The mobile Internet underclass: Reality or hyperbole?

Abstract

Scholars have warned that because mobile Internet access offers lower levels of functionality and content availability, operates on less open and flexible platforms and contributes to diminished levels of user engagement, content creation and information seeking compared with traditional fixed solutions, mobile Internet has created a mobile Internet ‘underclass’. As a growing proportion of the online population is now ‘mobile only’, it is imperative to understand the extent to which mobile Internet may represent a meaningful substitute for traditional PC-based Internet access. Based on operational data from a major Chinese telecommunications carrier, this study provides a comparative analysis of the mobile Internet usage patterns associated with different access channels. Our results show that, in terms of mobile Internet usage, there is not enough evidence to support the ‘mobile underclass’ argument. However, our analysis reveals that the primary concern about the mobile ‘underclass’ remains as the first-order digital divide ‘access’ issue.

Keywords

Digital divide Internet usage mobile Internet usage

1. Introduction

Studies have shown that mobile technology has the potential to bridge the digital divide. Mobile phone and mobile Internet are found to be determined less by demographics, socio-economic status and technological readiness, suggesting that barriers to Internet access are lower for mobile users than is the case for computer users and that mobile Internet could be a potential way to bridge the digital divide in areas that lack fixed line facilities [1 –3]. However, scholars have also warned that because mobile Internet access offers lower levels of functionality and content availability, operates on less open and flexible platforms and contributes to diminished levels of user engagement, content creation and information seeking compared with traditional fixed solutions, mobile Internet has created a mobile Internet ‘underclass’ [4]. That is, mobile Internet access represents a significantly inferior form of Internet access compared with traditional PC-based access. Because both technological and skill disparities may diminish gradually over time, the usage patterns associated with mobile Internet seem to pose a more troublesome problem. Particularly, when a growing proportion of the online population is now ‘mobile only’, it is imperative to understand the extent to which mobile Internet may represent a meaningful substitute for traditional PC-based Internet access.

This study utilises objective ‘system-generated’ data from a major telecommunications service company in China. The sample was drawn from the company’s mobile users in a mid-sized city in southwest China. The data include individual-level information on mobile Internet usage activities. Our primary objective is to explore whether mobile-only users’ mobile Internet usage patterns are inferior to those of people with multiple channels of access. We contribute to the ongoing debate about the ‘mobile underclass’ from the perspective of a developing country.

2. Literature review

A recent meta-analysis of existing empirical research on mobile Internet usage intensity levels and potential determinants of respective usage behaviour at the individual subscriber level identified 175 scholarly publications between 2001 and 2012. Five antecedents – enjoyment, behavioural intention to use mobile Internet, educational level, subscription at a flat rate and ease of use – had the largest sample size– and measurement error–corrected average correlations with mobile Internet usage criteria [5]. However, so far, there has been limited comparative analysis of PC and mobile Internet usage. Particularly, none of the 175 publications reviewed by Gerpott and Thomas [5] specifically differentiated mobile-only users from general Internet users.

One of a few sporadic studies – a 2008 study based on 443 mobile Internet users in Germany – found that mobile-only customers who lived in homes without a broadband landline to the Internet generated significantly more mobile IP traffic than those who had fixed-broadband Internet home access; there seemed to be a customer segment whose members use their mobile Internet access as a substitute for, rather than as a complement to, fixed broadband access [6]. However, a recent study based on a survey of Swedish mobile Internet users found that those mobile users who have fixed broadband at home are more likely to use the video, music and social networking applications, suggesting that the use of some applications through mobile broadband was complementary to having fixed broadband at home [7]. In Korea, females, members of the older generation, those with less income and less education and those with poor social capital are found to use mobile devices less frequently and less effectively. Particularly, there is a noticeable divide between low and high socio-economic (defined by education and income) groups in terms of innovativeness and competence, which are required to use mobile media [8]. Similarly, in Japan, it is found that people engage in a wider variety of activities via their PC computers than their mobile phones. Interestingly, the level of education and income are not influential on the use of smartphone applications (APP) in terms of the breadth (total number of applications downloaded) and depth (the level of application usage by type) in Korea [9]. Based on semi-structured interviews with 21 American and German mobile Internet users, it was found that PC-based Internet usage is primarily immersive (more focused, outcome-oriented usage) and mobile Internet usage is primarily extractive (process-oriented and sometimes serendipitous usage); thus, having only a mobile phone at hand could foster extractive use and make immersive Internet use less likely [10]. Based on a 2011 survey in Chicago, it was found that mobile-only users are more likely to be low-income, less educated and Black, and more importantly, the growth in mobile phone use has not erased inequalities in online participation. Having broadband at home seems to still be critically important for digital citizenship. Mobile-only users compare unfavourably with home broadband users in terms of performing many political and economic activities online and for all indicators of skill; they are also more likely than home broadband adopters to use the Internet for entertainment [11].

Even fewer studies have been conducted in developing countries. The results of a study conducted in Armenia show that while mobile Internet has the potential to be an alternative route to Internet resources and reduces some gaps over time, mobile-only users are, in general, less likely than PC-based or dual users to use the Internet as frequently or as long, to engage in ‘capital-enhancing’ Internet activities or to engage in as much activity breadth, creating a device divide and decreasing the potential benefits of the Internet [12]. In Indonesia, a study shows that Internet usage is more prevalent among younger and more educated people on both smart and feature phones, and people experienced in PC-based Internet are more likely to own smartphones, the use of which is associated with higher information acquisition in daily life [13]. A survey of 500 students from very low-income areas in Cape Town, South Africa revealed that mobile phones are usually associated with casual use and leisure, whereas computers are used for ‘serious’ work, such as conducting research for school, looking for medical information or accessing the school and job market [14]. In an ethnographic study also conducted in Cape Town, South Africa, eight low-income women who had never before used a personal computer were trained to access the Internet on mobile handsets. They had encountered various barriers 6 months after training, including affordability, limited functionality, the dearth of mobile-friendly online content and difficulty of use [15,16].

Overall, PC-based Internet usage studies have indicated that differentiated usage of the Internet increasingly reflects the relationship between the use of traditional media and known offline economic, social and cultural attributes, including inequalities, and it also contributes to reproducing such inequalities [17 –19]. While research on the mobile Internet is far from conclusive, it generally leaves the impression that skill and usage gaps also exist, and may have worsened, among mobile-only users.

3. Analytical framework

In the previous digital inequality literature, there is little theoretical discussion of the appropriate dimensions of Internet use. As a dependent variable, Internet usage in general has been measured in different units or dimensions, including frequency, length of time spent online, data volume, types of online activities and, in rare cases, revenues [5]. Blank and Groselj [20] reviewed previous studies that created a typology of Internet users or uses and found a great deal of confusion about the dimensions of Internet use; the existing typologies were inconsistent, rigid and primarily data-driven, with few theoretical guidelines. They argued that Internet use should be measured in a three-dimensional property space, based on the amount of use, variety of different uses and types of use; these theoretical dimensions of Internet use should not be collapsed into a single empirical dimension with multiple categories [20]. We ground our empirical study in this analytical framework and adopt it in exploring mobile Internet usage behaviours associated with different access channels.

3.1. Amount of use

The most widely used measurement of the first dimension of Internet use, amount of use, refers to the frequency of Internet use in day-to-day life (also called ‘frequency of use’ in many other studies). Most, if not all, Internet inequality studies rely on one-shot surveys or, infrequently, longitudinal panels to measure this variable [17 –19,21 –26]. For instance, Oxford Internet Surveys (OxIS) asked respondents how often they do each one of 48 Internet activities on a 6-point Likert scale, where ‘never’ = 0 and ‘more than once per day’ = 5; the sum of those scales generates a continuous variable representing the total amount of Internet use, ranging from 0 to 240 [20]. In fact, in a meta-analysis of the general mobile Internet usage literature, it is found that nearly 60% of the 175 contributions have used self-reported surveys to collect data [5].

There are some noticeable methodological limitations to using a subjective method to collect usage data. Respondents tend to overestimate their usage; they may fail to record or remember the frequency, duration or variety of their usage accurately; and they may be unaware of the ‘idle traffic’ automatically generated by networks, devices or applications, which are not under (direct) consumer control [5]. Alternative objective methods, such as handset monitoring, traffic measurement, usage billing and log collection, usually require direct access to service providers’ core systems [27]. Particularly, network service providers’ billing systems could provide accurate Internet data volume information, which could be used as a usage measure. However, due to the understandable reluctance of the network service providers to permit access to this sensitive information, only a small fraction of existing studies, 15 out of 175 studies reviewed, based their usage measurements on billing system records [5].

For the purpose of this study, an objective ‘system-generated’ mobile Internet use intensity measure was extracted from the collaborating carrier’s billing engine. In addition to this measure of data volume, we also use two other measures to represent the concept of amount of use. We use a measure of the amount of App usage to indicate a user’s App usage frequency on a mobile phone. Considering the difference between App usage and website visits, we also use a measure of the number of website visits to indicate a user’s website visit frequency on a mobile phone. While these measures may be correlated, each of them captures a different aspect of a user’s mobile Internet usage behaviour.

3.2. Variety of use

The concept of variety of use represents a user’s variety-seeking behaviour when surfing the Internet via a mobile phone. Variety is logically distinct from amount of use and is a separate property of Internet use because users with the same amount of use may have different numbers of Internet uses. For example, someone may use a video App 10 times during the day, while during the same amount of time another user may use 10 different Apps one time each. For the metric of amount of App usage, the amount of use is the same for these two users, but the variety of use is different (1 versus 10).

3.3. Type of use

Exploring differences in mobile Internet usage requires a classification of common contemporary Internet activities. Moving beyond the details of specific, individual activities requires a relatively small and manageable set of internally consistent types of Internet activities [19]. Following the suggestions by Blank and Groselj [20] that the activity type needs to be defined on a conceptual level, a coding process was conducted in which the individual Apps and websites were screened and categorised by two researchers for the purpose of classification. If coding diverged, it was discussed until consensus was reached to minimise the subjectivity of coding.

4. Research questions

Efforts have been made to understand the usage patterns associated with the mobile Internet. As reviewed in the previous section, existing studies generally suggest that skill and usage gaps also exist, and may have worsened, in mobile Internet usage – particularly for mobile-only users. However, objective empirical evidence in support of this claim is sparse. It is therefore worth posing the following research question: In what way does mobile Internet usage differ depending on whether users are mobile-only or mobile-fixed-bundled subscribers?

In addition, very few studies to date have explored the distribution of use behaviours of mobile Internet users. Previous research has been inconclusive on the characteristics of mobile Internet usage distribution. Some have argued that it is normally distributed within the mobile Internet adopter population. Others have found that the skewness statistics for the distribution of mobile usage were strongly positive, that is, few adopters caused a disproportionately large share of traffic [5,28]. Indeed, DiMaggio et al. [29] argued that the level of socio-economic inequality in access to information online was greater than in other traditional media outlets and that persons of higher socio-economic status employ the Internet more productively (e.g. seeking financial information, learning about public issues and obtaining work assistance) than their less privileged peers. Studies based on self-reported data have also found that different social positions (occupation, village membership (insider or outsider of a village) and social status) lead to differences in acquisition, usage, demand and conception of information and communications technology (ICT) products and services. In this view, advantaged social groups enjoy more services and information while marginalised social groups find it harder to obtain useful information, and thus, information technologies fail to bridge existing divides and facilitate social integration [30]. While such claims do not necessarily lead to the conclusion that a certain group of users engages in more mobile Internet activities than others, it does raise the interesting question cited above, on the exploration of the actual distribution of mobile usage. Consequently, we also explore the following research question: In what way does the skewness of the distribution of mobile Internet usage differ depending on whether users are mobile-only or mobile-fixed-bundled subscribers?

5. Methodology

5.1. Data and sample

Our research is based upon a unique set of data of mobile Internet users’ behaviours. Our sample was drawn from the population of mobile users in a typical mid-sized city in southwest China who subscribe to a major telecommunications service company. At the time of the study, the city had a gross domestic product (GDP) per capita of about USD7000, comparable to the national average of USD6978. Unlike previous studies that rely on survey or interview data, we empirically observed and extracted records of mobile Internet users’ behaviours from their Internet service provider (carrier). Aside from its billing system, the carrier keeps records of the subscribers’ online behaviours. These records include detailed information on daily App usage and website visits using the mobile platform. While we can identify specific Apps and the frequency with which they are used, we are limited to aggregated information on website visits because the carrier has classified the web URLs into several categories due to the enormous number of URLs. By combining this behavioural data with other user information, such as gender, age, access channel, handset and data and service usage, we obtained a complete profile of each mobile user.

The market structure makes it possible to identify mobile-only users. There are three telecommunications carriers in the city under study. While all the three carriers provide mobile service, the carrier for this study is the only residential fixed-broadband service provider. As a result, if a customer subscribes only to the mobile service of the carrier, we can reasonably assume that he or she does not have fixed broadband at his/her residence and thus can be classified as a mobile-only user.

We collected data on users’ mobile Internet usage behaviour during September 2013. The data include individual-level information on mobile Internet usage activities. The carrier keeps records of users’ mobile Internet behaviours, including their usage of Apps and their Web browsing visits. All the Apps and websites that are used or visited by a mobile user are recorded. A total of 105 Apps were used by mobile users during the period of the study. Websites are classified into 18 categories by the carrier, namely, sports, entertainment, travel, video services, gaming, finance, shopping, education, reading, news, media, social interaction, everyday life, IT information, car information, cell phone information, digital products and carrier specific. It should be noted that the carrier’s data on mobile Internet behaviours are reported on a daily basis. However, as the temporal dynamics of mobile Internet usage are not the concern of this study, we construct mobile usage measures by summing the data of the corresponding measurement for the 30 days. We, then, combine this mobile Internet usage data with Internet access channel information to identify whether a user has mobile-fixed-bundled Internet access or mobile-only Internet access.

In our data, 44.3% of mobile users with mobile-fixed-bundled access generate at least one record of App or website usage, while only 26.6% of mobile-only users generate at least one record of App or website usage. A logit model is estimated to model the dichotomous outcome variable of whether a mobile user engages in mobile Internet activities, using predictors of Internet access channel, gender, age, number of voice calls and short message services (SMSs), monthly bill and past mobile Internet experience. The analysis confirms that the Internet access channel has a significant negative impact on the odds of mobile Internet activities (as we code access channel 1 for mobile only and 0 for mobile-fixed-bundled). In addition, the data also show that mobile-only users on average make fewer calls, send fewer short messages and are more likely to use cheap and feature phone than are the mobile-fixed-bundled customers. Nevertheless, as we are interested in mobile Internet usage activities, those users with no App usage or website visits during the period of the study are excluded from analysis. The final dataset includes data records from 210,208 mobile users.

The privacy of participants is guaranteed by conforming to Chinese regulations, and data were processed after anonymisation. The tasks of data collection and data analysis were separated so that no personal information could be used for identification.

5.2. Measurement

In line with the conceptual framework of this study, the concept of ‘mobile Internet usage behaviour’ is threefold: it comprises amount of use, varieties of use and types of use. Measuring the amount of use includes three metrics: first, the data volume indicates the total amount of mobile Internet data usage; second, the amount of App usage represents a user’s APP usage frequency on a mobile phone; and third, the number of website visits represents the user’s Internet browsing frequency when using a browser on a mobile phone.

For the measurement of variety of use, two metrics are proposed, namely, the variety of App usage, representing the total number of different Apps that are used by a user at least once in a month, and the variety of website visits, representing the total number of different categories of websites that are visited by a user at least once in a month.

For the measurement of type of use, three conceptual types of use are defined for both App usage and website visits. The first type is ‘knowledge-intensive’ activities, including finance, commercial transactions, information, news, education, reading and so on. The second type is ‘leisure-related activities’, including games, sports, entertainment, video, social interaction and so on. The first type offers users more opportunities and resources for moving forward in their careers, work, education and societal positions than does the second type, which are mainly consumptive or entertaining. Other activities are classified as a third type, ‘other’. These three types are mutually exclusive, although a user may engage in mobile Internet activities of more than one type.

Based on this classification scheme, shares of the different types of App usage and shares of the different types of website visits are computed. For example, a user’s share of knowledge-intensive App usage is defined as the percentage of Apps the user uses that are classified as knowledge-intensive, with logical boundaries of 0–1. The definitions for the measurement of mobile Internet use are listed in Table 1.

Table 1.

Measurement of mobile Internet use

Level of analysis	Sub-level of analysis	Definition
Amount of use	Data volume	Volume of uploaded and downloaded IP-switched data traffic via an MNO’s network
	App	Frequency of using an App on a mobile phone
	Website	Frequency of visiting a URL when using a browser on a mobile phone
Variety of use	App	Total number of Apps that are used by a user at least once in a month
Variety of use	Website	Total number of different categories of websites that are visited by a user at least once in a month
Type of use	App	Share of different types of App usage
Type of use	Website	Share of different types of website visits

MNO: mobile network operator.

In addition to these focal variables, the two most frequently analysed general demographic variables in the mobile Internet usage literature, gender and age, are also measured. For example, while the current cumulative evidence does not allow us to conclude whether the typical mobile Internet usage level of males exceeds (or falls below) that of females, males were found to be more likely to spend their time online engaging in more task-focused activities (e.g. reading the news, getting financial information) and for entertainment and leisure, whereas women use their time online primarily for interpersonal communication [31,32]. In terms of age, while the usage differences among different generations are gradually diminishing, older Internet users tend to have distinct age-specific needs, preferences and predispositions, such as maintaining family and social connections, accessing information on health and performing routine activities [33 –36].

To better understand the factors that may correlate with mobile Internet usage behaviour, we also collected data on the usage of traditional mobile services (such as voice calls and SMS). In the literature to date, it is generally agreed that two established mobile service categories – voice calls and SMS – are positively correlated with mobile Internet usage. However, whether this correlation still holds true is subject to debate, and it should be tested given the increasing proliferation of mobile Internet–based voice and messaging apps on smartphones [5]. While income has an effect on Internet access as well as on types of online activities [20], we do not have income information on the individual users in our sample. However, we do consider their bill charges, which are supposed to correlate with income. Another factor accounted for is mobile experience. We use the length of time since a user initially subscribed to the service provider’s mobile Internet service to measure his/her experience with mobile Internet usage.

6. Results

6.1. Descriptive statistics

The summary statistics of the key variables are provided in Table 2, which shows that mobile-only users account for 23% of our sample. Male users account for a proportionally larger share than female users (66% versus 34%). The mean age of our sample users is approximately 37, with a standard deviation of 11.38. An average user makes approximately 565 calls and sends approximately 41 SMSs a month. The average bill charge per user is 51.14 Yuan. The average user had been using the mobile Internet for 23.4 months before September 2013.

Table 2.

Summary statistics

	Variable	Mean	SD
Mobile Internet usage	Amount of use
	Data volume (MBs)	179.56	403.78
	Frequency of App usage	3406.74	5230.64
	Frequency of website visits	2261.45	5028.65
	Variety of use
	Variety of App usage	14.35	8.43
	Variety of website visits	7.18	3.58
	Type of use
	Share of knowledge-intensive App usage	0.27	0.25
	Share of leisure-related App usage	0.53	0.30
	Share of knowledge-intensive website visits	0.51	0.32
	Share of leisure-related website visits	0.28	0.29
Access channel	Internet access (1 = mobile Internet only, 0 = mobile-fixed-bundled Internet)	0.23	0.42
Demographics	Gender (1 = male, 0 = female)	0.66	0.47
Demographics	Age	37.39	11.38
Other variables	Number of calls made	565.02	565.40
	Number of SMS sent	41.08	87.19
	Charge (RMB Yuan)	51.14	68.62
	Mobile Internet experience (months)	23.40	15.86

SD: standard deviation; SMS: short message service.

As for mobile Internet usage, an average user generates 179.56 MB of traffic volume using mobile Internet. During a 1-month period, the users in our sample used an average of 14.35 Apps, accumulating 3406.74 instances of App usage, and visited an average of 7.18 categories of websites, accumulating 2261.45 website visits.

By comparing the usage distribution of different categories between App usage and website visits, an interesting pattern emerges: for App usage, the share of leisure-related activities (53%) is higher than that of knowledge-intensive activities (27%); for website visits, a reverse pattern exists (28% versus 51%). It may be inferred that generally speaking, mobile Internet users are more likely to engage in leisure-related activities when using Apps, and they are more likely to engage in knowledge-intensive activities when browsing the Internet via a browser. This finding adds to the understanding of how people spend their time online with regard to different tasks.

As shown in Table 3, the correlations between the three metrics of amount of use (data volume, amount of App usage and number of website visits) are all below 0.50 (0.30, 0.34 and 0.45, respectively). This suggests that none of the three metrics alone can account for the total amount of use, and each of them may represent a different aspect of a user’s amount of mobile Internet usage. It is shown that while the correlation between the two metrics of variety of use (variety of App usage and variety of website visits) is as high as 0.81, the correlation between amount of use and variety of use is relatively small, thus providing empirical support for the theoretical dimensions of Internet use.

Table 3.

Correlations between variables (amount of use and variety of use)

	Data volume	Frequency of App usage	Frequency of website visits	Variety of App usage	Variety of website visits
Data volume	1.00
Frequency of App usage	0.30	1.00
Frequency of website visits	0.34	0.45	1.00
Variety of App usage	0.43	0.54	0.43	1.00
Variety of website visits	0.37	0.44	0.38	0.81	1.00

6.2. Mobile-fixed-bundled versus mobile-only

Tables 4 and 5 provide our findings on the comparative behaviour of mobile Internet use for users with different forms of Internet access. Concerning the amount of use, mobile-only users outperform mobile-fixed-bundled users in each of the variables of amount of use. Mobile-only users generate 229.17 MB of data traffic, compared with the 165 MB of data traffic generated by mobile-fixed-bundled users. In terms of App usage, mobile-only users use Apps more frequently than mobile-fixed-bundled users (3739 times versus 3309 times). Meanwhile, when surfing the Web, mobile-only users visit more web pages than mobile-fixed-bundled users (2533 times versus 2182 times).

Table 4.

Amount of use by access channel

Statistics	Mobile-only			Mobile-fixed-bundled
Statistics	Data volume	Frequency of App usage	Frequency of website visits	Data volume	Frequency of App usage	Frequency of website visits
Mean	229.17	3739	2533.15	165	3309.19	2181.69
Skewness	16.47	4.28	9.17	20.27	2.74	12.45
CV	2.29	1.53	2.1	2.18	1.53	2.26
Percentile 0–25	0.16	0.43	0.41	0.32	0.49	0.49
Percentile 25–50	3.88	4.44	4.15	4.30	4.28	3.88
Percentile 50–75	17.32	18.63	16.03	17.27	17.82	14.76
Percentile 75–100	78.64	76.50	79.40	78.11	77.40	80.86

CV: coefficient of variation.

Table 5.

Variety of use and type of use by access channel

	Mobile-fixed-bundled			Mobile-only
	Mean	CV	Skewness	Mean	CV	Skewness
Variety of use
Variety of App usage	7.13	0.5	0.23	7.38	0.51	0.14
Variety of website visits	14.22	0.58	0.52	14.79	0.61	0.45
Type of use
Share of knowledge-intensive App usage	0.27	0.95	1.19	0.27	0.95	1.19
Share of leisure-related App usage	0.54	0.56	−0.3	0.53	0.58	−0.3
Share of knowledge-intensive website visits	0.51	0.61	−0.03	0.49	0.65	0.03
Share of leisure-related website visits	0.28	1.01	0.91	0.29	1	0.88

CV: coefficient of variation.

The distribution of amount of use was examined by appraising the skewness and variation coefficients of the three amount-of-use metrics in the two groups as defined by the Internet access channel. Furthermore, mobile users were split into four use quartiles. The results in Table 4 show that the skewness statistics of the amount of use measure are strongly positive in both of the customer groups. For data volume and frequency of website visits, the skewness statistic is lower for the sample with mobile-only access than the sample with mobile-fixed-bundled Internet access, while the skewness statistic for frequency of App usage is higher for the sample with mobile-only access than the sample with mobile-fixed-bundled Internet access. This finding provides mixed evidence that may contradict the hypothesis of an earlier study that mobile Internet use intensity is less unevenly spread among consumers accessing the Internet with a handset than among mobile subscribers with a desktop computer [28].

The three metrics of amount of use unanimously indicate that a large share of mobile Internet usage can be attributed to a few heavy users, and the percentage of the top quarter exceeds 75%. There is no substantial difference in percentile distributions between mobile-only users and mobile-fixed-bundled users. Generally speaking, across both customer groups under study, there was a small group of subscribers generating a very large proportion of the sample’s overall mobile Internet usage. In other words, the analysis indicates that the majority of customers used the mobile Internet in a very limited way.

The measure of variety of use indicates that mobile-only users seek a larger variety of mobile Internet usage than mobile-fixed-bundled users. As shown in Table 5, mobile-only users use an average of 7.38 Apps and visit an average of 14.79 categories of websites, whereas mobile-fixed-bundled users use an average of 7.13 Apps and visit an average of 14.22 categories of websites.

For the third dimension, there is hardly a difference between mobile-fixed-bundled and mobile-only users’ behaviour in share of usage across the three App types. A common pattern emerges for both mobile-fixed-bundled users and mobile-only users. For App usage, the share of leisure-related activities outweighs that of knowledge-intensive activities.

There is a minor difference between bundled and mobile-only users’ behaviour in share of usage across different website types. The share of knowledge-intensive browser visits for mobile-fixed-bundled users is slightly higher than that of mobile-only users. On the contrary, mobile-only users invest slightly more time in leisure-related activities when browsing the Internet than do mobile-fixed-bundled users. For both mobile-fixed-bundled users and mobile-only users, knowledge-intensive activities account for a larger share of browsing activities online and a smaller share of their App usage, compared with leisure-related and other activities. As far as the distribution of different types of use is concerned, an interesting finding emerges. For App usage, the skewness statistic of the share of leisure-related activity is much closer to zero compared with that for the share of knowledge-intensive activity; while for website visits, the skewness statistic of the share of knowledge-intensive activity is much closer to zero compared with that of the share of leisure-related activity. These findings suggest that the share of leisure-related App usage is not only higher but also more evenly distributed among the sample than the share of knowledge-intensive App usage, and a reverse pattern exists for website visits.

6.3. Mobile-only breakdown

To better understand the mobile Internet usage behaviour of mobile-only users, two analyses were performed. First, we investigated the factor of gender. Table 6 shows the difference between male and female mobile-only users. On average, male mobile-only users generate more data volume and use Apps more frequently than female mobile-only users, while female mobile-only users tend to visit more web pages when surfing the Web on a mobile phone. In terms of variety of use, male mobile-only users seek more variety in their App usage and less variety in their website visits than female mobile-only users, though the differences are small. When using Apps, the preference for leisure-related Apps over knowledge-intensive Apps is more prominent for female mobile-only users than their male counterparts.

Table 6.

Mobile Internet usage by gender, mobile-only users.

	Male	Female
Amount of use
Data volume	254.01	235.99
Frequency of App usage	4103.9	3948.75
Frequency of website visits	2660.12	2790.21
Variety of use
Variety of App usage	7.68	7.54
Variety of website visits	15.56	15.65
Type of use
Share of knowledge-intensive App usage	0.28	0.25
Share of leisure-related App usage	0.53	0.57
Share of knowledge-intensive website visits	0.49	0.49
Share of leisure-related website visits	0.28	0.32

Moreover, we are interested in how type of use is related to amount of use. The mobile-only users are ordered according to data volume and then divided into four equal-sized groups. Then, the share of knowledge-intensive/leisure-related activities for both App usage and website visits is compared between each of the four quarters. As indicated in Figure 1, compared with the upper quarter (group 4), the lower quarter (group 1) has a higher share of knowledge-intensive App usage and a lower share of leisure-related App usage. The same pattern persists for website visits. In other words, users who generate lower amounts of data usage comparatively engage in more knowledge-intensive and fewer mobile Internet activities, and users with higher amounts of data usage comparatively engage in more leisure-related activities. This provides evidence for the justification of the analytical framework in which mobile Internet usage is a multi-dimensional concept and should be addressed with caution in studies of mobile behaviour.

Figure 1.

Shares of different types of use for four groups defined by amount of use.

7. Conclusion

Our analysis provides a comparative understanding of the mobile Internet usage patterns associated with different access channels. Unlike previous studies that primarily, if not entirely, relied on survey or interview data, our research is based on operational data from a major Chinese telecommunications carrier. Generally speaking, in terms of online usage, our analysis hardly seems to support the ‘mobile underclass’ claim that has been of concern to digital divide scholars since mobile broadband first emerged.

As a matter of fact, mobile-only users outperform mobile-fixed-bundled users in each of the variables of amount of use. Mobile-only users generate more data traffic, use Apps more frequently and visit more web pages than mobile-fixed-bundled users. In addition, mobile-only users exhibit a greater variety of mobile Internet usage than mobile-fixed-bundled users. In terms of the type of App usage, there is hardly any difference between mobile-fixed-bundled and mobile-only users. It is found that the share of knowledge-intensive browser visits for mobile-fixed-bundled users is slightly higher than that of mobile-only users. However, the gap between the two groups is too small to draw any meaningful inferences. Thus, no significant difference is found between mobile-only and mobile-fixed-bundled users in their mobile Internet usage behaviour.

Within the mobile-only user group, there are some noticeable differences between male and female users. On average, male mobile-only users generate more data volume. Males seem to prefer Apps and females are inclined to use browsers to access information, although the differences are small. Particularly, the preference for leisure-related Apps and websites over knowledge-intensive Apps and websites is more prominent for female mobile-only users than their male counterparts. In addition, while a large share of mobile Internet usage was attributed to only a few heavy users, it is interesting that users who generate lower amounts of data usage comparatively engage in more knowledge-intensive and fewer mobile Internet activities, and users with higher amounts of data usage comparatively engage in more leisure-related activities. It appears that users tend to first fulfil their utilitarian needs, and they do so with limited data usage.

Our research also reveals another important, but yet often neglected, issue in previous studies. Surprisingly, 44.3% of mobile users with mobile-fixed-bundled access generate at least one record of App usage or one website visit, while only 26.6% of mobile-only users generate at least one record of App usage or one website visit. That is, nearly three-quarters of mobile-only users are not using mobile Internet. Once these users are online, they demonstrate little difference in their usage pattern compared with mobile-fixed-bundled users. Thus, at least from the perspective of developing countries, the primary concern about the ‘mobile underclass’ remains as the first-order digital divide ‘access’ issue. Making mobile Internet access more affordable is still the primary concern for policy makers and businesses. Traditionally, China’s communications policy has focused on the supply side by investing in infrastructure. Interestingly, the State Council issued a directive policy in 2015 on broadband development, entitled ‘Guiding Opinions of the General Office of the State Council on Accelerating the Construction of High-speed Broadband Networks, Boosting Internet Speed and Lowering Internet Charges’, in which the government specifically requires telecommunications carriers to lower charges on broadband access. Some policy measures included in this guideline are highly detailed; for example, it encourages carriers to offer free mobile data packages during non-peak periods, prohibits the compulsory clearing of unused data and allows the subscriber to transfer unused data to others as a gift. Hopefully, this seemingly user-centred approach would help to lower the barrier to mobile Internet access.

Like most empirical work, this study is not without limitations that call for improvement. First, because we do not have access to fixed-line broadband usage data, while this study adequately meets the primary objective of our research, comparing Internet usage on the mobile Internet network, mobile-fixed-bundled users’ mobile Internet usage might be underestimated because they may intentionally limit their mobile Internet usage due to the comparatively higher cost of using the mobile Internet. Thus, the amount of use is probably underestimated in our study. However, this deficiency does not impact our analysis of the other two dimensions of Internet use, namely, variety of different uses and types of use. Second, we have limited control over the availability and format of the data. For example, we are limited to aggregated information on website visits because the carrier has already classified the visited URLs into several categories. To that end, a subjective data-collection method would be a useful tool to complement this study.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

This research was supported by the National Social Science Fund of China (14BGL117), Science and Technology Project of Chengdu (2015-RK00-00030-ZF), the Science and Technology Project of Sichuan Province (2017ZR0031), the Major Program of the National Natural Science Foundation of China (71490722) and Young Scholars Development Fund of SWPU (201499010068).

References

Akiyoshi

Ono

. The diffusion of mobile Internet in Japan. Inform Soc 2008; 24: 292–303.

Wareham

Levy

Shi

. Wireless diffusion and mobile computing: implications for the digital divide. Telecommun Policy 2004; 28: 439–457.

Srinuan

Bohlin

. An analysis of mobile Internet access in Thailand: implications for bridging the digital divide. Telemat Inform 2012; 29: 254–262.

Napoli

Obar

. The emerging mobile Internet underclass: a critique of mobile Internet access. Inform Soc 2014; 30: 323–334.

Gerpott

Thomas

. Empirical research on mobile Internet usage: a meta-analysis of the literature. Telecommun Policy 2014; 38: 291–310.

Gerpott

. Communication behaviors and perceptions of mobile Internet adopters. Info 2010; 12: 54–73.

Kongaut

Bohlin

. Investigating mobile broadband adoption and usage: a case of smartphones in Sweden. Telemat Inform 2016; 33: 742–752.

Lee

Kim

. Socio-demographic gaps in mobile use, causes, and consequences: a multi-group analysis of the mobile divide model. Inf Commun Soc 2014; 17: 917–936.

Jung

Chan-Olmsted

Kim

. From access to utilization: factors affecting smartphone application use and its impacts on social and human capital acquisition in South Korea. Journalism Mass Comm 2013; 90: 715–735.

10.

Humphreys

Von Pape

Karnowski

. Evolving mobile media: uses and conceptualizations of the mobile Internet. J Comput-Mediat Comm 2013; 18: 491–507.

11.

Mossberger

Tolbert

Hamilton

. Measuring digital citizenship: mobile access and broadband. Int J Commun 2012; 6: 2492–2528.

12.

Pearce

Rice

. Digital divides from access to activities: comparing mobile and personal computer Internet users. J Commun 2013; 63: 721–744.

13.

Puspitasari

Ishii

. Digital divides and mobile Internet in Indonesia: impact of smartphones. Telemat Inform 2016; 33: 472–483.

14.

Kreutzer

. Internet and online media usage on mobile phones among low-income urban youth in Cape Town. Int J Educ Dev ICT 2009; 5: 1–21.

15.

Gitau

Marsden

Donner

. After access: challenges facing mobile-only Internet users in the developing world. In: Proceedings of the SIGCHI conference on human factors in computing systems, Atlanta, GA, 10–15 April 2010, pp. 2603–2606. New York: ACM.

16.

Donner

Gitau

Marsden

. Exploring mobile-only Internet use: results of a training study in urban South Africa. Int J Commun 2011; 5: 574–597.

17.

Witte

Mannon

. The Internet and social inequalities. New York: Routledge, 2010, 180 pp.

18.

Zillien

Hargittai

. Digital distinction: status-specific types of Internet usage. Soc Sci Quart 2009; 90: 274–291.

19.

Van Deursen

Van Dijk

Ten Klooster

. Increasing inequalities in what we do online: a longitudinal cross sectional analysis of Internet activities among the Dutch population (2010 to 2013) over gender, age, education, and income. Telemat Inform 2015; 32: 259–272.

20.

Blank

Groselj

. Dimensions of Internet use: amount, variety, and types. Inf Commun Soc 2014; 17: 417–435.

21.

Goldfarb

Prince

. Internet adoption and usage patterns are different: implications for the digital divide. Inf Econ Policy 2008; 20: 2–15.

22.

Van Deursen

Van Dijk

. The digital divide shifts to differences in usage. New Media Soc 2014; 16: 507–526.

23.

Hargittai

Hinnant

. Digital inequality: differences in young adults’ use of the Internet. Commun Res 2008; 35: 602–621.

24.

Robinson

DiMaggio

Hargittai

. New social survey perspectives on the digital divide. It Soc 2003; 1: 1–22.

25.

Graham

Smith

. Dividing lines. Inf Commun Soc 2010; 13: 892–908.

26.

Gutiérrez

Gamboa

. Determinants of ICT usage among low-income groups in Colombia, Mexico, and Peru. Inf Soc 2010; 26: 346–363.

27.

Kivi

. Measuring mobile service usage: methods and measurement points. Int J Mob Commun 2009; 7: 415–435.

28.

Gerpott

Thomas

Weichert

. Characteristics and mobile Internet use intensity of consumers with different types of advanced handsets: an exploratory empirical study of iPhone, Android and other web-enabled mobile users in Germany. Telecommun Policy 2013; 37: 357–371.

29.

DiMaggio

Hargittai

Celeste

. From unequal access to differentiated use: a literature review and agenda for research on digital inequality. In: Neckerman

(ed.) Social inequality. New York: Russell Sage Foundation, 2004, pp. 355–400.

30.

Guo

Chen

. Digital divide and social cleavage: case studies of ICT usage among peasants in contemporary china. China Quart 2011; 207: 580–599.

31.

Weiser

. Gender differences in Internet use patterns and Internet application preferences: a two-sample comparison. Cyberpsychol Behav 2000; 3: 167–178.

32.

Jackson

Ervin

Gardner

. Gender and the Internet: women communicating and men searching. Sex Roles 2001; 44: 363–379.

33.

Zickuhr

Madden

. Older adults and Internet use. Pew Internet & American Life Project, Pew Research Center, Washington, DC, 6 June 2012.

34.

Vroman

Arthanat

Lysack

. ‘Who over 65 is online?’ Older adults’ dispositions toward information communication technology. Comput Hum Behav 2015; 43: 156–166.

35.

Zickuhr

. Generations 2010. Pew Internet & American Life Project, Pew Research Center, Washington, DC, 16 December 2010.

36.

Jones

. Generations online in 2009. 2009, http://www.pewinternet.org/2009/01/28/generations-online-in-2009/