shared a strict consistency in both the overall trend and fluctuations. It could be preliminarily inferred that advertising investment and online WOM volume have an impact on game adoptions.
In addition, as WOM might also be influenced by advertising investment, the research showed the relationship between advertising investment and WOM volume [Figure 3(a)], WOM valence [Figure 3(b)] and WOM variance [Figure 3(c)] in Figure 3. The dotted line in Figure 3 represented advertising investment, whereas the solid lines represented WOM volume, WOM valence, and WOM variance respectively. In the graphs, the three curves representing WOM values and the curve representing advertising investment showed strong consistency.
Control variables: To control the potential impact of factors other than advertising and WOM, the research calculated a series of control variables. First, as mobile phone games were released in two regions (Taiwan and Southeast Asia), and there were great differences in the mobile game markets of the two regions, so there might also be differences in advertising and WOM effects. To control the potential impact of regional differences, the research used regional dummy variables to distinguish Taiwan from Southeast Asia. Second, game types might also have an impact on game adoptions. Game types in the data were mainly divided into role-playing games and strategy simulation games, which covered more than 80 per cent of the popular games in the market. The popularity of different types of mobile games among consumers might vary because of different themes, highlights and operating rules. Therefore, the study used game-type dummy variables to distinguish between RPG and SLG to control the impact of game type on results. Third, since the times of game release varied much, seasonal factors such as holidays and winter and summer vacations might result in differences in game adoptions. Therefore, the research used quarterly dummy variables to control the impact of time factors on game adoptions. Specifically, three dummy variables were generated based on the first quarter (January-March), representing the second quarter (March-June), the third quarter (June-September) and the fourth quarter (September-December).
4. Empirical model
4.1 Model specification
This paper took two impacts into consideration when building the empirical model: the impact of advertising on WOM and the joint impact of advertising and WOM on game adoptions. As the two effects exist simultaneously and are inter-related, the most ideal analysis method is to consider the two effects simultaneously by establishing simultaneous equation models (Greene, 2008). Based on the model of Onishi and Manchanda (2012), this research proposed the following linear simultaneous equations model:
In the model above, subscript i: 1,… N stands for games, subscript t: 1,… N stands for time.
Equations (1)-(3) investigated the impact of online ads on online WOM. Specifically, Equations (1)-(3) were set up with the WOM volume (lnVolit), WOM valence (lnValit) and WOM variance (lnVarit) as dependent variables. On the right side of the three equations, the variable lnAdit was first added to analyze the impact of advertising on WOM volume, valence and variance. In addition, the model also used lagged variables lnVolit-1, lnValit-1, lnVarit-1 as explanatory variables, because previous studies had found that online WOM can be affected by previous WOM (Moe et al., 2011). The research did not add a lagged variable for online advertising, because the impact of advertising on online WOM occurs rapidly in a very short period of time, and there is no lag effect in terms of days (Fossen and Schweidel, 2016). In addition, the research added three control variables: Geoi, the dummy variable of the game release area (Southeast Asia = 0, Taiwan of China = 1), Genrei, the dummy variable of game type (SLG = 0, RPG = 1) and Seat, the quarterly dummy variable.
Equation (4) investigated the joint impact of online WOM and online ads on downloads. In equation (4), the dependent variable lnDlit represented downloads of game i on day t. The independent variable lnAdit represented advertising investment of game i on day t, lnVolit represented WOM volume of game i on day t, lnValit represented WOM valence of game i on day t, and lnVarit represented WOM variance of game i on day t. Then, this study added Geoi, Genrei and Seat to control region differences, game type differences and seasonal factors. In addition, as existing literature showed that product sales in the previous period will have an impact on that in the current period (Arellano and Bond, 1991), this study added the lag term lnDlit−1 of game adoptions to control the lagging effect.
This study used two-stage least squares to estimate the simultaneous equation model above. In the first stage, this study used exogenous variables lnAdit, lagged variables lnVolit−1, lnValit−1, lnVarit−1 and control variables Geoi, Genrei and Seat to estimate the WOM variables lnVolit, lnValit and lnVarit in equations (1) to (3) to analyze the impact of online advertising on online WOM. In the second stage, this study used the predicted values of the first stage WOM variables to estimate equation (4) and investigate the joint impact of online WOM and online advertising on sales or adoptions. Since both dependent variables and independent variables were transformed by natural logarithms, the estimates of regression coefficients, i.e. the effect of independent variable variance ratio on the dependent variable variance ratio, were elastic.
4.2 Impact of online ads on online word-of-mouth and dynamic changes
To analyze the overall impact of online advertising on online WOM, this study used the overall sample to perform regression analysis of Equations (1)-(3). Table VI showed the results of regression analysis. The goodness of fit of the model, R2, was 0.75, 0.48 and 0.33, respectively, which showed that advertising investment best explained WOM volume, followed by WOM valence, and then WOM variance. Overall, the fitting results of all models were very significant.
The results in Table VI Column (1) reflect the impact of advertising investment on WOM volume. The coefficient of lnAdit was positively significant (b1 = 0.026, p < 0.01), indicating that the more the advertising investment was, the more WOM volume obtained by the game would be. In addition, early WOM also had an impact on WOM volume. The coefficients of lnVolit−1 and lnValit−1 were both positive and significant (b2 = 0.857, p < 0.01; b3 = 0.110, p < 0.01), indicating that the more WOM volume and the better WOM valence were, the more WOM volume would be obtained in the current period. The coefficient of lnVarit−1 was negative and significant (b4 = –0.152, p < 0.01), indicating that the smaller WOM variance was in the earlier period, the smaller WOM volume would be in the current period.
The results in Table VI Column (2) reflect the impact of advertising investment on WOM valence. The coefficient of lnAdit was positive and significant (c1 = 0.015, p < 0.01), indicating that the more the advertising investment was, the higher the WOM valence would be. In addition, the coefficients of lnVolit−1 and lnValit−1 were both positive and significant (c2 = 0.165, p < 0.01; c3 = 0.352, p < 0.01), indicating that the more WOM volume and better WOM valence were, the higher the game’s WOM valence would be in the current period. The coefficient of lnVarit−1 was not significant (c4 = –0.000, p > 0.10), indicating that the earlier WOM variance had no significant effect on the WOM valence in the current period.
The results in Table VI Column (3) reflect the impact of advertising investment on WOM variance. The coefficient of lnAdit was positive and significant (d1 = 0.024, p < 0.01), indicating that the more the advertising investment was, the greater the WOM variance would be. In addition, the coefficients of lnVolit−1, lnValit−1 and lnVarit−1 were all positive and significant (d2 = 0.129, p < 0.01; d3 = 0.073, p < 0.01; d4 = 0.110, p < 0.01), indicating that the volume, valence and variance of WOM in the earlier period had positive effects on WOM variance in the current period.
Overall, the above regression results supported H1 proposed earlier.
Then, to study the dynamic changes in the impact of online advertising on online WOM, the study divided all samples into several sub-samples in different stages of the product life cycle, and then made regression analysis on each sub-sample.
Table VII showed the average life cycle of mobile games provided by Renren Game Company. According to relevant literature (Zhu and Zhang, 2010), the life cycle of game products has two characteristics: firstly, the whole life cycle of a game is usually relatively short, and the update and iteration speed is very fast; secondly, the number of users of a new game will accumulate rapidly after product launch, and the product will pass the early stage and enter the mid-stage quickly. As you can see from the table, the average life cycle of mobile games is about one year. After data validation, the length of this life cycle also conforms to life cycles of the five mobile games analyzed in this study. As mentioned earlier, mobile games’ life cycle is characterized by a rapid transition from introduction to maturity, so this study divided the life cycle into three stages: the early stage, the mid-stage and the later stage. The specific criteria are as follows.
Early stage: About one month starting from the game launch to the end of the first month. During the early stage, early users of the game begin to accumulate. Game companies usually increase advertising awareness at this stage through increasing advertising investment. At the same time, this is the initial stage of Online WOM.
Mid-stage: About five months starting from the second month to the sixth month after game launch. During the mid-stage, the user number and activity of the game increase rapidly and reach the peak. The advertising investment of game companies gradually decreases, WOM volume increases rapidly, and WOM valence becomes more stable.
Later stage: Usually six months starting from the seventh month to game exiting from the market. During the later stage, users of games begin to decrease, and user activities decline significantly. Game companies restart a certain amount of advertising investment. Increase of WOM volume slows down, and WOM valence almost doesn’t change any longer.
Based on the above criteria, this study divided the overall sample into three sub-samples: samples at early stage, mid-stage and later stage. The study also used two-stage least squares regression method to process the sub-samples. Table VIII showed the regression results from Equations (1)-(3). Among them, the dependent variable in Columns (1)-(3) is lnVolit, corresponding to equation (1) in the model; the dependent variable in Columns (4)-(6) is lnValit, corresponding to equation (2) in the model; the dependent variable in Columns (7)-(9) is lnVarit, corresponding to equation (3) in the model. In Table VIII, this study focused on the estimated coefficients of the variable lnAdit in each column, because the variation of lnAdit coefficients could reflect dynamic changes in the impact of advertising on WOM in different stages of the game’s life cycle.
The regression results from Columns (1)-(3) reflect the dynamic impact of online advertising on WOM volume. By comparing the regression results of the three columns, this study found that the positive effect of advertising on WOM volume decreased gradually as the life cycle developed. The estimated coefficient of lnAdit was 0.042 (p < 0.01) in the early stage, 0.028 (p < 0.01) in the mid-stage and 0.012 (p < 0.01) in the later stage. The result supported H2a proposed earlier.
The regression results from Columns (4)-(6) reflect the dynamic impact of online advertising on WOM valence. This study found that the positive effect of advertising on WOM valence was gradually strengthened as the life cycle developed, which is just opposite to the result of WOM volume. In the early stage, the estimated coefficient of lnAdit was 0.009, and only marginally significant (p < 0.10); in the mid-stage, the estimated coefficient of lnAdit increased to 0.017 (p < 0.01); and in the later stage, the estimated coefficient of lnAdit further increased to 0.026 (p < 0.01). The result supported H2b proposed earlier.
The regression results from Columns (7)-(9) reflect the dynamic impact of online advertising on WOM variance. This study found that the impact of advertising on WOM variance also gradually diminished as the life cycle developed. The estimated coefficient of lnAdit was 0.034 (p < 0.01) in the early stage, 0.024 (p < 0.01) in the mid-stage and 0.018 (p < 0.01) in the later stage. The result supported H2c proposed earlier.
Figure 4 showed the dynamic impact of online advertising on online WOM more clearly. Throughout the product life cycle, the impact of advertising on WOM volume and WOM variance gradually weakened, yet the impact on WOM valence gradually strengthened.
4.3 Impact of online ads and online word-of-mouth on product adoptions and dynamic changes
To analyze the overall impact of online advertising and online WOM on adoptions, this study used the overall sample to make regression analysis of equation (4), and the results are shown in Table IX. The study used stepwise regression method to analyze equation (4). In the beginning, this study only made regression analysis of dependent variables and control variables. Then, on this basis, the study gradually added in the regression analysis of the independent variable related to advertising and WOM.
Table IX Column (5) shows the regression results of the complete model, which illustrated the relationship between variables. First, this study found that the lnAdit coefficient was positive and significant (a1 = 0.066, p < 0.01), indicating that the more the advertising investment was, the higher the number of downloads would be. Second, the coefficients of lnVolit and lnValit were both positive and significant (a2 = 0.105, p < 0.01; a3 = 0.149, p < 0.10), indicating that the volume and valence of WOM both had a significant and positive impact on game adoptions, while the coefficient of lnVarit was negative and insignificant (a4 = 0.126, p > 0.10), indicating that WOM variance didn’t have a significant impact on game adoptions. Regression results above supported H3a and H4a proposed earlier.
In addition, for control variables, the coefficient of Geoi was very small and insignificant (a5 = 0.039, p > 0.10), indicating that there was no significant difference in game adoptions between Taiwan and Southeast Asia; the coefficient of Genrei was significant and positive (a6 = 0.142, p < 0.01), indicating that role-playing games were more popular with consumers than strategy simulation games. As Geoi coefficient was not significant, this study removed this variable, and made another regression analysis of the model to see whether the results would change. Column (6) shows the regression results after Geoi was kicked out. Overall, the coefficient estimates of variables did not change significantly.
To study the dynamic changes of the impact of online advertising and online WOM on game adoptions, this study divided the overall sample into three sub-samples, namely samples at early stage, mid-stage and later stage, and then made regression analysis of each sub-sample. Table X shows the relevant regression results.
The regression results in Table X show that advertising had a significant and positive impact on game adoptions throughout the life cycle, which was consistent with the regression results in Table IV. The coefficient of lnAdit was 0.157 (p < 0.01) in the introducing stage, 0.056 (p < 0.01) in the mature stage and 0.166 (p < 0.01) in the declining stage, which showed that, throughout the product life cycle, the impact of online advertising on game adoptions was a positive U-shaped curve. This result supported H3b proposed earlier.
In addition, online WOM variables played different roles in different stages of the game life cycle. In the early stage, the coefficient of lnVolit was positive and significant (a2 = 0.842, p < 0.01), whereas the coefficients of lnValit and lnVari were both insignificant (a3 = –0.724, p > 0.10; a4 = 0.736, p > 0.10), indicating that only WOM volume had an impact on game adoptions in the early stage. In the mature stage, coefficients of lnVolit, lnValit and lnVarit were all positive and significant (a2 = 0.237, p < 0.01; a3 = 1.457, p < 0.01; a4 = 0.170, p < 0.01), indicating that the overall impact of WOM on game adoptions increased significantly. In the later stage, only lnVolit coefficient was positive and significant (a2 = 0.517, p < 0.05), whereas lnValit and lnVari coefficients were insignificant (a3 = –0.242, p > 0.10; a4 = –0.192, p > 0.10), indicating that the overall impact of WOM on game adoptions was also declining. This showed that throughout the product life cycle, the impact of online WOM game adoptions was an inverted U-shaped curve, which supported H4b proposed earlier.
Figure 5 shows more clearly the dynamic impact of online WOM and online ads on game adoptions. The solid and dotted lines represented the dynamic effects of WOM and advertising on downloads, respectively. Based on the relationship proposed in Figure 5, the impacts of online WOM and online ads on game adoptions complemented each other during the product life cycle: in the early and later stages, advertising had a stronger impact on adoptions, whereas WOM was less effective in increasing adoptions. On the contrary, in the mid-stage, advertising had a weakened impact on adoptions, whereas WOM became more effective.
5. Conclusion and discussion
This study discussed the dynamically changing impacts of online WOM and online ads on product adoptions from the perspective of integrated marketing communication. A simultaneous equation model was built using data collected from five mobile games launched by Renren Game Company from 2012 to 2014. Based on the analysis of the model, the research drew the following conclusions:
First, the research studied the impact of online ads on online WOM, and dynamic changes in the impact. As the first step, the study made a regression analysis of the overall sample. Regression results indicated that online ads had a significant and positive impact on WOM volume, WOM valence and WOM variance. In the second step was, the study divided the sample into three sub-samples in accordance with the mobile game life cycle: the early stage, the mid-stage and the later stage. Results of the regression analysis of sub-samples indicated that the influence of online ads on WOM volume and WOM variance gradually weakened, but the influence on WOM valence gradually strengthened.
Second, the research studied the joint impact and dynamic changes of both online WOM and online ads on product adoptions. In the first step, regression results of the overall sample indicated that WOM volume, WOM valence and online ads all had a significant and positive impact on the game adoptions, whereas WOM variance had no significant impact. In the second step, the whole sample was divided into three sub-samples according to their stages in the life cycle. Regression results indicated that the dynamic impact of online WOM and online ads on product adoptions complemented each other throughout the product life cycle. The impact of WOM on sales appeared in an inverted U-shape curve: WOM was less effective in encouraging adoptions in the early stage, yet its impact increased and peaked in the mid-stage, and weakened in the later stage. On the contrary, the dynamic impact of online ads on product adoptions was a positive U-shaped curve: Advertising was more effective in encouraging adoptions in the early stage, yet its impact gradually weakened in the mid-stage, and then restored in the later stage.
Theoretical contributions of this paper mainly included the following four aspects. First, the study brought online advertising and online WOM into the research framework of digital marketing communication, and studied the relationship among mass media (online ads), interpersonal media (online WOM) and product adoptions in the context of the Internet. The research results greatly enriched the theory of digital marketing communication. Second, the study found that the impacts of online ads and online WOM on product adoptions varied in different stages of the product life cycle. This conclusion provides an empirical basis for the inference of the dynamic development of media effect in the new product diffusion theory (Bass, 1969; Mahajan et al., 1995; Rogers, 2003). Third, the study found that the impacts of online ads and online WOM on product adoptions complemented each other – the impact of online ads appeared in a “first decline and then rise” U-shaped curve, yet the impact of online WOM appeared in a “first rise and then fall” inverted U-shaped curve. This conclusion could be an answer to the debate in the theoretical literature on whether the relationship between the two impacts is complementary or alternative (Villanueva et al., 2008; Onishi and Manchanda, 2012). Fourth, the study also found that online ads had an impact on WOM volume and WOM valence, and the impact on the WOM volume gradually decreased, whereas the impact on WOM valence became more salient over time. Results of this study also enriched the research on the antecedents and dynamic effects of online WOM to some extent.
5.2 Managerial implications
Online ads and online WOM are the two most representative marketing communication methods in the digital era. It is a major concern for managers to effectively integrate different types of marketing communication tools and carry out targeted marketing communication in different stages of the product life cycle. According to the conclusion of this study, this paper offered three marketing suggestions for companies:
First, companies should adjust the investment strategy of online ads and online WOM in different stages of the product life cycle, so as to increase product adoptions by using limited marketing resources. Our empirical results indicated that online ads have a positive U-shaped effect on product adoptions throughout the product life cycle, whereas online WOM has an inverse U-shaped effect on product adoptions, indicating that the two effects on product adoptions are heterogeneous in different stages, and can be complementary to each other. Therefore, companies should adjust the investment strategy of online ads and online WOM in different stages to make best use of these two marketing communication tools. For example, companies can adopt the “high-low-high” advertising investment strategy and the “low-high-low” WOM management strategy throughout the product life cycle. By doing this, companies would be able to achieve the goal of increasing product adoptions with limited marketing resources by strengthening the effect of online ads in the early and later stages and the effect of online WOM in the mid-stage.
Second, the interaction between various online marketing communication tools should be fully considered to monitor and optimize the overall marketing communication effect. This study found that online ads have an impact on product adoptions through two paths. On the one hand, online ads have a direct impact on product adoptions by directly acting on consumers. On the other hand, online ads also have an indirect impact on product adoptions by increasing the volume and valence of online WOM. Therefore, companies should not only pay attention to the direct impact of online advertising, but should also notice the important role that online WOM plays in the process. In the early stage after product launch, online ads have a more salient impact on WOM volume but less on WOM valence, so companies should consider other more effective methods (e.g. expert recommendation, trial, sharing, etc.) to encourage consumers to make more positive comments, so as to persuade other consumers to buy. In the middle and later stages, the impact of online ads on WOM volume will gradually decrease, so companies should focus on other marketing strategies that can enhance the WOM volume (e.g. reviews or forwarding incentives, etc.) to ensure sustainable effect of WOM.
Third, companies should ensure the sustainability of online ads and online WOM investment throughout the product life cycle, so as to achieve sustainable marketing communication effect. The research found that online ads and WOM have significant impacts on product adoptions throughout the product life cycle. In addition, the volume of advertising investment and WOM volume in the early stage will have different degrees of positive impacts on the follow-up marketing and communication. Therefore, companies should consider the sustainability of online advertising investment and online WOM management when making marketing communication strategies. Marketing managers should also adjust strategies of advertising investment and online WOM management in a planned and sustainable way according to the characteristics of different marketing communication tools in different stages of the product life cycle, so as to generate a cumulative effect and achieve better marketing outcomes.
Last but not least, this study held the following prospects for future research. First, this study was based on the mobile game industry, so future research can consider more research objects in other industries to enhance the universality of research conclusions. Second, this study analyzed the impact of online advertising investment on online WOM and product adoptions. However, apart from the amount of ad investment, the content (e.g. rational, emotional, etc.) and forms (e.g. online and offline ads) of advertising are also important factors affecting online WOM and product adoptions, which should be taken into consideration in future research. Third, future research can further consider the interactive effects of ads and WOM, and the dynamic changes of the effects throughout the product life cycle, so as to reveal the synergy of these two factors on product adoptions. Fourth, product adoptions might be affected by both the marketing communication activities of companies and the local macro environments in different countries and regions. For example, in this research, the number of game adoptions might also be affected by the degree of Internet penetration in the region, the number of Internet users, the number of adolescents, the level of economic development and other macroeconomic factors. Therefore, if the panel data from various countries and regions can be obtained in future studies, the interaction between marketing communication activities and regional macro factors can be further studied to provide a theoretical basis for companies to formulate marketing communication strategies in line with local conditions.
Relevant literature and incremental contributions of this study
|Author(s)||Research domain||Key variables of WOM|
|Godes and Mayzlin (2004)||TV program||Volume, dispersion|
|Chevalier and Mayzlin (2006)||Book||Volume, valence|
|Clemons et al. (2006)||Beer||Valence, variance|
|Dellarocas et al. (2007)||Movie||Volume, valence|
|Duan et al. (2008)||Movie||Volume|
|Lu and Feng (2009)||Restaurant||Volume, valence|
|Chintagunta et al. (2010)||Movie||Valence, variance|
|Zhu and Zhang (2010)||Game||Volume, valence, variance|
|Moe and Trusov (2011)||Online retail||Volume, valence, variance|
|Sun (2012)||Book||Valence, variance|
|Gong et al. (2012, 2013)||Book||Volume, valence, variance|
|Rui et al. (2013)||Movie||Volume, valence|
|Gopinath et al. (2013)||Movie||Volume, valence|
|Ho-Dac et al. (2013)||Blue-ray DVD player||Volume, valence|
|Baker et al. (2016)||Various industries||Valence|
2,497 samples in total
Descriptive statistics (by different regions and game types)
|Variables||Taiwan, China||Southeast Asia||SLG||RPG|
|Proportion of high-scored WOM (>4 stars) (%)||67.3||62.4||77.3||56.4|
The data in the table are shown in terms of mean value
Macro-economic and demographic factors (by difference regions)
|Factors||Variables||Taiwan, China||Southeast Asia|
|Internet penetration||Internet user ratio (%)||80.00||51.56|
|Internet user number||Internet Broadband access number (Million)||7.0||23.6|
|Youth population||Population aged 10-24 (Million)||4.5||78.2|
|Economic development||GDP per capita (Dollar)||41,538||25,590|
|Education development||Literacy rate of population aged above 15 (%)||98.50||95.78|
Southeast Asia includes five countries: the Philippines, Malaysia, Thailand, Singapore and Vietnam. The numbers of internet penetration, internet cost, economic development and education development in Southeast Asia were mean values, whereas the numbers of internet users and adolescents were sums
|lnDlit||Adoptions per day per game||After natural logarithmic transformation|
|lnAdit||Advertising investment per game per day||After natural logarithmic transformation|
|lnVolit||Online WOM volume per game per day||After natural logarithmic transformation|
|lnValit||Online WOM valence per game per day||After natural logarithmic transformation|
|lnVarit||Online WOM variance per game per day||After natural logarithmic transformation|
|Geoi||Region dummy variable||Southeast Asia = 0, Taiwan = 1|
|Genrei||Game type dummy variable||SLG = 0, RPG = 1|
|Seat||Quarterly dummy variable||Three dummy variables were generated based on the first quarter (January-March), representing the second quarter (March-June), the third quarter (June-September) and the fourth quarter (September-December).|
Impacts of online ads on online WOM
|Variables||DV: lnVolit (1)||DV: lnValit (2)||DV: lnVarit (3)|
|lnAdit||0.026 (0.003)***||0.015 (0.003)***||0.024 (0.003)***|
|lnVolit−1||0.857 (0.018)***||0.165 (0.019)***||0.129 (0.018)***|
|lnValit−1||0.110 (0.022)***||0.352 (0.023)***||0.073 (0.022)***|
|lnVarit−1||−0.152 (0.022)***||−0.000 (0.022)||0.110 (0.021)***|
|Geoi||−0.026 (0.022)||0.012 (0.023)||−0.052 (0.021)**|
|Genrei||−0.224 (0.028)***||−0.233 (0.029)***||−0.226 (0.028)***|
|Constant term||0.296 (0.033)***||0.400 (0.034)***||0.269 (0.032)***|
Values in brackets were standard errors; **p < 0.05 and ***p <0.01
Typical life cycle of mobile games
|Game type||Average life||Early stage||Mid-stage||Later stage|
|Mobile game||1 Year||1 Month||5 Months||6 Months|
Dynamic impact of online ads on online WOM
|Variables||DV: lnVolit||DV: lnValit||DV: lnVarit|
|(1) Early stage||(2) Mid-stage||(3) Later stage||(4) Early stage||(5) Mid-stage||(6) Later stage||(7) Early stage||(8) Mid-stage||(9) Later stage|
|lnAdit||0.042 (0.010)***||0.028 (0.004)***||0.012 (0.005)**||0.009 (0.005)*||0.017 (0.004)***||0.026 (0.009)***||0.034 (0.008)***||0.024 (0.004)***||0.018 (0.006)***|
|lnVolit−1||0.743 (0.060)***||0.817 (0.026)***||0.343 (0.056)***||0.061 (0.034)*||0.184 (0.024)***||0.210 (0.076)***||−0.085 (0.044)*||0.112 (0.025)***||0.130 (0.063)**|
|lnValit−1||−0.038 (0.116)||0.007 (0.036)||0.041 (0.037)||0.578 (0.066)***||0.372 (0.033)***||0.192 (0.051)***||0.250 (0.087)***||0.147 (0.035)***||0.012 (0.042)|
|lnVarit−1||−0.196** (0.081)||−0.156 (0.031)***||−0.057 (0.033)*||−0.038 (0.046)||−0.006 (0.028)||−0.031 (0.045)||0.131 (0.061)**||0.115 (0.030)***||0.035 (0.037)|
|Geoi||−0.070 (0.076)||−0.026 (0.034)||0.064 (0.034)*||−0.023 (0.043)||0.038 (0.032)||0.109 (0.046)**||−0.129 (0.058)**||−0.081 (0.034)***||0.013 (0.038)|
|Genrei||−0.612 (0.117)***||−0.093 (0.040)**||−0.271 (0.046)***||−0.362 (0.066)***||−0.154 (0.038)***||−0.467 (0.063)***||−0.592 (0.088)***||−0.162 (0.039)***||−0.168 (0.052)***|
|Constant term||0.292 (0.201)||0.243 (0.052)***||0.317 (0.047)***||0.246 (0.113)**||0.374 (0.049)***||0.551 (0.064)**||0.409 (0.151)***||0.275 (0.051)***||0.195 (0.053)***|
Value in brackets were standard errors, *p < 0.10, **p < 0.05, ***p <0.01
Impact of online ads and online WOM on game adoptions
|lnAdit||0.068 (0.004)***||0.064 (0.004)***||0.064 (0.004)***||0.066 (0.005)***||0.064 (0.005)***|
|lnVolit||0.132 (0.024)***||0.093 (0.033)***||0.105 (0.038)***||0.098 (0.037)***|
|lnValit||0.110 (0.065)*||0.149 (0.090)*||0.104 (0.051)*|
|lnVarit||−0.126 (0.201)||−0.007 (0.180)|
|Geoi||−0.019 (0.027)||−0.035 (0.026)||−0.027 (0.025)||−0.031 (0.025)||−0.039 (0.028)|
|Genrei||−0.127 (0.031)***||0.040 (0.031)||0.133 (0.034)***||0.163 (0.036)***||0.142 (0.051)***||0.161 (0.049)***|
|lnDlit−1||0.884 (0.008)***||0.783 (0.010)***||0.746 (0.012)***||0.738 (0.012)***||0.740 (0.013)***||0.739 (0.013)***|
|Constant term||0.644 (0.056)***||0.872 (0.054)***||0.916 (0.053)***||0.891 (0.055)***||0.740 (0.013)***||0.875 (0.054)***|
lnDlit, namely, downloads of a single game in a single day, was the dependent variable. Values in brackets were standard errors, *p < 0.10, ***p <0.01
Dynamic impact of online ads and online WOM on game adoptions
|Variables||(1) Early stage||(2) Mid-stage||(3) Later stage|
|lnAdit||0.157 (0.032)***||0.056 (0.006)***||0.166 (0.007)***|
|lnVolit||0.842 (0.251)***||0.237 (0.031)***||0.517 (0.231)**|
|lnValit||−0.724 (0.600)||1.457 (0.141)***||−0.242 (0.213)|
|lnVarit||0.736 (0.811)||0.170 (0.022)***||−0.192 (0.491)|
|Genrei||0.882 (0.557)||0.168 (0.054)***||0.036 (0.082)|
|lnDlit−1||0.403 (0.041)***||0.743 (0.024)***||0.904 (0.024)***|
|Constant term||1.815 (0.411)***||0.817 (0.090)***||0.328 (0.070)***|
Notes: lnDlit, namely, downloads of a single game in a single day, was the dependent variable. Values in brackets were standard errors; **p < 0.05, ***p <0.01
This study has reached a confidentiality agreement with Renren Game Company, so the names of the five games could not be revealed in this article.
According to statistics released by International Monetary Fund (IMF) in 2011, based on the US dollar purchasing power = 1 as the purchasing power parity index benchmark, China (RMB) was 1.550, Taiwan (TWD) was 1.879, the Philippines (PHP) was 1.837, Thailand (Thai baht) was 1.742, Malaysia (Ringgit) was 1.666, Singapore (Singapore dollar) was 1.213 and Vietnam (VND) was 2.447. First, we calculated the average purchasing power parity index of currencies in Southeast Asia. .Then, the advertising investment cost was converted according to the purchasing power parity index of the corresponding region, that is, the advertising investment cost. (
Statistics were from results of internal market research provided by Renren Game Company.
Statistics from Renren Game Company on the company’s and domestic mobile game products.
Another regression method is to introduce dummy variables representing different stages (early stage, mid-stage and later stage) in the same regression model, and then set up an interactive term between dummy variables and independent variables to observe and compare the coefficients in different stages. This study also used this method for robustness test and got the same result. Due to the limitation of length, the results of robustness test were not included in the article, and readers interested in the robustness testing are welcome to write to us and ask for results.
Here, we used the dependent variable and independent variables in the same period to conduct regression analysis. In addition, considering that the impact of advertising and WOM may be hysteretic, this study also used the independent variable of lag1 variable to make a regression analysis of equation (4) to verify the robustness of the regression results. The regression results stayed consistent.
This study used the model in Column (6) of Table IX for regression analysis, which removed the dummy variable Geoi because it was insignificant.
In the study, the value of WOM effect was the sum of the coefficients of lnVolit., lnValit and lnVarit, aiming to reflect the cumulative effect of these three dimensions of online WOM. Since the estimation results of the coefficients of lnValit and lnVarit were not significant, this study set the values of these two coefficients to 0.
Ambler, T. and Vakratsas, D. (1999), “How advertising works: what do we really know?”, Journal of Marketing, Vol. 63 No. 1, pp. 26-43.
Aravindakshan, A. and Naik, P.A. (2011), “How does awareness evolve when advertising stops? The role of memory”, Marketing Letters, Vol. 22 No. 3, pp. 315-326.
Arellano, M. and Bond, S. (1991), “Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations”, The Review of Economic Studies, Vol. 58 No. 2, pp. 277-297.
Baker, A.M., Donthu, N. and Kumar, V. (2016), “Investigating how word-of-mouth conversations about brands influence purchase and retransmission intentions”, Journal of Marketing Research, Vol. 53 No. 2, pp. 225-239.
Bass, F.M. (1969), “A new product growth for model consumer durables”, Management Science, Vol. 15 No. 5, pp. 215-227.
Berger, J. and Schwartz, E.M. (2011), “What drives immediate and ongoing word of mouth?”, Journal of Marketing Research, Vol. 48 No. 5, pp. 869-880.
Brown, S.P. and Stayman, D.M. (1992), “Antecedents and consequences of attitude toward the ad: a meta-analysis”, Journal of Consumer Research, Vol. 19 No. 1, pp. 34-51.
Bruce, N.I. (2008), “Pooling and dynamic forgetting effects in multitheme advertising: tracking the advertising sales relationship with particle filters”, Marketing Science, Vol. 27 No. 4, pp. 659-673.
Campbell, M.C. and Keller, K.L. (2003), “Brand familiarity and advertising repetition effects”, Journal of Consumer Research, Vol. 30 No. 2, pp. 292-304.
Chatterjee, P., Hoffman, D.L. and Novak, T.P. (2003), “Modeling the clickstream: implications for web-based advertising efforts”, Marketing Science, Vol. 22 No. 4, pp. 520-541.
Chen, Y. and Xie, J. (2008), “Online consumer review: word-of-mouth as a new element of marketing communication mix”, Management Science, Vol. 54 No. 3, pp. 477-491.
Chevalier, J.A. and Mayzlin, D. (2006), “The effect of word of mouth on sales: online book reviews”, Journal of Marketing Research, Vol. 43 No. 3, pp. 345-354.
Chintagunta, P.K., Gopinath, S. and Venkataraman, S. (2010), “The effects of online user reviews on movie box office performance: accounting for sequential rollout and aggregation across local markets”, Marketing Science, Vol. 29 No. 5, pp. 944-957.
Clemons, E.K., Gao, G.G. and Hitt, L.M. (2006), “When online reviews meet hyperdifferentiation: a study of the craft beer industry”, Journal of Management Information Systems, Vol. 23 No. 2, pp. 149-171.
Dellarocas, C., Zhang, X.M. and Awad, N.F. (2007), “Exploring the value of online product reviews in forecasting sales: the case of motion pictures”, Journal of Interactive Marketing, Vol. 21 No. 4, pp. 23-45.
Derbaix, C. and Vanhamme, J. (2003), “Inducing word-of-mouth by eliciting surprise–a pilot investigation”, Journal of Economic Psychology, Vol. 24 No. 1, pp. 99-116.
Duan, W., Gu, B. and Whinston, A.B. (2008), “Do online reviews matter? An empirical investigation of panel data”, Decision Support Systems, Vol. 45 No. 4, pp. 1007-1016.
Duan, W., Gu, B. and Whinston, A.B. (2009), “Informational cascades and software adoption on the internet: an empirical investigation”, MIS Quarterly, Vol. 33 No. 1, pp. 23-48.
Erdem, T., Keane, M.P. and Sun, B. (2008), “A dynamic model of brand choice when price and advertising signal product quality”, Marketing Science, Vol. 27 No. 6, pp. 1111-1125.
Evans, D.S. (2008), “The economics of the online advertising industry”, Review of Network Economics, Vol. 7 No. 3, pp. 359-391.
Fossen, B.L. and Schweidel, D.A. (2016), “Television advertising and online word-of-mouth: an empirical investigation of social TV activity”, Marketing Science, Vol. 36 No. 1, pp. 105-123.
Godes, D. and Mayzlin, D. (2004), “Using online conversations to study word-of-mouth communication”, Marketing Science, Vol. 23 No. 4, pp. 545-560.
Gong, S., Liu, X. and Liu, Y. (2012), “Does online word-of-mouth determine product’s fate? An empirical analysis of online book reviews”, Nankai Business Review, Vol. 15 No. 4, pp. 118-128.
Gong, S., Liu, X. and Zhao, P. (2013), “How does online word of mouth influence product sales? An empirical study based on online book reviews”, China Soft Science, No. 6, pp. 171-183.
Gopinath, S., Chintagunta, P.K. and Venkataraman, S. (2013), “Blogs, advertising, and Local-Market movie box office performance”, Management Science, Vol. 59 No. 12, pp. 2635-2654.
Greene, W.H. (2008), Econometric Analysis, 6th ed., Pearson Prentice Hall, Upper Saddle River, NJ.
Hoban, P.R. and Bucklin, R.E. (2015), “Effects of internet display advertising in the purchase funnel: model-based insights from a randomized field experiment”, Journal of Marketing Research, Vol. 52 No. 3, pp. 375-393.
Ho-Dac, N.N., Carson, S.J. and Moore, W.L. (2013), “The effects of positive and negative online customer reviews: do brand strength and category maturity matter?”, Journal of Marketing, Vol. 77 No. 6, pp. 37-53.
Hu, H. (2008), “Consumer behavior determined by word of mouth”, China Business News.
Jansen, B.J., Zhang, M., Sobel, K. and Chowdury, A. (2009), “Twitter power: tweets as electronic word of mouth”, Journal of the American Society for Information Science and Technology, Vol. 60 No. 11, pp. 2169-2188.
Kolsarici, C. and Vakratsas, D. (2010), “Category- versus brand-level advertising messages in a highly regulated environment”, Journal of Marketing Research, Vol. 47 No. 6, pp. 1078-1089.
Lewis, R.A. and Reiley, D.H. (2014), “Online ads and offline sales: measuring the effect of retail advertising via a controlled experiment on Yahoo!”, Quantitative Marketing and Economics, Vol. 12 No. 3, pp. 235-266.
Liu, Y. (2006), “Word of mouth for movies: its dynamics and impact on box office revenue”, Journal of Marketing, Vol. 70 No. 3, pp. 74-89.
Lovett, M.J., Peres, R. and Shachar, R. (2013), “On brands and word of mouth”, Journal of Marketing Research, Vol. 50 No. 4, pp. 427-444.
Lu, X. and Feng, Y. (2009), “Value of word of mouth – an empirical study based on online restaurant reviews”, Management World, No. 7, pp. 126-132.
Luo, X. (2009), “Quantifying the long-term impact of negative word of mouth on cash flows and stock prices”, Marketing Science, Vol. 28 No. 1, pp. 148-165.
Mahajan, V., Muller, E. and Bass, F.M. (1991), “New product diffusion models in marketing: a review and directions for research”, Diffusion of Technologies and Social Behavior, Springer, New York, NY.
Mahajan, V., Muller, E. and Bass, F.M. (1995), “Diffusion of new products: empirical generalizations and managerial uses”, Marketing Science, Vol. 14 No. 3, pp. 79-88.
Manchanda, P., Dubé, J.P., Goh, K.Y. and Chintagunta, P.K. (2006), “The effect of banner advertising on internet purchasing”, Journal of Marketing Research, Vol. 43 No. 1, pp. 98-108.
Martin, J., Barron, G. and Norton, M.I. (2007), “Choosing to be uncertain: preferences for high variance experiences”, London Business School Trans-Atlantic Doctoral Conference.
Meyer, R.J. (1981), “A model of multiattribute judgments under attribute uncertainty and informational constraint”, Journal of Marketing Research, Vol. 18 No. 4, pp. 428-441.
Miniard, P.W., Bhatla, S. and Rose, R.L. (1990), “On the formation and relationship of ad and brand attitudes: an experimental and causal analysis”, Journal of Marketing Research, Vol. 27 No. 3, pp. 290-303.
Moe, W.W. and Trusov, M. (2011), “The value of social dynamics in online product ratings forums”, Journal of Marketing Research, Vol. 48 No. 3, pp. 444-456.
Naik, P.A., Mantrala, M.K. and Sawyer, A.G. (1998), “Planning media schedules in the presence of dynamic advertising quality”, Marketing Science, Vol. 17 No. 3, pp. 214-235.
Olney, T.J., Holbrook, M.B. and Batra, R. (1991), “Consumer responses to advertising: the effects of ad content, emotions, and attitude toward the ad on viewing time”, Journal of Consumer Research, Vol. 17 No. 4, pp. 440-453.
Onishi, H. and Manchanda, P. (2012), “Marketing activity, blogging and sales”, International Journal of Research in Marketing, Vol. 29 No. 3, pp. 221-234.
Pechmann, C. and Stewart, D.W. (1988), “Advertising repetition: a critical review of wearin and wearout”, Current Issues and Research in Advertising, Vol. 11 Nos 1/2, pp. 285-329.
Rogers, E.M. (2003), Diffusion of Innovations, Free Press, New York, NY.
Rui, H., Liu, Y. and Whinston, A. (2013), “Whose and what chatter matters? The effect of tweets on movie sales”, Decision Support Systems, Vol. 55 No. 4, pp. 863-870.
Sun, M. (2012), “How does the variance of product ratings matter?”, Management Science, Vol. 58 No. 4, pp. 696-707.
Trusov, M., Bucklin, R.E. and Pauwels, K. (2009), “Effects of word-of-mouth versus traditional marketing: findings from an internet social networking site”, Journal of Marketing, Vol. 73 No. 5, pp. 90-102.
Villanueva, J., Yoo, S. and Hanssens, D.M. (2008), “The impact of marketing-induced versus word-of-mouth customer acquisition on customer equity growth”, Journal of Marketing Research, Vol. 45 No. 1, pp. 48-59.
Zhang, X.X.M. (2006), Tapping into the Pulse of the Market: Essays on Marketing Implications of Information Flows, MA Institute of Technology, Cambridge, MA.
Zhu, F. and Zhang, X. (2010), “Impact of online consumer reviews on sales: the moderating role of product and consumer characteristics”, Journal of Marketing, Vol. 74 No. 2, pp. 133-148.
Li, X. and Hitt, L.M. (2008), “Self-selection and information role of online product reviews”, Information Systems Research, Vol. 19 No. 4, pp. 456-474.