A Test of the Mobile Phone Appropriation Model: A Comparison between Chinese and US Samples

The MPA Model

The MPA model developed by Wirth et al. (2007, 2008) proffers an integrative framework for explaining individuals’ mobile phone appropriation patterns from a communication perspective. This model synthesizes factors from five extant frameworks: diffusion of innovation (Rogers, 2003), theory of planned behavior (Ajzen, 1985), cultural studies (Silverstone & Haddon, 1996), frame analysis (Höflich, 2003), and uses and gratifications theory (Katz et al., 1973). Building upon previous frameworks, the MPA model conceptualizes mobile phone appropriation as a dynamic and creative process rather than binary choices and attempts to merge the divergence between qualitative and quantitative perspectives. The model contains three major components (see Figure 1): metacommunication, usage beliefs, and actual usage behavior.

Figure 1.

The MPA Model Proposed by Wirth et al. (2008)Note. Wirth, W., von Pape, T., & Karnowski, V. (2008). An integrative model of mobile phone appropriation. Journal of Computer-Mediated Communication, 13(3), 593–617. (CC BY 4.0) p. 606, Figure 1.

The first dimension, metacommunication (i.e., communication about mobile phones), originally included three subdimensions: (a) interpersonal communication concerning features and functions of mobile phones, (b) mass communication regarding mobile phone usage, and (c) observations of other people’s usage behavior. The second dimension, evaluation, is people's assessment of (a) functional, (b) symbolic, (c) normative, and (d) restriction dimensions of mobile phones. Finally, actual usage of mobile phones contains two parts: functional (i.e., control, distraction/ pastime, management of daily schedule, and relationship maintenance) and symbolic (i.e., social and psychological) usage. Starting from metacommunication, the MPA model posits individuals develop beliefs about different usage dimensions, which eventually impact their usage behavior, both functionally and symbolically (Wirth et al., 2008). Presenting mobile phone appropriation as a constant circulation, the model strives to capture the complex dynamics of everyday mobile phone usage.

The MPA model has made an impact in the field of mobile communication research. The 89-items scale accompanying this model was originally developed in German by Wirth et al. (2007) and later translated into English by Lee et al. (2016). Most subscales achieved good reliability after modification with the English scales via confirmatory factor analyses (CFAs). On the qualitative side, Humphreys et al. (2013) conducted interviews guided by MPA in the US and Germany to explore the perceptions of those characteristics of mobile Internet as well as extractive versus immersive usage, surprisingly finding little cross-cultural difference in metacommunication and handling of mobile Internet. Research by Aricat et al. (2015) simplified the circular model and focused on the interactions amongst three components: metacommunication, pragmatic usage, and symbolic usage (usage prestige and social identity). The findings revealed four types of mobile phone users (i.e., convenience seekers, experimenters, group communicators, and tabula rasa) in an Indian Malayali community in Singapore, related with three mindsets toward migrant acculturation (i.e., culture campaigner, culture connoisseur, and culturally petrified). On the quantitative side, Lee and Cionea (2023) further examined the relationships between the proposed factors with a US sample and found the overall MPA model received partial support via structural equation modeling. These studies established MPA to be a viable theoretical framework for assessing mobile phone appropriation across cultures.

As a follow-up study to Lee and Cionea (2023), the present study first analyzes the MPA model’s compatibility with a Chinese sample using a translated (and back-translated) Chinese scale. Based on the original MPA model proposed by Wirth et al. (2008), the initial hypotheses are proposed as below, which are later revised based on the results of the multigroup CFA in the current paper.

H1: Levels of metacommunication will be positively associated with levels of (a) functional, (b) symbolic, (c) normative, and (d) restrictive evaluations.
H2: Levels of functional evaluations will be positively associated with levels of (a) functional and (b) symbolic aspects of usage.
H3: Levels of symbolic evaluations will be positively associated with levels of (a) functional and (b) symbolic aspects of usage.
H4: Levels of normative evaluations will be positively associated with levels of (a) functional and (b) symbolic aspects of usage.
H5: Levels of restrictive evaluations will be negatively associated with levels of (a) functional and (b) symbolic aspects of usage.

Cultural Influences on Mobile Phone Appropriation

Cross-Cultural Comparisons of MPA

As discussed above, we expect to observe some nuances in mobile phone appropriation despite overarching cross-cultural similarities. Given the large IC difference, Chinese people’s beliefs and usage may be more closely linked to what is considered socially normative in the cultural context than Americans’ (Meng & Kim, 2020). With high collectivistic values, Chinese people may pay more attention to other community members’ opinions on social occasions and attach greater importance to communication regarding social norms of mobile phone usage than Americans. As the metacommunication component of the MPA model spotlights how personal beliefs and evaluations of mobile phones can be jointly shaped by personal connections, social environment, and mass media and highlights the social discourse of mobile phone appropriation the effects of metacommunication may be stronger in China.

It is also possible that mobile phones are appropriated with more collectivistic purposes (e.g., social obligations, cohesion, support) in China than in the US due to the cultural emphasis on kinship and relationship maintenance. Predictably, Chinese people may draw greater connections between metacommunication and relationship maintenance as well as control. Furthermore, Chinese people may attach greater symbolic values to mobile phones, encompassing both social status display and psychological attachment, as indicated by past research on purchase intentions (Cui et al., 2007; Hoang, 2015; Lee et al., 2013).

The patterns of mobile phone appropriation may differ across cultures owing to distinct cultural norms and values, and the strength of cultural norms and social influence may also vary across cultures. Thus, we consider whether the associations between different components significantly differ by culture with established sample comparability. Although we could speculate over the potential differences, we propose the following research questions to maximize parsimoniousness in multigroup comparisons:

RQ1b: What differences, if any, are present in the factor structure of the MPA in the Chinese and US sample?
RQ1c: What differences, if any, are present in the path magnitude of the MPA in the Chinese and US sample?

Participants

Both Chinese and US samples were drawn from general populations, and the sampling parameters for the Chinese sample were set in the way to maximize its comparability with the US sample, which was collected by hiring the same online survey panel service. There were 510 completed surveys in total by Chinese and 502 by US participants. For the Chinese sample, approximately half of the participants were males (n = 260), and half were females (n = 250). For the US sample, there were 251 male participants, 250 female participants, and 1 non-binary participant. For the Chinese sample, age was distributed as follows: 18 to 29 (20.6%); 30 to 39 (21.96%); 40 to 49 (21.6%); 50 to 59 (21.6%); and above 60 (14.31%). For the US sample, age was distributed as follows: 18 to 29 (19.9%); 30 and 39 (19.9%); 40 to 49 (20.1%); 50 and 59 (19.9%); and above 60 (20.1%).

Most Chinese participants were Han Chinese (95.9%), followed by Hui Chinese (0.8%), Uighurs (8.5%), Mongolians (0.6%), Tibetans (0.2%), and others (2.5%). Many American participants were non-Hispanic white (73.5%), followed by Hispanic (9.0%), African American (8.8%), Asian American (6.6%), native American (0.6%), and others (1.4%). For educational backgrounds, 75.7% of Chinese participants had college education, with 17.5 % high school education or below and 6.8% graduate-level education; whereas 61.8 % of American participants had college education, with 22.5 % high school education or below and 15.7% graduate-level education. The medium monthly income of Chinese participants was between 7,000 and 9,999 RMB (between approximately 1000 and 1,500 USD), and the medium monthly income for US participants was between 4,000 and 4,999 USD.

Procedures

The data for both samples were collected through a professional survey company, SurveyMonkey, during summer 2016. As compensation, the researchers paid $5 per response to the company for collecting data from their Chinese and US panelists. Two bilinguals between Chinese and English participated in translation and back-translation of the Chinese MPA scales. Before filling out the survey, participants were informed of the general study purpose and asked to report their demographic information. See Appendix A for descriptive statistics of the key variables across cultures and scale reliability scores and Appendix B for the Chinese and English measures. The reliability scores reported below are calculated with two samples combined.

Measures

Multigroup Confirmatory Factor Analysis of MPA Scales and Invariance Testing

Before running structural equation modeling (SEM), a multigroup confirmatory factor analysis (MGCFA) was conducted with the R package, lavaan 0.6-14 (Rosseel, 2012) software with maximum likelihood (ML) estimation. Hu and Bentler (1999) emphasize that “it is difficult to designate a specific cutoff value for each fit index,” but nonetheless suggest values for a “relatively good fit” (RMSEA ≤ .06, CFI ≥ .95, and SRMR ≤ .08; p. 449). High correlations (Tabachnick & Fidell, 2019) were found between three subconstructs of metacommunication (r between .94 and .96 across samples). Likewise, the FEDIS and FEORG were highly correlated in the Chinese sample (r_China = .97, r_US = .75), and FECONT and FECONN were also highly correlated (r_China = .94, r_US = .98). SESO and NE were also highly correlated (r_US = .92 and r_China = .99). Each of these values indicated multicollinearity, and the subdimensions were collapsed for (a) metacommunication and (b) FECONN and FECONT, as advised by Brown (2015). Because FEDIS and FEORG are more varied constructs, and because FEDIS had two poorly loading indicators, only FEORG was retained. Further, given the construct overlap between NE and SESO and the poor loadings for six NE indicators across samples (as in Lee & Cionea, 2023), we chose to drop NE.

DeVellis (2016) suggests that latent variables and indicators should share a moderate relationship (r² > .30, see Appendix A), we used this criterion for retaining indicators in both sampled groups. In line with measures reported above, five indicators with low loadings were iteratively removed and three indicators were removed as modification indices demonstrated cross-loading on multiple latent constructs. Error covariances were allowed between measurement items sharing similar phrasing or meaning within any given latent construct, but not between constructs. The final model met some, but not all, of Hu and Bentler’s (1999) criteria: RMSEA = .051, CFI = .904, and SRMR = .064. Because this combined sample exceeds 1,000 participants and because CFI is definitionally contingent upon unique variances and is harmed in models with high levels of shared variance among latent constructs (Moshagen & Auerswald, 2018), we consider the fit acceptable. Given that normative evaluations and distract/pastime dimension of functional evaluations were excluded and that subdimensions of metacommunication as well as functional evaluations of control and connections were combined, some parts of the initial hypotheses were not testable. In the following, we only report results of those hypotheses that were testable after these modifications.

MGCFA facilitates measurement invariance testing to assess the psychometric equivalence of the models across the US and Chinese samples (Putnick & Bornstein, 2016). The model proceeded through three steps. First, establishing a baseline configural model which assesses model fit given multiple groups. This model assesses if the latent constructs have “the same meaning across groups” (Kühne, 2013, p. 155). The second model tests for metric invariance. Metric invariance implies “each item contributes to the latent construct to a similar degree across groups” (Putnick & Borenstein, 2016, p. 75). To test this assumption, we constrained the loadings across groups and computed a nested version of the configural model. The third, more stringent model tests for scalar invariance. Scalar invariance is constrained to both loadings and intercepts, which is also called strong factorial invariance (Kühne, 2013). When present, scalar invariance “generally supports cross-group comparisons of manifest (or latent) variable means on the latent variable of interest” (Rutkowski & Svetina, 2014, p. 35). Fit statistics across each invariance test are presented in Table 1.

Table 1.

MGCFA Testing Model Invariance

Given the complexity of the models, it was unlikely that changes in χ² between the nested models which follow would be non-significant. This is because “χ² is overly sensitive to small, unimportant deviations from a ‘perfect’ model in large samples” (Putnick & Bornstein, 2016, p. 78). Thus, we used the alternative fit indices (i.e., ΔRMSEA, ΔCFI, and ΔSRMR) to assess fit. Specifically, we followed the guidance suggested by Rutkowski and Svetina (2014), ΔRMSEA < .03, ΔCFI < -.02, and ΔSRMR ought not exceed .03 for metric invariance (Putnick & Bornstein, 2016). For scalar invariance, these criteria are more stringent: ΔRMSEA < .01, ΔCFI < -.01, and ΔSRMR < .015 (Putnick & Bornstein, 2016). The models met assumptions for configural and metric invariance, signaling the latent constructs are psychometrically similar and comparable across cultural contexts. But the models did not quite meet the criteria for scalar invariance, signaling that it was inappropriate to compare means between the two countries. The baseline configural model and constrained comparison model statistics are presented in Table 1.

Analysis Plan

With metric invariance established, we utilized multi-group SEM (MGSEM) for hypothesis testing, we used maximum likelihood estimation. Though the results are similar to Lee and Cionea’s (2023), our results differ slightly because we use the multigroup approach to revise the model, including scale item inclusion. MGSEM yields separate χ² values but a combined set of fit indices. For the Chinese data, χ² = 4713.83, df = 1828. For the US data, χ² = 4824.48, df = 1828. The model fit reasonably well, given the complexity: RMSEA = .06, CFI = .88, and SRMR = .09. Figure 2 presents the results from the US and Chinese sample.

Figure 2.

Paths Coefficients and Variances Explained in the Chinese (left) and US SampleNote. Dashed paths are significant in the opposite direction of the hypotheses. *p < .05; **p < .01

Hypotheses Testing

Measurement of the MPA across Cultures

The CFA results revealed some problems of this 89-item MPA scale. First, the high correlations in the study suggested that the factor structures of metacommunication and symbolic evaluations should be reconsidered. Corresponding with Lee and Cionea (2023), the present study found three subconstructs of metacommunication (i.e., interpersonal, mass-mediated, and observational) had poor discriminant validity and issues with multicollinearity. Different aspects of metacommunication might have been measured without enough distinction, or the participants might have processed metacommunication holistically.

As media technology rapidly advances, it is no longer easy to make clear distinctions between interpersonal and mass communication via any technology, as described by O’Sullivan and Carr’s (2018) concept of masspersonal communication. The participants might not have precisely identified information sources (e.g., a person, or a person shown in the mass media) when asked to recall metacommunication related to mobile phones.

Additionally, the present study also found empirical overlaps between functional evaluations for control and connectivity. As argued by Lee and Cionea (2023), the dimension of control evaluations seemed to include two types of control: accessibility (e.g., “It is important for me to be available 24/7”) and environmental surveillance (e.g., “It is important for me to be in control of my surroundings”). The former emphasizes control exerted on self-behavior, which also overlaps with the relationship maintenance need, while the latter focuses on control exerted on the external environments, which may have negatively impacted the internal consistency of this subscale. Also, the connectivity dimension appears to overlap with control by asking participants to indicate how important it was for them to always know what was going on with significant others (environmental surveillance) and to always stay in touch with friends and family (accessibility), which can explain why the two factors were highly correlated in both samples. Moreover, the evaluation subscales differed from the corresponding usage subscales for connectivity and control in the way that control usage subscale only focuses on accessibility, and the relational usage subscale only focuses on relationship maintenance. Thus, the conceptualization and operationalization of control and connectivity evaluations should be reconsidered in their future measurement and to align better with the ones for usage.

Noticeably, two dimensions were dropped from the scale: functional beliefs on distraction/pastime and normative evaluations. The distraction subscale was excluded due to its high correlation with the organization of daily lives subscale, indicating the dimensions highly overlapped for the participants. Two items from the distraction/ pastime subscales were closely related with management of daily schedules and chronemic expectations, such as “It is important to me that I do not waste my time with anything during the day,” which offers an explanation why their factor loadings were poor. The subscale for normative evaluations was excluded partially due to the lack of internal consistency. This subscale contained both positively and negatively worded statements (see Appendix B), encompassing both communication norms (when and where to use mobile phones) and social implications of usage (how other people think of certain use behavior), which makes the subscale problematic. Herein, we advocate for further revisions of these two subscales.

Model Structure

Given that only H1 was fully supported, with other hypotheses being partially supported, the findings revealed some practice of mobile phone appropriation in the Chinese sample that deviated from the predictions made by the MPA model. The MGCFA and MGSEM results further unmasked some interesting cultural differences. In this section, we discuss the missing paths after comparing two samples, highlight those paths with reversed relationships, and elucidate the observed differences in path magnitude.

First, the analysis revealed seven significant paths in the Chinese sample that the US sample did not contain: from functional evaluations of connectivity and control to (a) functional usage for relationship maintenance and (b) social usage; from psychological evaluations to (a) usage for control and (b) psychological usage; and from restrictive evaluations to (a) usage for schedule management, (b) usage for relationship maintenance, and (c) psychological usage. The results potentially indicated that for Chinese people, evaluations for relational usage penetrate more aspects of mobile phone use. Possibly, relationship maintenance behavior is incorporated more into people’s everyday agenda and is motivated more by their needs for environmental surveillance and accessibility in the Chinese context, given the emphasis on social connections by collectivistic cultures. The perceived usefulness of mobile phones as a tool for relationship maintenance reinforces its symbolic usage for socializing, which means Chinese people convey social messages by mobile phone usage not only to the social environment in general but also to their significant others. Such messages can encompass both identity display and relational messages (e.g., importance, closeness, intimacy).

That Chinese people’s collectivism and interdependence emphasizes the social environment more could be a reason behind the association between psychological importance and usage for control. Indeed, in past work the fear of missing out was greater among Asians than Americans (Karimkhan, & Chapa, 2021), so the importance of mobile phones might have been boosted by their stronger desires to stay in the loop through mobile phone use. The absence of the predicted association between psychological evaluations and psychological usage in the US sample could be explained by habitual attachment: living in a more tech-savvy environment than Chinese participants, American participants might have been psychologically attached to their mobile phones regardless of how they evaluated such attachment.

It was particularly interesting that restrictive evaluations positively predicted usage for relationship maintenance and daily schedule management in the Chinese context, which contradicted the model. This implies the more costly Chinese participants perceived mobile phones usage to be, the more they used it for relationship maintenance and daily schedule management, which distinguished them from American participants. The logic behind this finding might be that the more financial, technical, temporal, or cognitive difficulties Chinese people perceived, the more they perceived mobile phones usage to be a form of privilege or luxury. As reflected by Aricat et al. (2015), even though mobile phones and mobile services were considered expensive, many migrant workers still chose to use them because such use could not only display their financial resources but also boost their social status as they were treated as technical experts with respect in the community. In the same sense, Chinese participants who perceived more restrictions and barriers might have paradoxically increased their usage as a means to gain both social respect and relational communication with their significant others. Therefore, it is important to consider the inverse relationship identified between affordability and functional usage when examining mobile phone appropriation under the Chinese context.

The US sample also contained some paths that were missing from the Chinese sample: (a) from social evaluations to functional usage for relationship maintenance and (b) from restrictive evaluations to usage for control. The reason why this proposed relationship between social evaluations and relational usage was not observed in the Chinese sample was potentially due to social evaluations being constructed with individualistic values. The items from the subscale emphasized uniqueness and independence, which deviates from the collectivistic self-construal of harmony and interdependence (Markus & Kitayama, 1991). To pursue cross-cultural applicability, the may need to be revised to reflect more universal values accompanying mobile phone appropriation. Moreover, the absent path from restrictive evaluations to usage for control in the Chinese sample may be explained by the fact that using mobile phones for control is less of a social display in general. Because staying reachable (accessibility) is less socially observable as a constant state than calling friends (relationship maintenance) as an episodic activity, the paradox between affordability and usage may be less salient in mobile usage for control.

Besides, the analysis revealed several negative relationships supported by the Chinese sample, which were positive in the US sample: between the functional evaluations of connection and control and functional usage for distraction/pastime, between social evaluations and usage for control, as well as between psychological evaluations and (a) usage for daily schedule management and (b) social symbolic usage. The negative association between the functional evaluations of connection/control and distraction/pastime usage in the Chinese sample may reflect the particularly negative connotations for distracted usage under the cultural context. Chinese users who valued functions of relational maintenance and control might have considered themselves to be serious users, which made them despise usage for distraction and entertainment and perceive such uses to be frivolous. Therefore, the more importance they attached to the pragmatic value of mobile phones, the less they used them to kill time. That the Chinese data revealed a negative association between social evaluations and usage for control may be explained by the similar division that Chinese people draw between serious users and non-serious users and the notion that control usage is less observable in the public sphere. The less they considered mobile phones to be objects for social display, the more they pursued pragmatic uses of mobile phones for accessibility, and vice versa.

By contrast, there were two paths with positive valence in the Chinese sample but negative valence in the US sample: (a) the functional evaluations on organization of daily lives and social aspect of usage, and (b) psychological evaluations to usage for distraction/pastime. The first difference indicates the more US participants valued mobile phones for facilitating management of daily activities, the less they used them for social display. One reason for this phenomenon could arguably be that Americans draw more clear boundaries between the professional and the personal when it comes to technology usage than Chinese people (Caporael & Xie, 2003). Because of this, the more US participants considered mobile phones to be a tool for (work) schedule management, the less they were willing to use it symbolically for identity or status display.

As for the negative path from psychological evaluations to distraction/pastime, one candidate explanation is that distraction/pastime usage is socially stigmatized so there were more errors in self-reporting caused by social desirability bias and cognitive dissonance. American participants might not have been very honest about how often they used mobile phones for distraction/ pastime because of their perception of those negative images associated with it. Another possibility is that usage for distraction/pastime has become so common that people habitually kill time with mobile phones usage, no matter what they think of psychological attachment (similar to how people keep smoking or driving inefficient vehicles, despite knowing the risks or harms). In this case, they may play with mobile phones due to habits and customs even if they deem psychological attachment to be undesirable. Under either condition, these dynamics require further exploration.

Last, there were four paths with significantly different magnitude across the two samples. The path from metacommunication to the psychological evaluations was stronger for American participants than for the Chinese, indicating that Americans may communicate more about mobile communication interpersonally, through mass media, or by observation of others, and this active metacommunication aligns with their psychological evaluations more. It was also observed that metacommunication better predicted restrictive evaluations in the US sample, and our guess is that Chinese people prefer not to talk about restrictions as much interpersonally and that it is not discussed as often by mass media, given the social implications attached to it. Acknowledging ignorance or inability may be more embarrassing and face-threatening for Chinese users. Another guess is that because Chinese people recognize the symbolic values of mobile phones, they tend to disregard metacommunication concerning use restrictions and believe the claims to be less relevant relying more on their own experience to form their perspectives.

Next, the path from psychological evaluations to relational usage was found to be stronger in the Chinese participants, potentially because their psychological attachment to social networks (Shi & Wang, 2011) was a bigger part of their attachment to mobile phones, under the collectivistic appropriation of mobile phones (Arpaci, 2019; Holmes et al., 2015; Yoon, 2003).

Limitations and Future Directions

One limitation of the present study is that the data was collected in 2016, and mobile technology has greatly advanced since then. Thus, future studies may scrutinize how mobile phone appropriation has evolved over the years including its usage for health aspects (e.g., walking, heart rate, connection with smart watches) and short form videos consumption (e.g., TikTok and Instagram Reels). Another limitation is that two dimensions (i.e., functional evaluations on distraction/ pastime and normative evaluations) were dropped from the model testing, which inhibited some of the initial hypotheses and comparisons. Furthermore, the causality of the hypothesized relationships could not be determined with our cross-sectional data. Additionally, self-report data might have increased measurement errors especially for dimensions of usage behavior. Future researchers should improve ways to capture mobile users’ actual behavior, not based on their recall or perception. The data will be more accurate and reliable if collected directly from tracking mobile devices instead of self-reporting (Kobayashi & Boase, 2012).