Mobile news access, mobile news repertoires, and users’ tendency to talk about the news – an experience sampling study on mobile news consumption

2.1 Inter- and intra-personal variation in news consumption

Across their daily lives, people encounter the news in many different situations. They might wake up in the morning listening to their favorite news podcast and then, throughout the day, check a news website on their phone during a commute, get a chat message referring to a news article, or encounter news items in social media posts. These examples demonstrate that news access, and most importantly mobile news access, varies considerably within one person from one situation in their daily lives to another. Put differently, we can uncover intra-individual variation when studying mobile news access.

At the same time, we also observe differences between people. Some people prefer listening to news podcasts, others heavily rely on news being pushed to them primarily through social media (as captured in the concept of the news-finds-me perception; Gil de Zúñiga and Diehl 2019), and others still enjoy reading news articles in a news app. However, a recent meta-analysis has revealed that the bigger part of the variation in media use resides within the person, not between different users (Schnauber-Stockmann et al. 2023). To gain a more comprehensive understanding of mobile news consumption, we need to acknowledge and study this intra-personal variation in mobile news access, i.e., we need to study inter- and intra-personal variation simultaneously.

2.2 Changes in news access

Media ecologies have changed tremendously during the past decades, transforming former low-choice to high-choice media environments (Van Aelst et al. 2017). Not at least, these changes have also affected news consumption processes. Current news exposure can occur via many communication channels and devices, with users encountering multiple sources and news providers. The underlying mechanisms leading to this situation are manifold and often overlapping. Thorson and Wells (2016) proposed a compelling systematization in their curation framework, differentiating several curation processes influencing and governing current news exposure and use. Without discussing this framework and the different curation processes in detail, we want to draw attention to the fact that these different, simultaneously ongoing curation processes – supposedly – lead to considerable intra-personal heterogeneity in news access in our current media ecologies.

This intra-personal variance in accessing the news is rarely considered by research on news consumption and its effects. Instead, extant research mainly describes and investigates the impact of inter-personal differences (for an exception, see the tradition of media and news repertoires discussed below). To better grasp these current heterogeneous news behaviors and thus provide a solid basis to assess their consequences, we need to look at these intra-personal differences. These intra-personal differences have, of course, already been researched in the age of legacy media (i.e., before the massive diffusion of online media), especially in the tradition of (contextual influences on) media selection processes (e.g., Webster and Wakshlag 1983). However, the rise of mobile media, i.e., ubiquitously usable devices such as the smartphone, has increased the interest in this perspective due to mobile media’s inherent nature of being used ubiquitously and across many different situations (Karnowski 2020; for an overview of studies employing this perspective see Schnauber-Stockmann and Karnowski 2020; Schnauber-Stockmann et al. 2023).

Therefore, rather than studying our overall media ecologies, we will concentrate on mobile media, specifically the smartphone, in this study for two reasons. First, as mentioned above, interest in intra-personal variation in media use has recently increased in the context of mobile media (for news consumption, see, e.g., Struckmann and Karnowski 2016; van Damme et al. 2020). Second, and most importantly, mobile media, such as smartphones, can be conceptualized as meta-media (Jensen 2016). Humphreys and colleagues argue that this results in their individual uses being undetermined, programmable, and configurable by users (2018). An unlimited number of constituent media, such as apps, can be embedded by users into the structure of the meta-medium, leading to individual and constantly changing configurations. For example, users can embed a radio application, access the news via a newspaper website using the smartphone’s browser, or use a social media application like Facebook. This individual assemblage of constituent media is constantly and freely (re)configurable by each user. Thus, these meta-media can be seen as a microcosm reflecting the high-choice environment of our broader media ecologies in one device.

We must analyze these constituent media used to analyze intra-personal variation in mobile news access. According to the framework suggested by Humphreys et al. (2018), we should investigate their specific features instead of just asking which constituent media are used. We, therefore, analyze two main features of constituent media. First, because smartphones are multimedia devices, they are open to all reception modes (i.e., reading, listening, and watching). The reception mode can thus be seen as one fundamental feature of constituent media worth analyzing. Second, constituent media can be classified by their broader class or type of platform (Humphreys et al. 2018), such as social media or news apps, for example. Hence, to assess types of mobile news access, we ask:

2.3 Mobile news repertoires

To provide a more comprehensive basis for assessing possible consequences of media use, we need to focus on inter-personal variation simultaneously. One way to grasp the intra-individual heterogeneity in media use and condense it is through media repertoires, which refer to “relatively stable cross-media patterns of media practices” (Hasebrink and Hepp 2017, p. 367). However, extant studies on news repertoires mainly rely on cross-sectional surveys and users’ overall assessment of usage frequencies of different platforms (e.g., Oh et al. 2021; Yuan 2011). We argue that such an approach cannot disentangle intra-personal from inter-personal variance in news consumption. Instead, we need to dig deeper and assess repertoires based on users’ actions captured in situ. Ohme et al. (2016) have adopted a similar approach to study exposure to political information, confirming the feasibility of such an approach. Studying (mobile) news use, we concentrate on users’ (mobile) news repertoires as part of their broader media repertoires (Peters and Schrøder 2018) and ask:

2.4 Communication as an effect of communication

Expected future interpersonal conversations about news are one of the strongest motives for news consumption and one of its main effects (Gehrau and Goertz 2010). This effect of news use is highly relevant to our democracies. It leads to a stronger sense of and reflection on one’s own opinion (de Boer and Velthuijsen 2001) and encourages civic and political involvement (Vaccari and Valeriani 2018). Along with the aforementioned changes in our media ecologies, the possibilities to access news and communicate about them have increased tremendously. News are no longer only a topic of conversation in face-to-face communication (Calabrese and Jenard 2018), but news can also be commented on and distributed on social media or discussed directly on news websites.

Whether people are willing to talk about news depends on personal, content, and context characteristics. In this study, we will focus on personal and contextual factors. Overall, general interpersonal communication motives (e.g., general tendencies to speak) and individual news consumption perceptions (e.g., news-finds-me perception) are important influences on whether people talk about the news (Park and Kaye 2020; Porten-Chee 2017; Ziegele 2016). In addition, the context of news use, such as the medium used to access the news, is also influential. For example, television news is discussed more often than news from the newspaper or radio (Gehrau and Goertz 2010). Considering our case of mobile news use, we assume broad intra-personal variance in specific types of mobile news access (RQ1), condensed in mobile news repertoires (RQ2). To assess the consequences of mobile news repertoires along with users’ characteristics, we therefore ask:

3.1 Measures

Online survey. In the pre-questionnaire, we assessed users’ characteristics. News-find-me perception was measured using the following four items (Gil de Zúñiga et al. 2017) on a scale of 1 (= do not agree at all) to 5 (= agree completely): “I rely on my friends to tell me when something important happens in the news,” “I feel well informed, even if I don’t actively follow the news,” “I don’t worry about staying up to date because I know the news will find its way to me,” and “I rely on receiving news and information based on what my friends link to and follow on social media” (M = 2.18, SD = 0.70, Cronbach’s α = 0.64).

Tendency to talk about the news was assessed with four items derived from Friemel’s (2013) theoretical considerations, assigning these conversations both an information function (talking about news) and a participation function (discussing news). Furthermore, a distinction was made as to whether the conversation about news occurs online or offline. These considerations led to the four possible courses of action: “Spread news online (e.g., by sharing the message, linking someone under it, or sending the message),” “Discuss news online (e.g., by liking the message, commenting on it, or replying to a comment),” “Tell news to someone in a personal conversation or on the phone,” and “Discuss news with someone in a personal conversation or on the phone.” The participants were asked to indicate on a scale of 1 (= very rarely) to 5 (= very often) how often they engage in these activities (M = 2.61, SD = 0.60, Cronbach’s α = 0.61).

3.2 MESM study

The in-situ questionnaires focused on participants’ mobile news access within the last hour. We asked about the features of apps used for news access and potential interpersonal conversations following news use.

First, the participants were asked to indicate whether they had read, watched, or listened to the news on their smartphone within the previous hour (reception mode). If the participants had consumed news several times during that period, they were asked to answer regarding the most recent news use situation.

Second, we asked about the type of platform used. Depending on the reception mode participants had selected, they could choose from limited options to keep the in-situ questionnaire as brief as possible. Overall, the types of platforms used included a website via a browser, a news app, a news widget for the smartphone, a social media app, a messenger app, an e-mail, a media library/streaming service, and a web radio/radio app.

We further subdivided social media and websites, as these platforms represent broad categories. When participants chose social media, they were asked to distinguish between Facebook, Instagram, Twitter, YouTube, Snapchat, and others (type of social media). For websites, we asked participants to differentiate between news websites and other websites (type of website).

Finally, following the operationalization used to assess users’ overall tendency to discuss the news (see above), we asked respondents whether they had talked about the news consumed in this situation (talking about news) with others or intended to do so (intention to talk about news).

3.3 Analytical strategy to assess types of mobile news access and types of mobile news repertoires

Data collection resulted in a hierarchical structure with (multiple) news access situations (560 news access situations) nested in participants (72 participants). Therefore, the need for specific statistical tools that account for non-independent data must be assessed. To simultaneously identify types of mobile news access (Level 1/L1; 560 situations) and types of users’ mobile news repertoires (Level 2/L2; 72 participants) (RQ1 & 2), we therefore conducted a multilevel latent class analysis (MLCA; Lukočienė et al. 2010) using Latent GOLD (Magidson and Vermunt 2016).

In its basic form, latent class analysis (LCA) allows the identification of clusters of homogeneous units regarding the clustering variables, assuming the existence of a discrete, unobserved variable with K categories triggering the homogeneity among the clustered units. LCA further hinges on the assumption of local independence (Lazarsfeld and Henry 1968).

However, as the current data set has a multilevel structure (situations nested in persons), a multilevel latent class analysis (MLCA; Vermunt 2008) must be applied. Following the strategy suggested by Vermunt (2008), we address the multilevel structure of the data set by introducing a group-level discrete latent variable (here: person-level) and allowing the model parameters to vary across latent classes of persons. The basic idea of this approach (i.e., the introduction of a person-level discrete latent variable) is that persons also belong to one of L person-level latent classes with a specific class membership of person j and a specific value of the unobserved variable, that is, a specific person-level latent class l (statistical model see Vermunt 2008, p. 38). These latent classes of persons can vary in the probability that situations of persons belong to the latent class k (i.e., on the situation-level). In the multilevel context, local independence then refers to the independence of situation response patterns of one another, given that person-membership is controlled for (Vermunt 2008).

To identify meaningful lower-level classes while considering the structure of the data, a model is calculated in which situation-level class membership probabilities vary across person-level classes but does not allow the parameters that define the class-specific conditional distributions for the situation-level variables to vary across person-level classes (Vermunt 2008). These calculations return several parameters, namely latent class probabilities and conditional probabilities. Latent class probabilities describe how classes of the latent variable are distributed. They indicate both the number and relative sizes of the classes. Conditional probabilities indicate the probability that a participant in a given latent class will be at a specific level of the observed variable.

4.1 Types of mobile news access (RQ1)

As elaborated above, we conducted an MLCA based on the type of platform and reception mode. Following our operationalization discussed above, this includes the variables reception mode, type of app used, type of social media, and type of website. To identify the number of situation-level and person-level classes, we compared several models following a three-step procedure by Lukočienė et al. (2010, procedure see foot of Table 1). As a general guideline, the solution with the lowest Bayesian information criterion (BIC, based on the number of higher-level cases) is considered the best solution (Lukočienė et al. 2010). This procedure resulted in a solution with five types of news access on the situation-level and four types of mobile news repertoires on the person-level, which will be described in the following (see Table 1 for model comparisons).

Regarding RQ1, we identified five types that reflect a broad spectrum of mobile news access. These types can be described as follows (see Table 2, class size in percent of the total number of situations in brackets):

Direct app access (41.2 %). This type stands out because it is the most frequent mobile news access. News are primarily read and accessed via a news application or the news widget of users’ smartphones. Hence, users are directly exposed to journalistically curated content, so this class is called ‘direct app access.’ There is also a small probability of accessing news via a messenger app or e-mail.

Listening & watching (16.9 %). This type is characterized by a high probability of listening to or watching news accessed via media-library/streaming-service apps (e.g., Zattoo, Spotify) and specific (web) radio apps.

Social media (textual) (15.6 %). The reception mode of reading characterizes the first of two social media-based types of mobile news access. When belonging to this type, there is a 54 % chance of reading news on Facebook and a 32 % chance of reading news on Twitter.

Social media (visual) (14.0 %). In contrast to the previous type, this social media-based type of mobile news access is characterized by reading and watching. However, what sets this type apart is the high probability of accessing news on Instagram and a more negligible probability of accessing the news via YouTube.

Direct website access (12.3 %). The smallest type of mobile news access is like the first in that it is almost exclusively characterized by reading the news. In contrast to the first type, in this type, news are accessed mainly via a news website (again representing journalistically curated content) and, in some cases, via other websites such as provider home pages.

4.2 Mobile news repertoires (RQ2)

As iterated above, we assume that the prevalence of these types of mobile news access varies within and between users. Multilevel LCA also identified four types of mobile news repertoires (see Table 1), indicating the prevalence of types of mobile news access across different users’ daily lives. These mobile news repertoires combine multiple types of mobile news access, with none relying on only one type. In addition, although these mobile news repertoires differ considerably in their combinations of types of mobile news access, all contained at least some amount of direct access to journalistic products (see Table 2, class size in percent of the total number of users in brackets):

Direct news access (34.1 %). Direct app access characterizes this repertoire. Hence, users mainly consume content from journalistically curated news applications or news websites.

Mixed readers (31.0 %). Users in the mixed readers repertoire show high variability in their mobile news access. The mobile news access types they engage in mostly are social media (textual) and – to a lesser degree – direct app access and direct website access. In contrast, the type of social media (visual) is underrepresented in this repertoire.

Visual social media and news apps (21.8 %). This repertoire is characterized by a high probability (54 %) of containing the mobile news access type social media (visual), supplemented by direct app access.

Listening and news apps (13.1 %). Participants showing this news repertoire constitute the smallest class. This repertoire also contains a mix of mobile news access types, with the highest likelihood of listening & watching (54.8 %). Social media (textual) is not part of this repertoire.

4.3 Tendency to talk about the news used on mobile media (RQ3)

RQ3 asked which mobile news repertoires go along with talking about the consumed news. We conducted two multiple linear regressions to answer this research question (see Table 3). Dependent variables were the actual behavior of talking about the news and the intention to talk about the news. Both variables were created by aggregating the binary situation variables via the mean, thus indicating the overall frequency with which a participant talked or intended to talk about the news. In 18.2 and 27.1 % of the situations, the participants talked about or intended to talk about the news they had just consumed. We entered the conditional probabilities of the person-level repertoire classes as predictors. We used class 2, the class with the lowest frequency of both dependent variables, as reference category. In addition, we controlled for person-level characteristics (gender, news-finds-me perception, tendency to talk about the news).

First, looking at the behavioral variable of talking about the news, we can see that only the general tendency to talk and gender emerged as significant predictors. The higher the general tendency to talk, the more participants spoke about the news. Men talked slightly more about the news than women. Overall, the model explained 11 % of the variance. Next, looking at the intention to talk about the news, we can see only one significant predictor. The more probable a user belongs to the mobile news use repertoire of visual social media and news apps, the more they intend to talk about the news compared to the reference cluster Mixed readers. Overall, the model explained 17 % of the variance.