Abstract
The rapid adoption of online streaming and over-the-top services has fundamentally changed at-home entertainment media consumption and given rise to new behaviors which are often characterized by a high intensity of watching (e.g., binge-watching). In this paper, we investigate how the watching intensity affects consumers’ engagement with media franchises in two areas: personal and interactive engagement. The former involves consumers’ adoption and consumption of franchise extensions and the latter concerns consumers’ content generation related to a focal media product they watched. Using individual-level data from an online anime (Japanese cartoons) platform, we find inverse U-shaped effects of watching intensity with the largest effects around three to five hours of watching per day on personal engagement and two to four hours a day on interactive engagement. The positive effects of watching intensity are larger for sequels than other types of franchise extensions. For interactive engagement, our results show that conditional on rating submission, higher watching intensity is associated with higher valence of anime ratings, the most prevalent form of UGC on the platform. We interpret this result as evidence that watching intensity can induce liking.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The global at-home media entertainment industry has fundamentally changed during the last 15 years due to online streaming, a new technology brought about by high-speed Internet. Prior to online streaming, at-home media consumption was mostly characterized by “appointment watching” or “linear watching” of TV programs, i.e., consumers turning to a specific TV channel to watch a show on a day and at a time determined by the TV network. While appointment watching still exists, its importance has declined. At-home media consumption through TV channels and satellite TV has been – to a large extent – replaced by over-the-top (OTT) media consumption.Footnote 1 OTT media services allow consumers to freely choose when and how much of a movie or TV series to watch. Consumers have overwhelmingly embraced this new technology giving rise to new media consumption behaviors such as “binge-watching,” which is characterized by rapid consumption of media content in a short period of time. This new technology has also changed the structure of the at-home media entertainment industry: streaming services such as Netflix and Hulu have emerged as major players in distributing and creating media content, traditional TV networks have launched their own streaming platforms, and new contenders such as Apple and Disney have launched their own OTT services.
Another notable trend in the entertainment and media industry is the rising success of media franchises on both big and small screens.Footnote 2 We define “media franchise” as a collection of media products in which several derivative works have been developed in response to the popularization of an original creative work and the commercial exploitation of such through licensing agreements (Aarseth, 2006). For example, all 10 top-grossing movies of 2022 in the U.S. belonged to franchises such as “Avengers,” “Jurassic World,” and “Top Gun.”Footnote 3 Franchise series also ruled the small screen as witnessed by the exploding traffic on Netflix drawn to “Breaking Bad” or “Bridgerton,” original series created by the streaming service.
Anecdotal evidence and common industry practices suggest that intensive media consumption might increase viewer engagement with media franchise series. Engagement in this context refers to the extent of the consumer’s interactions with media content, which can manifest as personal enjoyment (watching of related media content) or interactive enjoyment (talking or writing about media content). For example, “Breaking Bad” creator Vince Gilligan previously told Mashable that the show “may have met its demise after season two, had it not been for streaming video on demand. It ushered in new viewers and encouraged time-starved individuals to keep watching at their own pace resulting in enormous growth from season to season” that reached its climactic end in September 2013 with 10.3 million viewers (the show’s highest viewership ever) (Hernandez, 2014). Similarly, for popular series such as “Supernatural,” Netflix starts streaming previous season(s) shortly before the release of a new season (on traditional TV).Footnote 4 Netflix’s experience with “Squid Game” serve as another example. According to the platform, “It’s hard to imagine, for example, how a Korean title like Squid Game would have become a mega hit globally without the momentum that came from people being able to binge it. We believe the ability for our members to immerse themselves in a story from start to finish increases their enjoyment but also their likelihood to tell their friends, which then means more people watch, join and stay with Netflix.”Footnote 5
However, there is little systematic examination into the relationship between a consumer’s media consumption intensity and her engagement with a media franchise.Footnote 6 In this paper, we aim to fill this gap. To do so, we estimate the effects of different watching intensities ranging from low- to high-intensity watching. In general, low-intensity or slow watching (of a series) over an extended time period can encompass both appointment watching (on TV) and consumer-driven gradual viewing of a series (through an OTT service). High-intensity or “fast” watching of a series over a short time period may apply to both marathon releases (on TV) and consumer-driven intensive viewing of a series (through an OTT service). Higher levels of watching intensity therefore include but are not restricted to binge-watching.Footnote 7
To measure consumer engagement with media products, we adopt the categorization developed by Calder et al. (2009). They identify two types of consumer engagement: “personal engagement” such as enjoyment and relaxation directly derived from consuming the product and “interactive engagement” such as socialization and participation in a community facilitated by consuming the product. Calder et al. (2009) associate the former with an individual’s internal state of getting caught up in the flow of an activity and being absorbed by it Csikszentmihalyi (1997) and the latter with an individual’s voluntary content generation and promotion of a focal media product. Therefore, increased consumption intensity might result in the viewer watching subsequent seasons or other types of franchise extensions (see Table 1 for definitions and examples) of the same series (i.e., personal engagement) and/or in the viewer promoting the TV series and producing user-generated content (UGC) about it (i.e., interactive engagement).
If watching intensity increases consumers’ engagement with a media franchise, this finding would have important implications for media companies’ content production and release strategies. First, in line with this finding, both online streaming platforms and traditional TV networks could use watching intensity levels of a focal media product as an additional predictor for the success of its sequels and other types of franchise extensions, thus making the production planning more informed. Second, for online streaming services, the finding would validate their practice of releasing a whole season of a series at once and thereby making it bingeable. For TV networks, it would provide support for their new content release strategy of promoting a new season shown on traditional TV by making older seasons available through online streaming services. This strategy could garner an especially important benefit for TV networks since it would not only increase immediate profits through higher advertising revenues (for the new season on traditional TV), but also extend the “life” of a series, making it more likely to reach five seasons at which point the series is a candidate for syndication, a very profitable path for networks.Footnote 8
If watching intensity does not always increase media franchise engagement or if it does not do so for all shows, it is important to understand when and why this is the case. Does the level of watching intensity matter such that excessive watching would actually hurt consumer engagement with the media franchise? Does the timing of the release through online streaming services play a role? Or does the type of franchise extension make a difference? In this paper, through a systematic empirical investigation, we provide a nuanced description of consumers’ different levels of watching intensities and how they affect consumer engagement with media franchises.
Our data come from MyAnimeList.net, an online forum that attracts anime (Japanese cartoons) fans from all over the world. We observe an individual’s adoption timing of anime including the number of days it took a consumer to watch the whole season of an anime.Footnote 9 This information allows us to calculate the watching intensity for each consumer-anime combination. Further, we observe an individual’s self-generated content about an anime in the form of published posts on the discussion forum as well as submitted ratings and recommendations. Our data also contain information on a consumer’s decision to watch the next season (sequel) of an adopted anime and/or to watch other types of franchise extensions (see Table 1 for definitions and examples of franchise extensions). And lastly, we observe a consumer’s demographic characteristics, including the individual’s geographic location, age, and gender.
Our goal is to understand how the watching intensity (conditional on viewing the whole season) impacts a consumer’s actions related to media franchise engagement. Potential endogeneity of a user’s decision on how intensively to watch an anime is a concern for the causal interpretation of the estimated effects. Omitted variables, i.e., variables that are not observed in the data but are correlated with an individual’s watching intensity and influence their engagement, are the cause of the concern (see, e.g., Angrist and Pischke, 2009). In this paper, we address endogeneity concerns by exploiting our unique and rich individual-level data that allow us to incorporate a large number of fixed effects to control for various types of omitted variables (see, e.g., Wooldridge, 2001). The idea of the fixed effects approach is to partition the variation in watching intensity into that which is “clean” (Allenby & Rossi, 2019), i.e., as good as random (conditional on fixed effects and observed variables), and that which is not and only use the clean portion of variation to estimate the effect of watching intensity (see Angrist and Pischke, 2009). Specifically, our model incorporates the following fixed effects: anime, month, user-genre, and user-weekend/holiday fixed effects. We discuss in detail in Section 4.1 to what extent these fixed effects together with control variables can address potential endogeneity concerns and in Section 5.1 how the patterns in our empirical results alleviate a remaining endogeneity concern.
Our results show that a watching intensity of more than one hour per day, i.e., at a higher level than typically in linear watching, increases an individual user’s probability of watching a franchise extension (at any point in time in the future), and that the positive effects of watching intensity are larger for sequels than other types of franchise extensions. More interestingly, we find the effects of watching intensity to exhibit an inverse U-shaped pattern with the largest effects around three to five hours of watching per day. Additionally, the effects of watching intensity on an individual’s likelihood of finishing to watch a franchise extension critically depend on the availability of a franchise extension at the time of watching the focal season: the probability that a user finishes to watch a franchise extension only increases if the franchise extension was available at the time when the user watched the focal anime and, in that case, also follows an inverse U-shaped pattern. And lastly, watching intensity of more than three hours a day increases the probability that a user watches a franchise extension immediately next after watching the focal media product, and this effect is stronger when the franchise extension is a sequel.
Regarding the relationship between watching intensity and interactive engagement, i.e., the production of UGC, we find that intermediate levels of watching intensity around two to four hours per day increase the number of recommendations and ratings a user produces compared to the baseline of watching less than an hour per day. At the same time, our results also reveal that excessive watching intensity (i.e., more than five hours a day) can have a negative effect on UGC submission in the case of forum posts. We also find that, conditional on submitting a rating, consumers who intensively watch the focal anime rate it higher, suggesting that watching intensity positively affects consumers’ liking of a media product.
Our paper makes the following two contributions. First, we contribute to the consumer engagement literature by systematically examining the factors that drive consumer engagement in the context of a media franchise. By quantifying the effect of watching intensity on consumer engagement with a media franchise in two broad areas – interactive and personal engagement – our paper provides empirical evidence that the modus of consumption, on top of product adoption, influences consumer brand engagement. And second, our paper adds to the small but rapidly growing literature on watching intensity, binge-watching, and online streaming. To the best of our knowledge, we are the first to study the relationship between watching intensity and consumers’ subsequent media consumption and word-of-mouth behavior. Our results have important managerial implications for both online streaming services and traditional TV networks regarding content provision and the timing thereof.
The remainder of the paper is organized as follows: In the next section, we review the relevant literature. In Sections 3 and 4, we describe our data, introduce our model and estimation approach. We present our results in Section 5 and discuss managerial implications in Section 6. Finally, we conclude by summarizing our findings in Section 7.
2 Relevant literature
In this section, we draw from relevant streams of literature on customer engagement with a media franchise, watching intensity, and online movie streaming.
2.1 Customer engagement with a media franchise
Customer engagement has been extensively studied in the marketing literature (e.g., Bowden, 2009; Mollen and Wilson, 2010; Van Doorn et al. 2010; Vivek et al. 2012).Footnote 10 It differs from similar relational concepts such as participation or involvement in that it highlights customers’ interactive and co-creative experiences in networked relationships with multiple stakeholders including service personnel, firms, and/or other customers (Brodie et al., 2011). Empirical studies across various industries have shown that engaged customers play a key role in viral marketing activities by generating referrals and recommendations for products and services, in new product development, and in co-creating experiences and value (e.g., Nambisan and Nambisan, 2008; Brakus et al. 2009; Hoyer et al. 2010). However, to the best of our knowledge, no empirical study to date has systematically examined customer engagement in the context of media franchises.
To understand what drives customer engagement with media franchises, the first question is how customer engagement with a media product should be measured. In this regard, Calder et al. (2009) define media engagement in terms of the different motivational experiences that consumers have with a media product. Using confirmatory factor analysis, they identify two types of media engagement: personal engagement and interactive engagement. Personal engagement is associated with intrinsic motivation and includes individualistic experiences such as enjoyment and relaxation directly derived from consuming a media product. More specifically, a consumer’s personal engagement with a media product is driven by the “transportation” motive, i.e., by consuming a media product a consumer aims to be transported into a different state (e.g., from bored to happy) or to be transported into taking part in an activity, such as being absorbed into a story and shutting out the real world. Csikszentmihalyi (1997) describes a more general variant of the “transportation” experience as the internal state of an individual getting caught up in the “flow” of an activity and being absorbed by it. Interactive engagement, on the other hand, is associated with extrinsic motivation and includes interactive experiences such as socialization and participation in a community facilitated by the consumption of a media product. For example, after watching a movie, a consumer may have the urge to generate online word-of-mouth related to the movie by submitting a rating, publishing a review, or participating in discussion forums on social media. This online word-of-mouth has been shown to be effective in raising awareness and influencing opinions of other consumers, through which it increases the adoption of the movie (Ameri et al., 2019).
In this study, we follow the categorization by Calder et al. (2009) when examining consumers’ personal and interactive engagement with media franchises. In our empirical context of an online anime platform, we measure a user’s personal engagement with a media franchise by examining her self-enjoyment of the focal media product and the adoption of its franchised extensions including sequels, spin-offs, summaries, side stories, and remakes. We assess her interactive engagement through her content generation and promotion of a focal media product, i.e., her decision to submit recommendations, ratings, and comments in a community discussion forum regarding the focal anime.
2.2 Watching intensity
We adopt a time-based definition of watching intensity: average number of hours of watching the focal media product per day (conditional on the completion of watching the focal media product). While this definition results in a continuous measure of watching intensity, in our empirical models (see Section 4.2), we estimate separate coefficients for different watching intensity levels. Low-intensity watching, by our definition, encompasses both appointment watching (on TV) and consumer-driven slow viewing of a series (through an OTT service); while high-intensity watching, by our definition, includes watching of both marathon releases (on TV) and consumer-driven intensive viewing of a series (through an OTT service).
We would like to point out that by our definition, watching intensively includes, but is not restricted to binge-watching, i.e., the practice of watching multiple episodes (of a series) in one sitting.Footnote 11 There is disagreement in previous literature on how much watching is considered binge-watching. Based on a survey of their users, Netflix defines binge-watching as watching at least two episodes in one sitting (Netflix, 2013). The number of two episodes is not agreed upon by everyone though. For example, Amazon made the first 3 episodes of its series “Alpha House” available to its viewers at once, implying that it considers 3 episodes as a bingeable amount. Schweidel and Moe (2016) and Lu et al. (2017) use process-level streaming data to define binge-watching. For example, Schweidel and Moe (2016) consider “the consumption of multiple episodes of a television series in a short period of time” as binge-watching. Other studies rely on respondents’ perception of what is considered binge-watching without defining a specific amount (e.g., Devasagayam, 2014; Pena, 2015).
These definitions rely on the number of episodes without differentiating between one-hour dramas (about 40 minutes without commercials) and 30-minute sitcoms (about 20 minutes without commercials). It is debatable whether watching 8 episodes of a sitcom corresponding to about 2.5 hours should be considered as binge-watching. To circumvent these obstacles, in this study, we measure watching intensity based on the time spent watching a whole season. This definition is in line with the idea that watching a show very intensively is a violation of what is considered the norm, regular TV watching such as “appointment watching” or “linear watching.”
The underlying mechanism that drives high-intensity watching is related to the concept of flow (e.g., Hoffman and Novak, 1996), which describes a state of focused concentration, intrinsic enjoyment, and time distortion. Previous research has found that users who experience the flow are more likely to repeat their behaviors or even become addicted (e.g., Kubey and Csikszentmihalyi, 2002; Chou et al. 2003). This mechanism also provides a plausible explanation for the interplay between advertisements and binge-watching as documented in Schweidel and Moe (2016): advertisements in a viewing session discourage binge-watching and binge-watchers are less responsive to advertisements compared to non-binge-watchers.
While there has been a considerable amount of research on the reasons for high-intensity watching, few studies have focused on the consequences of high-intensity watching. In the TiVo (2015) study, 52% of respondents indicated that they feel sad when they finish bingeing a series; 31% reported that they have lost sleep due to bingeing. Binge-watching - due to the intensity of the experience and the flow it creates - has been suggested to create loyalty to a series, to lead to fandom, and to help the formation of one-sided, unconscious bonds between viewers and characters or, at the very least, behavior similar to fandom such as purchasing ancillary materials, creating fandom pages or posting or creating content (Devasagayam, 2014; Jenner, 2017). However, empirical evidence supporting these claims is very limited. To the best of our knowledge, this paper is the first to carry out a systematic empirical examination of the effects of watching intensity on consumer engagement with a media franchise.
2.3 Online streaming
Despite its wide popularity, research on online movie streaming is scarce. Studying consumer behavior within online streaming services, Zhang et al. (2013) develop a new class of “clumpiness” measures and, using data from Hulu.com, show that the “clumpiness phenomenon” is widely prevalent in digital content consumption. In a separate study, Zhang et al. (2015) extend the traditional recency/frequency/monetary value (RFM) segmentation framework to include the clumpiness measure (RFMC). In particular, they show that the RFMC framework can help companies with bingeable content (such as online streaming platforms) uncover previously unseen customer segments. Ameri et al. (2019) investigate the drivers of consumers’ anime adoption decisions in the context of online streaming. They find the average anime rating and the popularity rank from the community network, i.e., the platform, to have larger effects on consumers’ adoption decisions than the same two types of information obtained from the personal network, i.e., a consumer’s friends. And lastly and most closely related to this paper, Schweidel and Moe (2016) simultaneously examine the drivers of users’ binge-watching behavior and their responses to advertisements using data from Hulu.com. They find that binge-watchers are less responsive to advertising compared to non-binge-watchers.
3 Data
Our data come from MyAnimeList.net. This website is a consumption-related online community where online interactions are based upon shared enthusiasm for a specific consumption activity (Kozinets, 1999). It was created to allow anime fans to gather and share their excitement and opinions about animes. This website was established in November 2004, but its main activities did not begin until 2007 when the website moved to a public domain and its user base started to grow rapidly (see Fig. 1). At the point in time when we started the data collection in March 2015, there were more than 2.5 million users on the website.
Dates users joined MyAnimeList.Net
On MyAnimeList.net, both animes and users have their own pages. On a user’s page, information about the animes the individual has adopted (including the dates) and her opinion about adopted animes (via numerical ratings, forum posts, and recommendations) is shown in addition to personal information such as the individual’s geographic location, gender, age and the date when she joined the website. Users can create a list of animes that they have watched or plan to watch (we refer to this list as “watch list” throughout this paper).Footnote 12 Note that users add animes to their watch lists using a search function so that all animes are correctly and uniquely identified. Further, users can also indicate their opinion about the animes on their watch list by rating them on a scale ranging from 1 to 10 (10 being the highest rating). Throughout this paper, we refer to ratings given to animes on watch lists as “ratings.” Lastly, users can indicate the date they started watching an anime season and the date they finished watching an anime season. We use the start and end dates to infer the beginning and end of a user’s watch period for an anime season.
3.1 Estimation samples
We scraped data on 370,000 individuals from the website. Not all users list start dates for (all or any) animes they have adopted on their watch list. After excluding all user-anime combinations for which we did not have start dates, we were left with 92,273 individuals.Footnote 13 We then dropped (i) animes for which we did not have the release date or information on the number of episodes; (ii) user-anime combinations for which the watch period seemed unreasonably long, i.e. more than 3,000 days; (iii) observations for days on which individuals indicated to have watched animes for more than 24 hours or a single anime for more than 16 hours; (iv) observations with start dates before 2008 since, although the website was launched in 2004, its main activities did not start until mid 2007 (see Fig. 1); (v) individuals with unknown geographic location; (vi) observations with start dates after the end of 2014. Using the remaining 47,557 individuals and 4,364 animes (1,872,404 user-anime combinations), we took the following steps to get to our final data.
First, we dropped animes for which it would take an individual less than 2 hours to watch the whole season. Table 2 shows the frequency distributions for the 4,364 animes with respect to their number of episodes and durations of a season in hours.Footnote 14 Movies, i.e., animes with one episode, or short anime series generally take less than 2 hours to be watched and thus cannot be watched at higher intensity levels. Note that, even if an individual watches 3 movies back to back, if they are not part of the same franchise, we do not consider this instance as intense watching. Therefore, looking at the right half of Table 2, we dropped 1,626 animes in the first step because their season duration was less than 2 hours.
Second, we dropped user-anime combinations in which an individual did not watch the whole season. Even if a user intensively watches the first half of a season (and does not watch the second half of the season), her behavior might be different from someone who intensively watched and finished the whole season. To be able to attribute the difference in user behavior to the watching intensity and not to the completion of the whole season, we only consider cases in which the individual finished watching the whole season.
And third, we only consider user-anime combinations in which users have the option to intensively watch the anime, but may not do so, i.e., we only consider watching incidences after the season finale of an anime has been aired. It is noteworthy that most of our observations are for such cases. In Fig. 2, we show the number of days (after the original airing of the last episode in a season) after which individuals, who watched the anime at different intensity levels, started to watch it. Note that access to the anime after its original airing is not a reason for the delayed watching: almost all animes are available through online streaming within 3 days of the original episode airing (see also Ameri et al. 2019). After these steps, our final data sample for the empirical analysis related to interactive engagement contains 39,630 individuals and 2,467 animes with 739,969 user-anime combinations.Footnote 15
Number of days after season finale that animes were watched (truncated at 500 days)
For personal engagement, we need to constrain the final data sample further. More specifically, we can only consider animes that have a franchise extension, i.e., next season (sequel) or another type of franchise extension (i.e., side story, spin-off, summary or remake). After dropping animes that do not have any franchise extension, the data sample contains 37,516 individuals and 1,246 animes with 536,505 user-anime combinations, i.e., unique user-(focal)-anime combinations. Sometimes, animes have multiple types of franchise extensions (e.g., a spin-off and a summary). In such cases, we model the adoption of each type of franchise extension as a separate potential adoption. Sometimes, animes have multiple franchise extensions of the same type (e.g., two spin-offs). In such cases, we only model the first potential adoption among franchise extensions of the same type. Because of these two reasons, the number of observations in the model estimation is 641,424.
3.2 Definitions of watching intensity and engagement variables
We compute the watching intensity variable for each user-anime combination as follows: total duration of the anime season (measured in hours) divided by the number of days that it took the individual to watch all episodes of the anime season.Footnote 16 Thus, watching intensity is a continuous variable. In our main empirical analysis, we will use six dummy variables that capture the essence of the continuous watching intensity variable: whether a user spent less than 1 hour, between 1 to 2 hours, 2 to 3 hours, 3 to 4 hours, 4 to 5 hours, or more than 5 hours a day watching the anime season. In Section 5.3, we show that this simpler operationalization of watching intensity produces very similar results as using the continuous version of this variable.
We investigate three aspects of an individual’s personal engagement with media franchises by examining her consumption decisions related to franchise extensions of a focal anime season. First, watching an anime season intensively might affect a user’s likelihood of watching its franchise extension (at any point in time in the future). Second, conditional on watching a franchise extension, the watching intensity might affect a consumer’s likelihood of finishing to watch the franchise extension. And lastly, if a franchise extension is available at the time of watching the focal anime, watching intensively might also affect the likelihood of watching a franchise extension immediately next versus an unrelated anime. We operationalize these three aspects of personal engagement as binary indicator variables: (i) whether a user watched a franchise extension (at any point in time in the future), (ii) whether a user finished watching the franchise extension (conditional on starting to watch a franchise extension), and (iii) whether a user watched a franchise extension next (conditional on a franchise extension being available).
We examine an individual’s interactive engagement with media franchises by looking at her decisions to produce UGC related to adopted anime seasons. We investigate three types of UGC: recommendations, ratings, and posts on the discussion forum. Recommendations on this platform exhibit the following pattern: “If you like anime A, you will like anime B because of XYZ.” In that sense, individuals give a recommendation for which two animes are similar, but not necessarily a recommendation that an anime is particularly good. Posts on the discussion forum typically discuss topics such as new season release dates, voice cast decisions, story lines, specific characters, awards, or anime adaptations. Ratings are different from the other two forms of UGC in that they are numerical and a higher rating clearly indicates a more favorable opinion towards the rated anime season. Furthermore, while ratings are publicly visible to everybody, they are recorded by a user on her watch list and help her remember her preference for or liking of a particular anime.
For forum posts and recommendations, we study whether the watching intensity affects the number of posts and recommendations, respectively. For forum posts, we also investigate whether watching intensity affects the average sentiment and average length of submitted forum posts. For ratings, we examine whether the watching intensity impacts whether an individual submits a rating and its valence.
3.3 Self-reported watching data
Our anime watching data are self-reported. Thus, accuracy in the reporting of anime adoptions could be a potential concern due to factors such as social desirability noted by previous literature on response bias in survey research (e.g., Furnham, 1986; Fisher, 1993; Steenkamp et al. 2010). However, in contrast to surveys, that usually provide incentives for participation and have participants answer previously-determined questions, users’ self-reporting on anime-related behaviors on MyAnimeList.net is entirely voluntary. Therefore, there are no obvious incentives for them to falsely report their true anime watching behavior. Additionally, empirical evidence provided by Lovett and Staelin (2016) through a comparison of survey panelists’ self-reported viewing data and the actual streaming data in the similar setting of TV shows supports that people tend to correctly report their actual watching behavior.
To further check the reliability of our self-reported anime watching data, we display the total number of hours per day users in our data reported having watched animes (conditional on watching) in Fig. 3. Note that the number of hours in this figure includes everything the user watched, i.e., all animes watched that day (whether they belong to any franchise or not). The distribution is skewed with more than 40% of observations being for less than an hour of watch time. While there is heterogeneity in the number of hours per day, the data only contain a relatively small number of observations with watch times of more than six hours and essentially none with watch times of more than 11 hours.
And lastly, we accessed data from the 2014 American Time Use Survey collected by the Department of Labor.Footnote 17 Using these data, we created a figure showing the distribution of the number of hours individuals report watching TV per day. The distribution based on the 2014 American Time Use Survey (shown as Fig. 10 in Web Appendix A) exhibits a similar shape to one shown in Fig. 3. Thus, we are confident that the self-reported adoption data are reliable in our context.
Number of hours watching (any) animes per day
3.4 Data description
We present summary statistics for the 39, 630 individuals in our final sample in Table 3. 21,258 individuals in our final sample report their age. Among these individuals, the average age is 23 years. \(42\%\) of users are female and \(41\%\) of individuals are male with the remaining \(17\%\) of individuals not specifying their gender. \(45\%\) live in Europe, \(32\%\) come from North America, \(8\%\) from South America, \(11\%\) from Asia, and \(4\%\) from other geographies. Users, on average, have watched 15 anime during the last 30 days and a total of 207 animes over the course of their membership on the platform. Further, users, on average, started watching \(31\%\) and \(3\%\) of the animes on their watch lists on weekends and holidays, respectively.
Figure 4 shows the distribution of watching periods, i.e., the number of days between watching the first and last episode of an anime season, in our estimation sample. In 49% of the user-anime combinations, the individual watched a complete anime season within 5 days, with 27% of user-anime combinations being watched within a day or two. While Fig. 4 does not account for the length of a season in terms of the number of episodes or the length of an episode (in minutes), it nevertheless shows the possibility that a significant portion of user-anime combinations might be watched intensively.
Watching periods (truncated at 100 days)
Figure 5 shows the distribution of the watching activity that was done at different intensity levels. In about 45% of user-anime combinations, users watch the anime for less than an hour each day, while users viewed the anime at the highest intensity, spending 5 or more hours per day watching it, in 7% of user-anime combinations.
The boxplots in Fig. 6 show the fractions of adopted animes that were watched at each intensity level. The median user watches about 45% of adopted animes for less than an hour a day, about 18% of adopted animes for 1 to 2 hours a day, and about 10% of adopted animes for 2 to 3 hours a day. However, as the boxplots also show, there is significant variation in the fractions across users as shown by the large range of the 5th and 95th percentiles of the distributions. This empirical observation is consistent with previous findings (e.g., MarketCast, 2013; Schweidel and Moe, 2016).
Watching intensity
Fraction of users’ adopted animes watched at each intensity level
Next, we discuss our engagement variables. Table 4 shows statistics for our personal engagement variables. We observe several patterns in the table. First, personal engagement crucially depends on (i) the type of franchise extension (sequel vs. another type) and (ii) the availability of a franchise extension at the time of watching the focal media product. We observe the highest engagement when a sequel is available at the time of watching the focal anime for all three engagement variables. Second, watching intensity generally increases a user’s likelihood of starting to watch a franchise extension, of finishing to watch it, and of watching it immediately next. However, for the first two engagement variables, the positive effect seems to diminish at the highest watching intensity levels. This observation highlights the potential non-linearity in the effect of watching intensity on personal engagement.
In Table 5, we display statistics related to users’ interactive engagement. Note that, out of the three types of UGC on this platform (i.e., forum posts, ratings, and recommendations), ratings are the dominant form of UGC in terms of user participation: the submission rate is \(92.76\%\) for ratings compared to \(1.2\%\) for forum posts and \(0.1\%\) for recommendations. In terms of forum posts, users who watch an anime for 3 to 4 hours per day are most likely to post on the discussion forum. Conditional on posting on the forum, users who watch an anime season intensively make longer and more negative posts. In terms of anime ratings, users who watch an anime season for more than 4 hours per day are less likely to submit a rating compared to those who watch it at a very slow rate (e.g., less than 1 hour per day), but, conditional on submitting a rating, rate it higher. We do not observe significant differences between users who watch intensively and users who watch slowly in their recommendation behaviors.
4 Models and estimation
We investigate the relationship between the watching intensity and a consumer’s personal and interactive media franchise engagement. Potential endogeneity of the watching intensity is a concern. We first discuss how we address this concern and then present our empirical specification.
4.1 Potential endogeneity of watching intensity
Potential endogeneity of a user’s decision on how intensively to watch an anime is a concern when estimating the effects of watching intensity on media franchise engagement. Omitted variables, i.e., variables that are not observed in the data but are correlated with an individual’s watching intensity and influence their engagement, are the cause of the concern (see, e.g., Angrist and Pischke, 2009). In this paper, we address endogeneity concerns by exploiting our unique and rich individual-level data that allow us to incorporate a large number of fixed effects to control for various types of omitted variables (see, e.g., Wooldridge 2001). In the ideal situation, conditional on the fixed effects (and control variables), the watching intensity decision is independent of the error term, the coefficient estimates for the watching intensity variable can be interpreted as causal.
The idea of the fixed effects approach is to partition the variation in watching intensity into that which is “clean” (Allenby & Rossi, 2019), i.e., as good as random (conditional on fixed effects and observed variables), and that which is not and only use the clean portion of variation to estimate the effect of watching intensity (see Angrist and Pischke, 2009). Our specification includes the following fixed effects: anime, month, user-genre, and user-weekend/holiday. In the following, we discuss potential sources of endogeneity concerns and how our fixed effects address them.
First, the anime fixed effects control for time- and user-invariant and anime-specific unobservables that might affect users’ watching intensity. For example, some animes might have teasers at the end of each episode that induce individual viewers to watch more intensively.Footnote 18 Second, the month fixed effects control for anime- and user-invariant and month-specific unobservables that might influence individual users’ watching intensity. For example, during summer months, individual users might have more leisure time that they can spend watching animes due to summer breaks and vacations. Third, the user-genre fixed effects control for each user’s intrinsic preference for a specific genres of anime.Footnote 19 For example, a user might like romance and action animes the best, and tend to watch these two genres of animes more intensively. And lastly, the user-weekend/holiday fixed effects account for individual users’ preference or lifestyle which may lead to the user watching animes more during weekdays or weekends/holidays.Footnote 20
Another source of potential endogeneity concerns is a user’s unobserved interest in a specific anime. Although the user-genre fixed effects control for each user’s intrinsic preferences for specific genres of animes, an individual user’s interest in animes belonging to the same genre may vary across animes. For example, a user may be more interested in “Fullmetal Alchemist” than in “Bleach,” although both series are classic examples of action animes. This interest in a specific anime might affect both their watching intensity of the focal anime and personal/ interactive engagement. Ideally, we would like to include user-anime fixed effects to control for individual users’ unobserved interest in specific animes.Footnote 21 However, due to data limitations in the sense that we only have one observation for each user-anime combination, we are unable to do so. We acknowledge our inability to control for each user’s unobserved interest in a specific anime as a limitation of our study, and will discuss in Section 5.1 how this limitation affects the identification and interpretation of our empirical findings.
4.2 Empirical specification
We start by describing how we model a consumer’s personal engagement with a media franchise. The three personal engagement variables under study are operationalized as binary indicator variables and we estimate a linear probability model for each of the three dependent variables. Let \(i=1,...,N\) denote consumers and \(j=1,..., J\) denote animes. Consumer i’s utility from personally engaging with the media franchise is given by
The variable \(y_{ij}\) (whose realizations we observe in the data) equals 1 if user i, depending on the personal engagement variable of interest, watched at least one episode of the franchise extension, watched all the episodes of the franchise extension, or watched the franchise extension next without watching any other anime in between. In each case, \(y_{ij}\) equals 0 otherwise. \(B^k_{ij}\), \(k \in \{1-2, 2-3, 3-4, 4-5, 5+\}\), are dummy variables for watching intensity, indicating whether user i watched anime j for k number of hours on average per day. The baseline is set at less than one hour of watching anime j per day. By using this non-parametric specification to model the effects of watching intensity, we explore the potential non-linearity in the relationship between an individual user’s watching intensity and their franchise engagement behaviors. Our control variables \(C_{ij}\) include the popularity rank and community rating of anime j both measured at time t when user i watched the focal anime seasonFootnote 22 and, if the franchise extension was not available at the time of user i watching the focal anime season, the wait time until the franchise became available in days. And lastly, \(\gamma _j\) contains anime fixed effects, \(\lambda _t\) contains calendar month fixed effects, \(\tau _{ig}\) contains user-genre fixed effects, \(\nu _{it}\) contains user-weekend/holiday fixed effects, and \({\epsilon }_{ij}\) is an error term following a standard normal distribution.
Furthermore, because both the type and the availability of a franchise extension play important roles in precisely pinning down the effects of watching intensity (see also Section 3.4), we estimate the effects of watching intensity conditional on the type and the availability of a franchise extension. That is, we partition our data into four scenarios: when a franchise extension is a sequel and was available at time of user i watching the focal anime season, when a franchise extension is not a sequel and was available, when a franchise extension is a sequel and was not available, and when a franchise is not a sequel and was not available, and estimate the model separately under these four scenarios.Footnote 23
And lastly, we describe how we model a consumer’s interactive engagement with a media franchise. We operationalize the UGC variables as described in Section 3.2 and estimate linear regressions. For all UGC variables, consumer i’s engagement is modeled as follows:
\(B^k_{ij}\), \(k \in \{1-2, 2-3, 3-4, 4-5, 5+\}\), are the same dummy variables for watching intensity as we describe in equation (1) indicating whether user i watched anime j intensively for a number of k hours per day. \(C_{ij}\) contains control variables including the popularity rank and the average community rating. For dependent variables related to forum posts only, we also control for whether consumer i has ever published a forum post, the number of previous forum posts about, and the time since the last forum post on anime j published by anyone. For dependent variable related to recommendations only, we also control for the number of existing recommendations for anime j at the time of individual i’s adoption of the focal anime season and for dependent variables related to ratings only, we control for the number of ratings submitted by user i at the time of individual i’s adoption of anime j. Lastly, \(\epsilon _{ij}\) is the error term following a standard normal distribution.
5 Results
5.1 Personal engagement
We present the parameter estimates from our personal engagement models in Table 6. Columns (i) to (iv) describe an individual’s decision of whether to start watching a franchise extension (at any point in time) under varying scenarios of franchise extension type and availability. Recall that we estimate five coefficients allowing for non-linear effects of different watching intensities. Compared to the baseline of watching the focal anime for less than an hour a day, watching it for more than one hour per day increases an individual user’s likelihood of starting to watch a franchise extension of the focal anime. As an illustration, in Fig. 7, we plot the point estimates of the effects of different watching intensity levels on personal engagement variables for scenarios of franchise extension type and availability for which coefficients are statistically significant. As shown in Fig. 7(a), (b), (c) and (d) the effect sizes generally follow an inverse U-shaped pattern with the largest effect size when the individual watches between three to five hours per day.
Watching an anime series for three to five hours a day fits into an evening after work or school. The inverse U-shaped pattern is consistent with consumers initially experiencing a stronger flow when watching the focal anime more intensively, while satiation and fatigue kick in after watching the same anime for several hours. Comparing across the four extension types and availability scenarios (sequel/available, sequel/not available, other type/available, other type/not available), the effect of watching intensity on the adoption of a franchise extension is the largest when the extension is a sequel and was available at time when the user watched the focal anime (column (i) compared to columns (ii) - (iv) in Table 6). This finding is consistent with our expectation that consumers tend to experience a stronger and more seamless flow effect when watching a sequel compared to watching other types of franchise extensions. Among the different kinds of franchise extensions, sequels, i.e., next seasons, are the ones that continue the same story line of and share the same main characters with the prequel or previous season. Other franchise extension may have a different story line or center around different characters (e.g., “Better Call Saul,” a spin-off of “Breaking Bad,” tells the story of a lawyer who was a secondary character in “Breaking Bad”). As a result, it is natural for users to continue the flow after watching a season intensively by starting to watch the next season.
Point estimates for different levels of watching intensity
Next, we examine whether the watching intensity affects a user’s probability of finishing to watch the franchise extension (conditional on starting to watch a franchise extension) in columns (v) to (viii) of Table 6. Interestingly, the watching intensity of a focal anime increases a user’s probability of finishing to watch a franchise extension only if the extension was available for watching when the user watched the focal anime (see columns (v) and (vii) in Table 6 and Fig. 7(e) and (f)). In this case, we observe a similar inverse U-shaped pattern of the effect of watching intensity: the effect is the largest when the individual watches the focal anime for three to four hours a day. If the franchise extension was not available at the time when the user watched the focal anime, the watching intensity does not have a significant effect on the user’s probability of finishing to watch the extension. One explanation that can account for these results is that the positive flow effect due to intensive watching is transient. In other words, it disappears if the user cannot adopt the franchise extension right away due to its unavailability.
Lastly, we examine how the watching intensity of the focal anime affects the more immediate media watching behavior. In columns (ix) and (x) of Table 6, we pin down how watching intensively affects a user’s probability of watching a franchise extension (compared to an unrelated media product) immediately next (conditional on a franchise extension being available at the time of watching the focal media product). We find that individuals who watch a focal anime season for more than three hours a day are more likely to watch a franchise extension immediately next than individuals who watch at a slower pace, and that this effect is stronger when the franchise extension is a sequel (see also Fig. 7(g) and (h)). This result again speaks to the consumer’s tendency to continue the flow created by watching the focal media product intensively. A natural way to do so is to watch its franchise extension if one is available. Relative to other types of franchise extensions, sequels benefit more from the effect of watching intensity because of the stronger flow they create.
As discussed in Section 4.1, a potential endogeneity concern that we are not able to alleviate via fixed effects due to a data limitation is a user’s intrinsic interest in a specific anime. This unobserved interest may increase a user’s watching intensity as well as their likelihood to adopt the franchise extensions of the focal anime. However, two patterns in our empirical results largely ease concerns related to this potential source of endogeneity:Footnote 24 first, if a user’s anime-specific interest were to play an important role, we would not expect the availability of the franchise extension at the time of watching the focal anime to have such a key role as revealed in our findings. And second, a user’s interest in a specific anime, when unobserved, would suggest for watching intensity to have a positive linear effect on engagement, which, by itself, is inconsistent with the inverse U-shaped pattern in our empirical results. Taken together, although we cannot entirely eliminate this endogeneity concern, we believe that our results capture the effect of watching intensity on personal engagement and cannot be explained by a user’s unobserved anime-specific interest.
To summarize, watching intensity of more than one hour per day increases a user’s probability of watching a franchise extension (at any point in time in the future). Interestingly, the positive effects of watching intensity exhibit an inverse U-shaped pattern with the largest effects around three to five hours of watching per day. Further, conditional on starting to watch a franchise extension, watching intensity increases the probability that a user finishes to watch the extension only if it was available at the time when the user watched the focal anime. In this case, we find similar inverse U-shaped effects of watching intensity with the largest effect around three to four hours of watching per day. And lastly, watching intensity of more than three hours a day increases the probability that a user watches a franchise extension immediately next after watching the focal media product.
5.2 Interactive engagement
We now discuss the results for interactive engagement shown in Table 7. For illustrative purposes, in Fig. 8, we plot the point estimates of the effects of different watching intensity levels on interactive engagement variables that are significantly impacted by watching intensity. Columns (i) to (iii) in Table 7 display the coefficients for estimations related to forum posts. Our results in columns (i) indicate that a similar inverse U-shaped pattern: watching one to two hours a day significantly increases the number of forum posts a user will write about the focal anime while if a user watches the focal anime more than five hours a day, she will make significantly fewer forum posts about the anime. At this high level of watching intensity, satiation and potentially fatigue are likely reasons for the smaller number of forum posts. Similarly, the results related to the length of the forum posts in columns (iii) indicate that watching three to four hours a day significantly increases the length of the forum posts a user submits compared to watching more or less intensively.
Column (iv) in Table 7 shows the results for recommendations – another form of UGC. Recall that recommendations on this platform exhibit the following pattern: “If you like anime A, you will like anime B because of XYZ.” In that sense, individuals give a recommendation for which two animes are similar, but not necessarily an endorsement that either of these two animes is particularly good. As shown in Fig. 8(c), the familiar inverse U-shaped pattern emerges: watching two to four hours a day significantly increases the number of recommendations a user submits compared to watching more or less intensively. Albeit we acknowledge that the effect size is small.
Point estimates for different levels of watching intensity
Ratings are different from the two previously mentioned forms of UGC in that they are numerical and in that a higher rating clearly indicates a more favorable opinion towards the rated anime. Furthermore, while ratings are publicly observable by everybody, just like the other two forms of UGC, they are recorded by a user on her watch list and help her remember her preference for or liking of a particular anime. This partly explains why ratings are the dominant form of UGC in terms of user participation on this platform. Our results in columns (v) and (vi) in Table 7 (and illustrated in Fig. 8(d) and (e)) again show an inverse U-shaped pattern in the effects of watching intensity: if an individual watches a focal anime for two to four hours a day, she is significantly more likely to rate the focal media product. Furthermore, conditional on submitting a rating, the effect of watching intensity on the valence of the rating is the largest for users who watch two to four hours day. We believe that this result suggests that watching intensity increases consumers’ liking of an anime (compared to the baseline of watching less than an hour a day). This finding is consistent with previous research which suggests that bingeing induces loyalty and fandom-like behavior (Devasagayam, 2014; Jenner, 2017). However, at the same time, our results also show that too intensive watching of more than four hours a day decreases this effect, likely due to satiation and potentially fatigue, compared to more intermediate levels of watching intensity.
In summary, intermediate levels of watching intensity (in our case, around two to four hours per day) increase the quantity of UGC a user produces compared to the baseline case of watching less than one hour per day. However, excessively high levels of watching intensity (more than five hours per day) result in less UGC production in the case of forum posts. Our results also reveal that consumers who intensively watch the focal anime rate it higher conditional on submitting a rating, suggesting that watching intensity positively affects consumers’ liking of a media product.
5.3 Robustness checks
In this section, we demonstrate the robustness of our results with respect to different operationalizations of the watching intensity variable and a different model specification. In the first robustness check, we utilize the continuous version of the watching intensity variable in the model and estimate separate coefficients for each of the six hourly intervals. This specification uses more granular information as naturally contained in the continuous watching intensity measure (compared to the dummy variable operationalization in our main analysis) and also allows for potential non-linearity in the effects without assuming a specific functional form. Web Appendix B.I shows the coefficient estimates. The results are largely consistent with the results from our main analysis and we observe the familiar inverse U-shaped effects of watching intensity similar to those displayed in Figs. 7 and 8.
In a second robustness check, we also utilize the continuous watching intensity variable and assume that it enters our model linearly and quadratically. The coefficient estimates are displayed in Web Appendix B.II and exhibit the familiar inverse U-shaped pattern (with the exception of Fig. 13(g) and (h) just as in our main analysis).
Third, in our main model, we control for potential confounders through high-dimensional fixed effects and use the portion of variability in watching intensity that cannot be explained by these fixed effects to estimate the effects of watching intensity on media franchise engagement. To further test our assumption that these fixed effects control for potentially omitted variables, we estimate our model without any fixed effects (see Web Appendix B.III for the results). The comparison between the estimation results with and without fixed effects indicates that, without controlling for fixed effects, the estimates might capture not only the effect of watching intensity but also the influence of omitted variables such as individual preferences, seasonal trends, or specific anime characteristics. For example, the negative inverse U-shaped effect of watching intensity on the number of forum posts becomes flatter when including fixed effects, suggesting that some of the observed effects in the absence of fixed effects could be due to omitted variables such as anime- or time-specific factors.
Fourth, considering that a significant portion of our sample comprises younger individuals, we addressed potential concerns regarding the generalizability of our results by re-estimating our models with separate coefficients for younger and older individuals. We found no significant differences, even though around half of the platform’s users are minors (see Web Appendix B.IV).
Fifth, we estimate our regressions once excluding observations where the user adopted only one anime of a specific genre, and once excluding the top 25% of users in terms of animes adopted from our sample, i.e., users who adopted more than 100 animes out of 2,467 animes in our dataset. The results remain robust and are reported in Web Appendices B.V and B.VI.
Lastly, for our two interactive engagement dependent variables that are count variables, number of forums posts and number of recommendations, we reestimate our regressions using fixed effects Poisson model. Our results for the number of recommendations show a similar inverse U-shaped pattern. The results for the number of forum posts become insignificant, potentially due to large number of fixed effects (see Web Appendix B.VII).
6 Managerial implications
Our results offer important managerial implications for TV channels and online streaming platforms. First, watching a media product for more than an hour a day (compared to watching it for less than an hour day, e.g., once a week on a traditional schedule) has overall positive effects on both personal and interactive engagement with the media franchise. To put it differently, while the magnitudes of the effects of watching intensity vary depending on whether an individual watches a media product for, e.g., 1-2 or 3-4 hours a day, the effects of higher watching intensities are all positive compared to watching a media product less than an hour a day. This finding suggests that content providers should make larger amounts of content available at once.
Second, the generally positive effect of watching intensity on personal engagement suggests that content providers can use watching intensity of the current season as a reliable predictor for the success of future franchise extensions in addition to the viewer count of the current season. This information can be particularly useful when the decision of renewing a show has to be made right after the opening weekend, which is the case for many streaming platforms.Footnote 25
Third, taking a closer look at our results, the effect of watching intensity on personal and interactive engagement does not exhibit a straightforward linear relationship. Rather, we find inverse U-shaped effects of watching intensity on several key engagement measures with the largest impact typically found in the interval of 2 to 5 hours of watching an anime per day. Although we do not have direct evidence suggesting that watching shorter seasons yields higher engagement with franchise extensions, this finding nevertheless provides a rationale for the recently increasing popularity of mini series, a format that is longer than a movie but shorter than a typical season and can be finished in 6 to 8 hours of watching (e.g., on a weekend), right on the sweet spot when the maximum effect of watching intensity would be induced. This finding also validates the content release strategy of dropping the first few episodes of a new season all at once to hook the audience and then switch to a more linear fashion of releasing the remaining episodes one by one. For example, Amazon commonly drops the first three episodes of a season at once and then switches to a weekly release schedule (e.g., seasons 2 and 3 of “The Boys,” seasons 1 and 2 of “Wheel of Time”).Footnote 26 Similar release strategies include the split-release of seasons, where a platform splits the release of show so that the amount of new content made available can be consumed by viewers during a weekend (2-4 hours watching per day). Examples include Netflix shows such as “Witcher,” “Strange Things” or “Bridgerton.”Footnote 27
Fourth, watching intensity can boost viewership of franchise extensions. However, the availability of the franchise extension plays a crucial role. Media content providers have started to recognize this by making prior seasons available (for intensive viewing) shortly before the release of the next season. Figure 9 shows several examples from Netflix.
Examples of release dates on Netflix
Fifth, watching intensity does not boost viewership of all franchise extensions equally. Sequels benefit from a prior season that is available for intensive viewing more than other types of franchise extensions because sequels continue the flow viewers experience when watching the prior season intensively. Franchise extensions that differ significantly in story lines and/or main characters may not attract viewers who watched the prior season intensively. The historical lackluster performance of spin-offs speaks to the importance of staying close to the successful original series when developing franchised extensions.Footnote 28
And lastly, watching more intensively induces liking and content providers can target and encourage users who watched a show more intensively to generate favorable word-of-mouth.
7 Conclusion
With the introduction of high-speed Internet during the last 15 years, watching media products intensively – also known as binge-watching – has become very common. An open empirical question is whether the watching intensity has implications for user engagement compared to the traditional, linear way of watching. In this paper, using novel data coming from an online anime platform containing information on individual users’ adoptions of different animes and their UGC, we examine the relationship between watching intensity and consumers’ engagement with a media franchise as related to UGC and the adoption of franchise extensions. Our paper thus adds to the small but rapidly growing body of literature on consumers’ digital media consumption as well as on the online streaming industry. To the best of our knowledge, our paper is the first systematic empirical examination of the effects of watching intensity on user engagement with a media franchise.
Our findings reveal that the level of watching intensity matters for media franchise engagement and that its effects follow an inverse U-shaped pattern with the largest effect around three to five hours per day. Watching intensity increases the probability that a user start to watch a sequel or any other type of franchise extension. However, conditional on starting to watch a franchise extension, watching intensity only increases the probability that a user finishes to watch it if the extension was available at the time when the user watched the focal media series. This finding suggests that the timing of release of a franchise extension and not the type of franchise extension impacts engagement with a franchise: consumers are more likely to complete watching a franchise extension if it was released very close to the consumption the focal media series. We also find that watching intensity increases the probability that a consumer watches a franchise extension immediately after watching the focal media product if the user watched the focal media series for more than three hours per day. We believe these effects are driven by the balance between flow and satiation, two forces created by watching intensity but operating in opposite directions.
Regarding interactive engagement, our results suggest excessive watching (i.e., more than five hours per day) can hurt consumer engagement through a decrease in the number of forum posts a user writes, providing partial support for the avoidance tendency of binge-watchers proposed and documented in previous literature. However, intermediate levels of watching intensity positively affect the submission of recommendations and ratings, the most dominant form of UGC in our empirical context. Moreover, we also find that intermediate watching intensity increases rating valence conditional on rating submission.
Our research is not without limitations. First, a media franchise can also include merchandising items that are available for purchase, such as posters, coffee mugs, toys, and trading card games. In our data, we do not observe purchases of such ancillary products. It is left for future research to investigate whether the watching intensity affects such purchases. Second, even though we provide evidence for the validity of our data, measurement error in watching intensity due to its self-reported nature remains a potential concern. It is well-known that measurement error in an independent variable leads to attenuation bias, i.e. a bias of the coefficient towards zero. Thus our results should be interpreted as a lower bound of the effects of watching intensity.
Third, some TV shows or movies have a higher probability of being watched intensively than others due to their creative content, which we do not consider in our model. It is left for future research to study whether and how different content characteristics make a show more or less suitable for intensive viewing. Fourth, watching intensity of the media franchise can also be impacted by the watching intensity of the focal show. Our data only allows us to observe whether a user has started and finished watching a particular media franchise, and not the intensity of watching the franchise. More granular data on the intensity with which users watch the franchise would add further depth and nuance to our understanding of users’ engagement. And lastly, watching intensity may differ significantly depending on the methods or channels of watching such as online streaming websites, streaming platforms, DVDs, or piracy websites. An interesting direction for future research is to explore how these different channels should design and deploy user interfaces, advertising methods, and sequential watching strategies to influence individuals’ watching intensity.
Notes
An OTT media service is a streaming service that delivers film, TV, and video content directly to viewers over the internet without requiring users to subscribe to traditional cable or satellite pay TV. By 2022, over \(85\%\) of U.S. households subscribed to at least one video streaming service (https://www.kantar.com/north-america/inspiration/technology/85-per-cent-of-us-households-have-a-video-subscription-service). In the U.S., streaming captured more than \(38\%\) of total TV viewing in December 2022 and audiences streamed more than 19 million years worth of content (https://www.nielsen.com/insights/2023/streaming-services-remain-most-popular-destination-for-tv-viewing-in-december/).
In 1994, 1 out of the 10 top grossing movies was a franchise. In 2014, 7 out of the 10 top grossing movies were franchises (https://www.ft.com/content/192f583e-7fa7-11e4-adff-00144feabdc0?mhq5j=e5). In 2022, all 10 top grossing movies were franchises (https://www.boxofficemojo.com/year/2022/).
The total historic revenue from the “Star Wars” franchise was $69 billion (by mid 2021), $32 billion (by mid 2021) for the “Harry Potter” franchise, and $22 billion (by mid 2021) for the “Toy Story” franchise (https://www.statista.com/statistics/1257650/media-franchises-revenue/).
The empirical context of this paper are anime (Japanese cartoon) series. Thus, we use the terms “next season” and “sequel” interchangeably.
Note that all users in our estimation sample watch the whole anime season and that all episodes in a season are available for watching in our estimation sample. However, some finish watching the whole season in fewer days than others, i.e., we study the effects of watching intensity conditional on season completion.
Recently, a process-based definition of binge-watching has been suggested (see, e.g., Schweidel and Moe, 2016, Lu et al. 2017). It poses that streaming data are necessary to classify viewing behavior as binge-watching, i.e., the researcher has to observe that a consumer watched at least two episodes of a series directly in a row. Our data contain information on how many minutes of a series a consumer watched on average per day, but we do not observe whether this watching happened in one sitting. For more details, please see Sections 2.2 and 3.1.
The term “anime season” also subsumes movies, i.e., an anime season with one episode.
We refer readers to Brodie et al. (2011) for an extensive review of the marketing literature on engagement.
Many regard the element of control, i.e., the consumer’s control over whether to watch more episodes, as an essential part of binge-watching. In other words, binge-watching is not only about watching multiple episodes in one sitting, but also about a consumer’s control and decision of when and what to watch. Thus a consumer watching a marathon release on TV would not count as binge-watching.
We do not account for platform choice in this paper because, in general, users can watch animes either legally or illegally through a number of different channels such as netflix.com, hulu.com, funimation.com, crunchyroll.com, aniplexusa.com and others.
Individuals’ behavior on MyAnimeList.net is consistent with the well-known 90-9-1 rule in social media (see, e.g., https://www.nngroup.com/articles/participation-inequality/): a large proportion of individuals is inactive.
A larger number of animes have 12-13 or 26 episodes in a season.
Because of missing values in one of our explanatory variables (popularity rank), the number of observations in the model estimation is 658,542.
Note that a user might have watched more than 3 hours on a Sunday, but it took her Monday to Friday to gradually watch the remaining 3 episodes (about 1 hour) and finish the season. Our data do not allow us to compute different watching intensities within a user-anime combination.
Note that we also control for an anime’s time-varying popularity and (perceived) quality by using rankings and ratings at the time of focal anime adoption as control variables.
MyAnimeList.net assigns each anime to 3 - 4 out of 44 available genres (e.g., action, comedy, fantasy). We use this genre assignment to construct the user-genre fixed effects.
Note that users are located all around the world and that we use the local holidays and weekend definitions when constructing the variable.
Note that we control for observed anime characteristics in our empirical model.
The popularity rank is based on the number of users who adopted the anime. The community rating is the average rating of users who watched the anime. Both variables vary across time and users can see both of them on the platform.
Note that for the personal engagement form of whether to watch a franchise extension immediately next after watching the focal anime, we condition on a franchise extension being available at the time of watching the focal anime. Thus, for that equation, we only estimate the effects of watching intensity under two scenarios: when the extension is a sequel, and when the extension is not a sequel.
We thank the AE for their constructive suggestions on this issue.
Industry experts confirmed this practice in conversations.
Zhao et al. (2023) also suggest such a release strategy for online fiction.
Wikipedia lists 1,142 TV spin-offs on its website (https://en.wikipedia.org/wiki/List_of_television_spin-offs). Only 135 spin-offs (12%) ran for 5 or more seasons. 413 spin-offs (36%) ran for one season or less.
References
Aarseth, E. (2006). The culture and business of cross-media productions. Popular Communication, 4(3), 203–211.
Allenby, G., & Rossi, P. (2019). Inference for Marketing Decisions. In Dube, J.-P. & Rossi, P. eds, ‘Handbook of the economics of marketing’. North-Holland, pp. 69–150
Ameri, M., Honka, E., & Xie, Y. (2019). Word-of-mouth, observed adoptions, and anime watching decisions: The role of the personal versus the community network. Marketing Science, 38(4), 567–583.
Angrist, J., & Pischke, J.-S. (2009). Mostly Harmless Econometrics: An Empiricist’s Companion (1st ed.). NJ: Princeton University Press.
Bowden, J. (2009). The process of customer engagement: A conceptual framework. Journal of Marketing Theory and Practice, 17(1), 63–74.
Brakus, J., Schmitt, B., & Zarantonello, L. (2009). Brand experience: What is it? how is it measured? does it affect loyalty? Journal of Marketing, 73(3), 52–68.
Brodie, R., Hollebeek, L., Juric, B., & Ilic, A. (2011). Customer engagement: Conceptual domain, fundamental propositions, and implications for research. Journal of Service Research, 14(3), 252–271.
Calder, B., Malthouse, E., & Schaedel, U. (2009). An experimental study of the relationship between online engagement and advertising effectiveness. Journal of Interactive Marketing, 23(4), 321–331.
Chou, T., & Ting, C.h. (2003). The role of flow experience in cyber-game addiction. CyberPsychology & Behavior, 6(6), 663–675.
Csikszentmihalyi, M. (1997) Finding Flow: The Psychology of Engagement with Everyday Life. Basic Books
Devasagayam, R. (2014). Media bingeing: A qualitative study of psychological influences. Proceedings of the Marketing Management Association pp. 40
Fisher, R. J. (1993). Social desirability bias and the validity of indirect questioning. Journal of Consumer Research, 20(2), 303–315.
Furnham, A. (1986). Response bias, social desirability and dissimulation. Personality and Individual Differences, 7(3), 385–400.
Hernandez, B. (2014). All ‘breaking bad’ episodes are now on netflix. http://mashable.com/2014/02/24/all-breaking-bad-netflix/#XRHTdKSPkmqY
Hoffman, D., & Novak, T. (1996). Marketing in hypermedia computer-mediated environments: Conceptual foundations. Journal of Marketing, 60(3), 50–68.
Hoyer, W., Chandy, R., Dorotic, M., Krafft, M., & Singh, S. (2010). Consumer cocreation in new product development. Journal of Service Research, 13(3), 283–296.
Jenner, M. (2017). Binge-watching: Video-on-demand, quality tv and mainstreaming fandom. International Journal of Cultural Studies, 20(3), 304–320.
Kozinets, R. (1999). E-tribalized marketing?: The strategic implications of virtual communities of consumption. European Management Journal, 17(3), 252–264.
Kubey, R., & Csikszentmihalyi, M. (2002). Television addiction is no mere metaphor. Scientific American, 286(2), 74–81.
Lovett, M., & Staelin, R. (2016). The role of paid, earned, and owned media in building entertainment brands: Reminding, informing, and enhancing enjoyment. Marketing Science, 35(1), 142–157.
Lu, T., Bradlow, E., & Hutchinson, W. (2017). Binge consumption of online content. Working Paper
MarketCast (2013). Marathon tv: How binge-viewing is changing the way we watch. http://www.mcast.com/bingestudy/#sthash.oeqhmvqi.dpuf
Mollen, A., & Wilson, H. (2010). Engagement, telepresence and interactivity in online consumer experience: Reconciling scholastic and managerial perspectives. Journal of Business Research, 63(9–10), 919–925.
Nambisan, S., & Nambisan, P. (2008). How to profit from a better ‘virtual customer environment. MIT Sloan Management Review, 49(3), 53–61.
Netflix (2013). Binge watching is the new normal. https://www.prnewswire.com/news-releases/netflix-declares-binge-watching-is-the-new-normal-235713431.html
Pena, L. (2015). Breaking Binge: Exploring the Effects of Binge Watching on Television Viewer Reception, PhD thesis, Syracuse University
Schweidel, D., & Moe, W. (2016). Binge watching and advertising. Journal of Marketing, 80(5), 1–19.
Steenkamp, J. B. E., De Jong, M. G., & Baumgartner, H. (2010). Socially desirable response tendencies in survey research. Journal of Marketing Research, 47(2), 199–214.
TiVo (2015). Original streamed series top binge viewing survey for first time. https://www.4rfv.com/0NQZVITMU8CC/original-streamed-series-top-binge-viewing-survey-for-first-time.htm
Van Doorn, J., Lemon, K., Mittal, V., Nass, S., Pick, D., Pirner, P., & Verhoef, P. (2010). Customer engagement behavior: Theoretical foundations and research directions. Journal of Service Research, 13(3), 253–266.
Vivek, S., Beatty, S., & Morgan, R. (2012). Customer engagement: Exploring customer relationships beyond purchase. Journal of Marketing Theory and Practice, 20(2), 122–146.
Wooldridge, J. (2001). Econometric Analysis of Cross Section and Panel Data (1st ed.). Cambridge: MIT Press.
Zhang, Y., Bradlow, E., & Small, D. (2013). New measures of clumpiness for incidence data. Journal of Applied Statistics, 40(11), 2533–2548.
Zhang, Y., Bradlow, E., & Small, D. (2015). Predicting customer value using clumpiness: From rfm to rfmc. Marketing Science, 34(2), 195–208.
Zhao, C., Mehta, N., & Shi, M. (2023). An empirical investigation of a digital platform’s release strategy for serialized content. Working Paper
Acknowledgements
All errors are our own. The authors thank Pradeep Chintagunta, Sanjay Sood, Brian Ratchford, seminar participants at Emory University, Fudan University, and Shanghai University of Finance and Economics, and attendees of the 2017 Marketing Science/INFORMS conference for their comments.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of Interest
The authors did not receive support from any organization for the submitted work. All authors certify that they have no affiliation with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript. Elisabeth Honka serves on the Editorial Board of the Quantitative Marketing and Economics journal. Elisabeth Honka is not aware of any competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Web Appendix A: Daily watching behavior in 2014 American time use survey
Data are available on the Department of Labor website at https://www.bls.gov/tus/data/datafiles_2014.htm.
Daily watching - American time use survey 2014
Web Appendix B: Robustness checks
In this section, we demonstrate the robustness of our results with respect to using a different operationalization of the watching intensity variable in the model and a different model specification.
1.1 I. Continuous operationalization of watching intensity
In the first robustness check, we include the continuous version of the watching intensity variable in the model, but estimate separate coefficients for each of the six hourly intervals. This specification utilizes more granular information as naturally contained in the continuous watching intensity measure (compared to the dummy variable operationalization in our main analysis) and also allows for potential non-linearity in the effects without assuming a specific functional form.
Formally, we estimate the following equations for personal and interactive engagements, respectively:
and
\(B^k_{ij}\), \(k \in \{<1, 1-2, 2-3, 3-4, 4-5, 5+\}\), are dummy variables for watching intensity indicating whether user i watched anime j for k number of hours per day (similar to the dummy variables we described in Eq. 1)). \(w_{{ij}}\) is the number of hours per day user i spent watching anime j. The associated coefficients, e.g., \(\beta _1\), \(\beta _2\), \(\beta _3\), \(\beta _4\), \(\beta _5\), and \(\beta _6\), estimate the marginal effect of watching intensity for each of the six hourly intervals separately. All remaining variables are defined as in Eqs. 1 and 2.
Tables 8 and 9 show the coefficient estimates. The results are largely consistent with the results from our main analysis. For personal engagement, watching intensity increases a user’s probability of starting to watch a franchise extension (at any point in time in the future) for all six hourly intervals. However, watching intensity only positively affects a user’s likelihood of finishing to watch a franchise extension when the extension was available at the time when the user watched the focal anime. And lastly, watching intensity of more than three hours a day increases a user’s likelihood of watching a franchise extension immediately next after watching the focal media product. For interactive engagement, the effect of watching intensity on UGC production depends on the type of UGC: it decreases the number of forum posts but increases the recommendation and rating submissions. Conditional on submitting a rating, it also positively affects rating valence.
It is important to recognize that the estimated coefficients for each level of watching intensity indicate the slope of the effect of watching intensity on the dependent variable during each of the six hourly intervals. By comparing the magnitudes of these coefficients, we can determine whether the marginal effect of watching intensity (e.g., another 6 minutes or 0.1 hour) on the dependent variable increases or decreases as we move from one hourly interval of watching intensity to the next. However, these coefficients do not directly translate to a change in the level of engagement across the hourly watching intensity intervals. To understand the absolute changes in engagement at each intensity level, we calculate the predicted change in the dependent variable at the median value of watching intensity for each hourly interval. When plotting these changes in Figs. 11 and 12, we observe the familiar inverse U-shaped effects of watching intensity similar to those displayed in Figs. 7 and 8.
Predicted change in personal engagement for median watching intensity by hourly interval
Predicted change in interactive engagement for median watching intensity by hourly interval
Effects of watching intensity on personal engagement
Effects of watching intensity on interactive engagement
1.2 II. Linear and quadratic specification of watching intensity
In a second robustness check, we also utilize the continuous watching intensity variable and assume that it enters our model linearly and quadratically, i.e.,
and
The coefficient estimates are shown in Tables 10 and 11 and Figs. 13 and 14 display the familiar illustrations exhibiting an inverse U-shaped pattern (with the exception of Fig. 13(g) and (h) just as in our main analysis).
1.3 III. No fixed effects
1.4 IV. Age group estimates
1.5 V. Excluding extreme adoption users
1.6 VI. Excluding singleton user-genre observations
1.7 VII. Poisson regressions for count variables
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Ameri, M., Honka, E. & Xie, Y. Watching intensity and media franchise engagement. Quant Mark Econ (2024). https://doi.org/10.1007/s11129-024-09283-2
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s11129-024-09283-2
Keywords
- Watching intensity
- Media consumption
- Binge-watching
- Media franchise
- Consumer engagement
- Online movie streaming