You are on page 1of 15

Telecommunications Policy 37 (2013) 357371

Contents lists available at SciVerse ScienceDirect

Telecommunications Policy
URL: www.elsevier.com/locate/telpol

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
Torsten J. Gerpott a,n, Sandra Thomas b,1, Michael Weichert c,2
a

Chair of Strategic Management, Focus Telecommunications Management, Mercator School of Management, University of Duisburg-Essen,
Lotharstr. 65, D-47057 Duisburg, Germany
b
BearingPoint GmbH, Communications & Content, Gladbecker Str. 5, D-40472 D
usseldorf, Germany
c
Vodafone D2 GmbH, Consumer Web, Am Seestern 1, D-40547 D
usseldorf, Germany

a r t i c l e i n f o

abstract

Available online 18 June 2012

This work explores personal characteristics and mobile Internet (MI) use behaviors of
consumers equipped with four distinct types of advanced handsets for accessing the
Internet via cellular radio infrastructures of mobile network operators (MNO). Furthermore, it investigates the extent to which personal and mobile appliance characteristics
explain variance in actual MI use intensity. Data on two demographic variables, three
MNO relationship characteristics and actual MI use intensity (average monthly volume
of mobile IP trafc generated by a subscriber in May and June 2011) of 9321 adult
consumers with a at MI pricing scheme are extracted from customer les of the
German subsidiary of a large international MNO. 959, 2213, 2410 and 3739 of the
sample members use an Apple iPhone 3, an Apple iPhone 4, a model running with
Googles Android operating system (OS) and other MI-enabled mobile OS/phone types,
respectively. Compared to the adult population in Germany, persons at least 50 years of
age are clearly underrepresented among MI adopters with the four studied device types.
Differences between the four phone type groups with regard to gender, age, time from
enrollment and MI use experience emerge as statistically signicant, but they achieve
only minor substantial relevance. MI use intensity is highly positively skewed: In each
of the four appliance groups, a small number of users disproportionately add to the total
MI trafc generated by the subjects. Consumers advanced OS/handset type strongly
contributes towards explaining MI use intensity variance. iPhone subscribers generate
more trafc than Android customers who in turn show a higher MI activity level than
individuals running other web-enabled mobile models. Age is the only studied personal
characteristic consistently showing a (negative) association with MI usage, which both
is statistically and materially signicant. Conclusions are drawn for MNO on MI
marketing issues. Implications of study limitations for research on MI adoption and
use behaviors on the MI are also outlined.
& 2012 Elsevier Ltd. All rights reserved.

Keywords:
Advanced mobile handset types
Android
Apple
Consumer characteristics
Mobile Internet
Mobile phone operating system
Tariff type
Use intensity

Corresponding author. Tel.: 49 203/379 3109; fax: 49 203/379 2656.


E-mail addresses: torsten.gerpott@uni-due.de (T.J. Gerpott), sandra.thomas@bearingpoint.com (S. Thomas),
michael.weichert@vodafone.com (M. Weichert).
1
Tel.: 49 211/17143 4032; fax: 49 211/17143 8290.
2
Tel.: 49 211/533 4594; fax: 49 211/533 1698.
0308-5961/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.telpol.2012.04.009

358

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

1. Introduction
In the recent past the number of consumers owning sophisticated mobile handheld devices, so-called advanced or
smart phones, has rapidly increased. Such communication appliances enable people not only to make voice calls or to use
Short Message Service (SMS) while being on the move but also deliver a convincing approximation of the familiar wired
Internet (West & Mace, 2010, p. 282) with a myriad of Internet-based services including email, geo-location, streaming
video, and social networking (Kenney & Pon, 2011, p. 240). Furthermore, advanced mobile handsets are typically capable
of running various consumer and business software applications which can be installed on the device after purchase
(Verkasalo, 2011; Verkasalo, Lopez-Nicolas, Molina-Castillo, & Bouwman, 2010). According to market researchers,
worldwide sales of advanced mobile phones more than quadrupled from 119.7 million in 2007 to 481.3 million in 2011
and is expected to grow further to a sales volume of almost one billion devices by 2016 (Strategy Analytics, 2011; see also
Ovum, 2011).
This development in the mobile communications market has triggered a substantial number of scholarly investigations
on consumers willingness to adopt advanced multipurpose phones for obtaining mobile Internet (MI) access over the
infrastructures of mobile network operators (MNO) and on consumers (stated or objectively observed) use behaviors
concerning mobile data/Internet services beyond SMS. A few examples of this strand of research are Kim (2012), Choi, Kim,
and Kim (2011), Kim and Oh (2011), Pousttchi and Goeke (2011), Tojib and Tsarenko (2011), Tseng and Lo (2011),
Verkasalo et al. (2010) and Okazaki and Hirose (2009). Overall, this work has shown that consumers initial propensity to
adopt advanced mobile handsets and their continued intensity of utilizing such appliances for mobile access to Internetbased/-like services increase as consumers perceive the new data devices and services as more useful and easier to use.
Unfortunately, this insight is neither intriguing nor particularly helpful for MNO managers who develop marketing
measures to promote MI device adoption and use among existing and newly acquired customers: Practitioners simply do
not have the necessary information on MI usefulness or ease of use perceptions among potential or actual advanced device
owners and data service subscribers at hand. Rather, in real life MNO managers have to rely on less ambiguous, objectively
measureable personal characteristics of their (potential) MI customers such as gender, age, or length of business
relationship in order to derive proles of their (prospective) MI users as a starting point for targeting marketing programs.
In addition, most earlier research captured stated behavioral intentions to use MI or subjective self-reports of MI use
frequencies or time through questionnaires lled in by convenience samples of consumers (Peslak, Shannon, & Ceccucci,
2011; Son & Han, 2011; Tojib & Tsarenko, 2011). However, there is ample evidence revealing that MI use intention claims
or subjective use frequency reports correlate only very moderately with measures of actual adoption and use intensity (see
reference, Choi et al., 2011; Gerpott, 2011; Kim & Malhotra, 2005; Szajna, 1996; Verkasalo, 2008a). Hence, the validity of
many MI adoption and use measures in survey-based research is debatable (see reference, Legris, Ingham, & Collerette,
2003, p. 202).
Finally, the overwhelming majority of extant work on initial MI adoption and subsequent use intensity is customer premise
equipment agnostic. This means that it does not at all consider potential differences in personal characteristics and use
intensity of MI customers across distinct device types chosen by consumers to obtain MI access. This is unsatisfactory because
in the past business journalists and scholars conveyed the impression that at least the characteristics and use behaviors of
devoted consumers equipped with Apples iPhone do diverge from those of ordinary MI adopters relying on other handheld
appliances in getting MI access (Arruda-Filho, Cabusas, & Dholakia, 2010; Spehr, 2010).
The purpose of this research is to address some of the omissions and limitations of the literature on characteristics of
MI adopters, MI use intensity and factors explaining interindividual MI use intensity variance. We therefore empirically
explore how MI consumer groups with different types of advanced MI-enabled phones diverge in terms of plain personal
demographic and MNO relationship characteristics as well as with regard to an objective measure of their MI use intensity.
Besides we investigate how MI use intensity is affected by consumers mobile phone type, selected other device features
beyond its basis type and a set of personal characteristics of MNO customers. The studys focus has practical relevance. It
helps MNO managers to better understand the background proles and MI use intensity of MI adopters with different
device types as well as selected drivers of MI use intensity at the level of the individual consumer. Such an understanding
is a prerequisite for developing marketing programs, which account for differences and similarities between subscribers
with a specic MI appliance category.
The remainder of this paper is organized into ve parts. The subsequent section differentiates basic types of advanced
mobile handsets. In addition, it explains the research questions. Section 3 describes the empirical procedures pursued to
obtain measures of the study variables in a large sample of MI consumers of a global MNOs German subsidiary. The
empirical results are then presented. The article concludes in Sections 5 and 6 with a discussion of the practical and
research implications of our work.
2. Device-related foundations and research issues
2.1. Basic types of advanced web-enabled mobile handsets
MI consumers may choose their advanced mobile communication appliance from a large number of phone vendors and
models. Scholars and practitioners agree that from a high-level perspective a mobile devices operating system (OS)

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

359

constitutes the heart (Lin & Ye, 2009, p. 617) of an appliance, which determines a phones features, performance and
security, providing application programming interfaces for add-on applications and technical hooks to manage it
(Malykhina, 2007). Therefore, a mobiles OS is the most important criterion to structure the plethora of cell phones on the
market into a readily comprehensible set of basic handset types (Kenney & Pon, 2011; Lin & Ye, 2009; Peslak et al., 2011).

The analyses of several scholars (e.g., Campbell & Ahmed, 2011; Kenney & Pon, 2011; Muller,
Kijl, & Martens, 2011)
indicate that at this time the mobile OS (and advanced handset types, respectively) with the strongest market relevance
are Apples iOS (iPhone) and Googles Android OS (Android phone).
Apple launched its rst mobile, the iPhone (2), in mid 2007 with a proprietary OS labeled iOS. The model was able to
connect to WLAN and enhanced 2.5 generation [G]) GSM infrastructures, but not to 3G UMTS cellular radio networks
(West & Mace, 2010). Therefore and in accordance with Malykhina (2007), we suggest qualifying Apples rst mobile
handset generation not as an advanced fully web-enabled cell phone. Rather, the rst genuinely advanced mobile allowing
MI access with relatively high transmission speeds over 3G networks sold by Apple was the second generation of the rms
communication appliances, the so-called iPhone 3. The product was introduced in mid-2008. About two years later, in June
2010, Apple unveiled its third generation of mobiles, the iPhone 4 with incremental hard- and software improvements and
a revised design. The main factors distinguishing the iPhone 3 and the iPhone 4 from competing devices with other OS
were an intuitive custom use experience, a browser almost meeting PC standards, a large properly functioning highresolution touch screen and last but not least the availability of extensive audio, video and written contents and software
via Apples App Store (Kenney & Pon, 2011; Ling & Sundsy, 2010; West & Mace, 2010). Furthermore, outside the reach of
Apple the phenomenal uptakes of the iPhone 3 and 4 as well as offerings of other vendors of advanced mobiles were
supported by the increased spatial coverage and speed of cellular 3G radio infrastructures of MNO. These internal and
external factors helped Apple to boost its share in the worldwide advanced mobile phone market from 8.6% (12.6 million
units) in 2008 to 18.1% (87.2 million units) in 2011 (Strategy Analytics, 2011).
However, during the same 4-year period an even faster sales growth was recorded for advanced phones powered by the
mobile OS Android. The rst release of this OS was introduced by the Open Handset Alliance headed by Google in 2008 as
an open source, royalty-free OS. Googles prime motive to promote Android is to expand mobile trafc on the rms search

engine, which can be monetized through advertising (Kenney & Pon, 2011; Muller
et al., 2011). Until today Google was able
to convince a considerable number of major device vendors (e.g., HTC, Motorola, Samsung, Sony) to launch a variety of
advanced competitive phone models running with Android. Consequently, the share of Android mobiles in the global
advanced phone market rose from 0.4% (0.5 million units) in the year 2008 to 49.1% (236.5 million units) in 2011
(Strategy Analytics, 2011; see also Ovum, 2011).
Besides Apple and Android phones there are of course more mobile OS like Microsofts Windows Phone, Nokias Symbian
or RIMs Blackberry OS. However, because the market relevance of these other OS is either signicantly decreasing
(Symbian, Blackberry OS) or highly uncertain (Windows Phone) the present work chose to lump them together in a single
residual class, labeled as other MI-enabled phone.
In summary, the present investigation differentiates between a limited set of advanced mobile phone and corresponding OS types (iPhone3/4, Android, other). Based on this distinction it primarily explores the extent to which
consumers with various advanced appliance types differ in terms of easily discernible personal characteristics and their MI
use intensity.
2.2. Research questions
Regardless of a consumers MI access appliance the probably most frequently analyzed general demographic correlates
of initial MI adoption and subsequent MI usage are gender and age. Concerning gender the dominant tenet is that women
are less thrilled than men by relatively expensive consumer electronics gadgets such as advanced mobiles (e.g., Haverila,
2011; Hong & Tam, 2006; Jiang, 2008). Regarding age the standard line of reasoning is that younger buyers of technologyintensive durable goods such as computer-centric MI access appliances show more interest in using new mobile data
services and communication devices because they are open-minded about integrating innovative appliances into their
everyday activities (e.g., Armey, Vladar, & Pereira, 2011; Hong & Tam, 2006; Jiang, 2008; ODoherty, Hill, Mackay, &
McPherson, 2010). The gender and age considerations are in line with a study of 1000 iPhone users of a Norwegian MNO in
the second quarter of 2008. It found that 84% of them were male and that they were on average about 10 years younger
than users with other types of web-enabled phones (Ling & Sundsy, 2010). Furthermore, several device agnostic surveys
of consumer samples specically in Germany (the country for which we obtained our empirical data) revealed that in
early 2011 women and older persons (especially above 50 years) were underrepresented among MI adopters compared to
the adult population in Germany (Accenture, 2011; Eimeren & Frees, 2011; Pousttchi & Goeke, 2011).
However, based on Arruda-Filho and Lennon (2011) it may be argued that prior iPhone-related results do not hold
anymore because since that time consumers perceive the current iPhone generation as far less innovative and are no
longer convinced that it outperforms appliances of competing vendors. Therefore, the iPhone may have lost its elitist
status and should equally appeal to women and men as well as to consumers of all ages. In addition, there is another
reason why it is increasingly unlikely that users of Android or other advanced mobiles are still overwhelmingly younger
males, namely the massive proliferation of such appliances in the recent past. Particularly for Germany the installed base
of advanced web-enabled cell phones amounted to about 2528 million at the end of 2011 (Bitkom, 2012; Strategy

360

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

Analytics, 2011). Assuming that users of such devices typically fall in the age range between 17 and 65 (53.1 million
persons in Germany; see reference, Statistisches Bundesamt, 2011) this installed base implies that probably more than 50%
of the relevant population in Germany already had an advanced mobile at the end of 2011. Such a considerable penetration
with advanced mobile devices is hardly possible without delivering the appliances in large quantities also to females and
people across a broad range of ages.
Hence, given the sparse, incomplete, and somewhat dated empirical evidence on gender and age distributions among
iPhone, Android handheld and other MI-enabled phone users as well as the preceding supplementary considerations, our
rst two research questions (Q) read as follows:
Q1: In what way do the gender and age distributions of MNO customers differ depending on whether they use an
iPhone, an Android handheld, or any other advanced MI-enabled phone in obtaining mobile access to the Internet?
Q2: In what way do the gender and age distributions of MI users with advanced handhelds differ from those of the
adult population in the study country (Germany)?
In practice, MNO do not only segment their customer base by gender and age but also by attributes, which reect
temporal aspects of customers relationships with their mobile service supplier. Such relationship-oriented individual
characteristics include (1) the span of time that passed from the customers initial enrollment with the provider until
today (customer tenure), (2) the span of time since customers subscribed to their current price plan (tariff age), and (3)
the length of a customers previous MI use experience. Unfortunately, with the exception of length of MI (device) use
experience (Gerpott, 2010; Sugai, 2007) scholarly studies have not yet included such relationship variables for MI adopters
in general and for groups of MI users with distinct computer-centric access appliance categories in particular. In light of
this gap, the following research question is addressed:
Q3: In what way do customer tenure with the MNO, tariff age and length of MI use experience of MI adopters differ
depending on whether they operate an iPhone, an Android handheld, or any other advanced MI-enabled phone
when obtaining MI access?
To date, there is a dearth of scholarly work regarding the extent to which the volume of up- and downloaded IP trafc
generated by MI users (MI use intensity) varies as a function of the advanced phone type through which the Internet is
accessed. A few qualitative macro-level examinations of the competitive positioning of major handheld vendors and
mobile OS developers (Kenney & Pon, 2011; West & Mace, 2010) imply that consumers with iPhones should display a
higher MI use intensity than their counterparts relying on advanced mobiles from various vendors with other OS. Two
reasons are quoted for the assumed use intensity difference. First, the authors believe that Apples iPhone delivers a
mobile use Internet experience closer to a PC than any other previous mobile phone (West & Mace, 2010, p. 276). Second,
it is stated that the trafc of iPhone owners is positively inuenced by the possibility to download an unprecedented
variety of software from Apples App Store in order to expand and to customize their handhelds capabilities (Campbell &

Ahmed, 2011; Lin & Ye, 2009; Muller


et al., 2011).
These arguments receive support by studies of Amethon (2008), Ling and Sundsy (2010) and West and Mace (2010).
They provide quantitative MI trafc analyses, which unanimously suggest that at least in the years 2008 and 2009 the level
of mobile IP trafc produced by iPhone users signicantly exceeded that of MNO customers who access the Internet with
other handheld types. The research of these authors, however, did not control for the customers tariff type (useindependent/at versus use-dependent/metered) when comparing MI users with different phone types. Unfortunately,
earlier investigations (e.g., Gerpott, 2010; Harno, 2010) found strong rate scheme effects on MI use behaviors. Therefore, it
might well be that the observed above average MI use intensity of iPhone consumers was mainly due to differences in the
shares of at rate subscribers across mobile phone types and not caused by any reputed superior features of Apple
handhelds.
Furthermore, since the completion of the three comparative studies a substantial number of Apple competitors have
introduced a plethora of new models equipped with well working OS and improved user interfaces. This development may
have eroded prior competitive advantages of the second and third iPhone generations. The changed supply situation
together with the growing sales volumes of Apple phones not any longer only to the vanguard of early MI adopters (see
Section 2.1 above) could well have resulted in an equalization of the average mobile IP trafc generated by iPhone
consumers compared to that of users with other handheld device types. Indeed, in two recent investigations on MI use
behavior variations as a function of consumers device type (Ericsson, 2011; Matthews, Pierce, & Tang, 2009) iPhone and
Android mobiles were combined into the same category of advanced communications appliances because their OS were
supposed to provide equivalent functionalities to the consumer.
In view of the shortage of up-to-date empirical comparisons of MI use intensity variations by advanced mobile phone
type and the previous considerations, it appears worth to pose the following research question:
Q4: In what way does MI use intensity of MNO customers differ depending on whether they operate an iPhone, an
Android handheld, or any other advanced MI-enabled phone in obtaining MI access (when controlling for MI
tariff plan)?

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

361

To date, very few studies have explored the diversity or distribution of use behaviors of MI adopters. Based on the
applied parametric statistical signicance testing procedures it may be concluded that past empirical work assumed that
MI use intensity is normally distributed in the MI adopter population. Exceptions are the studies of Verkasalo (2008a,
2008b) who obtained actual MI use data for a sample of 565 Finnish consumers in 2005 and 2006 and of Gerpott (2010)
who analyzed use behaviors for a sample of 443 MI adopters of an MNO in Germany during the second quarter of 2008.
Both researchers found that the skewness statistics for the distribution of the in-/outgoing IP data volume generated by
mobile consumers were strongly positive, i.e., few adopters caused a disproportionately large share of a samples
cumulative MI data trafc. However, taking into account the recent progressing change of the MI business from a niche to
a mass market it is not clear whether the pertinent prior ndings are still valid. Therefore, we investigate the research
questions:
Q5: How skewed is the distribution of MI use intensity among consumers who obtain MI access through advanced
mobile phones?
Q6: In what way does the skewness of the distribution of MI use intensity among MNO customers differ depending
on whether they operate an iPhone, an Android handheld, or any other advanced MI-enabled phone in obtaining MI
access?
Research question 4 implicitly touches upon the quest for explaining interindividual variance in MI acceptance
measures such as pre-adoption MI subscription intentions or post-acquisition MI use intensity. It suggests itself to broaden
such a search for explanatory factors of MI use intensity beyond handheld OS types by exploring associations between the
personal demographic and MNO relationship characteristics of MI adopters taken up in research questions 13 on the one
hand and MI use intensity on the other. Consequently, we look into the following research question:
Q7: In what way are gender, age and MNO relationship characteristics of MI adopters associated with MI use
intensity?
Such an analysis can be pushed even further by incorporating additional device features discussed in the literature as
important potential MI use intensity drivers. One characteristic of this kind is the screen size (length of diagonal) of
advanced handsets (Ling & Sundsy, 2010, p. 221; West & Mace, 2010, p. 275). At a rst glance it appears obvious that
screen size should be positively related to MI use intensity because large displays make it easier to handle all sorts of
contents (e.g., emails, video, pictures, written documents). However, larger screen sizes may also have the effect that it is
less convenient for consumers to carry their devices along at any occasion which may in turn lower subscribers MI use
intensities (Kim, Sundar, & Park, 2011). Empirical work related to associations between screen size of a telecommunication
device and MI use intensity has produced equivocal results (Bao, Pierce, Whittaker & Zhai, 2011; Chae & Kim, 2004;
Ericsson, 2011; Matthews et al., 2009; Serif & Ghinea, 2008). Moreover, pertinent earlier work has not explored unique
screen size effects on MI use intensity after controlling for personal user characteristics and other device features
(particularly mobile OS). Hence, this study addresses the research question:
Q8: In what way is a handsets screen size associated with MI use intensity (when controlling for mobile OS)?
Another advanced mobile feature suggested as a potential driver of MI use intensity is the time that has passed since the
initial market release of a model (Ericsson, 2011; Ghose & Han, 2011). This characteristic constitutes a sort of black box
variable which embraces all kinds of technical and design improvements of newly launched phones compared to
predecessors. The analyses of Ghose and Han (2011) and Ericsson (2011) revealed that this factor was negatively
correlated with consumers MI use intensity. Explanations for the association include that more recently released devices
spark more usage because of their novelty (Arruda-Filho & Lennon, 2011; Shepard, Rahmati, Tossell, Zhong, & Kortum,
2010), better technical features and because of the social prestige gained through the demonstration of the ownership of a
cutting-edge device in the presence of important social reference groups. However, besides the studies of Ghose and Han
(2011) and Ericsson (2011) no empirical work could be located that included the time interval since the market launch of a
users phone model as a correlate of MI use intensity. We therefore incorporated the subsequent research question into our
analysis:
Q9: In what way is the time that has elapsed since the market release of consumers advanced phone model
associated with MI use intensity (when controlling for mobile OS)?
The preceding nine research questions did not differentiate between the two generations of genuinely advanced Apple
phones iPhone 3 and iPhone 4 which the rm has marketed up to the time when this article was written. The rationale
for doing so is as follows: Although the iPhone 3 and the iPhone 4 series display some differences with regard to their
internal memory (256 versus 512 MB), screen resolution (480  320 versus 960  640 pixel), number of cameras (1 versus 2),
camera resolution (3 versus 5 megapixel) and capabilities to support higher transmission speeds of the latest UMTS releases,
the core design features (mobile OS, large touch-screen, intuitive user interface, App Store access) remained identical. Thus, it
is justiable to refrain from distinguishing between iPhone 3 and iPhone 4 MI users in the research questions. Nevertheless,

362

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

in a spirit of exploration we designed the quantitative analysis such that we also looked for the necessity to distinguish
between iPhone 3 and iPhone 4 users when empirically exploring the data.
3. Empirical methodology
3.1. Data collection procedure and sample
The data were collected in collaboration with the German subsidiary of a globally active MNO, which has run 2G and 3G
networks in Germany for many years and has recently started to roll out a 4G infrastructure in this country. The rm
agreed to support an academic research project by granting the authors access to its German subscriber database in order
to extract the information required to address the study questions for a specic group of customers. The target group was
dened as private (non-business) adult (Z17 years) contract subscribers who had a at MI tariff at least during the time
period of April to June 2011.
The sample restriction to subscribers with a use-independent data tariff was chosen to control for rate scheme effects
on MI use behavior. The rationale for excluding business accounts and focusing on consumers was to partial out impacts of
diverging MI use purposes (business vs. personal) on MI use intensity also observed in prior work (e.g., Gerpott, 2011; Kim
& Garrison, 2009; Tojib & Tsarenko, 2011). Finally, the requirement that participants had not changed their data at rate
schemes during the second quarter of 2011 was imposed to ensure that subjects had at least one month of practical
experience with the MI at rate to adjust their MI use patterns to this type of tariff scheme before MI use intensity was
measured in May and June 2011.
Application of these selection criteria resulted in a population of about 0.8 million customers of the cooperating MNO
from which we were granted the permission to draw an anonymized random sample of 10,000 consumers. For each of
them the MNOs customer administration system contained information regarding the advanced mobile phone model
operated to obtain MI access. Therefore, we were able to identify 378 customers who used different MI access appliance
types during the three study months, i.e., who switched between the appliance types differentiated in this investigation.
They were discarded from the analysis as they could not be conclusively assigned to a single device category. Besides 301
cases were excluded because the MNOs billing system recorded that they generated no mobile Internet trafc both in May
and in June 2011 and hence were no true MI users during the study period. As a result, 9321 subscribers remained who
constitute the sample analyzed in the remainder of this work. They were distributed as follows across the considered
appliance types: iPhone 3:959 (10.3%), iPhone 4:2213 (23.7%) Android mobile: 2410 (25.9%) and other MI-enabled
phone: 3739 (40.1%). 50.1% of the persons in the other group were equipped with a Symbian mobile and the remaining
49.9% were fairly evenly distributed between phones with seven other mobile OS.
Of the selected customers, 62.3% were male (n 9300). The mean participant age as of June 30, 2011 was 33.6 years
(SD 12.0; median30.0; n 9321). The average number of months passed since a customers enrollment with the
collaborating MNO (customer tenure) amounted to 69.8 months (SD 53.4; median 59.3; n 9321). In the sample the
mean number of months that had passed between June 2011 and the month for which the MNO rst recorded an MI data
trafc volume exceeding the threshold of 1 MB (MI use experience) for the selected subscribers was 20.6 months
(SD 16.1; median14.0; n9321). Just 22.7% of the subjects had a relatively short MI use experience of not more than
six months. As of end of June 2011, the average participant had subscribed to her/his current data price plan 9.5 months
ago (SD 5.4; median 8.0; n 9321).
3.2. Variable measurements
Measures for the demographic and MNO relationship variables were extracted from archival customer les of the
cooperating supplier. Gender of the subjects (1male, 0 female) had been regularly recorded by the MNO when a new
customer had signed a contract. At the same occasion subscribers had to present their ID card from which the MNO noted
(among others) the date of birth of the customer. Each subjects age was quantied by calculating the time span between
the end of our study period (30 June 2011) and her/his birthday.
The rst MNO relationship characteristic of a participant, tenure with the provider, was captured by subtracting the date
at which a person initially signed the contract with the collaborating MNO from 30 June 2011. If a customer had quitted an
earlier contract with the focal rm before this signing date to switch to a competitor and then returned to the collaborating
supplier the aforementioned earlier contract had not been recorded in the MNOs customer data les. Thus, we measure of
a subscribers tenure underestimates the total length of a customers business relationship with the MNO for those
individuals with an unsteady transaction history.
The second MNO relationship characteristic, tariff age, was gauged by counting the number of months that had passed
between the date at which the customers had subscribed to their current data rate scheme and the end of the study period
(i.e., June 2011).
The last MNO relationship variable, length of MI use experience, was measured by searching the MNOs billing system for
the rst month during which mobile data trafc of at least 1 MB was recorded for a customer. The threshold of 1 MB was
chosen to ensure that accidentally generated low volumes of MI trafc (e.g., due to incorrect operation of a device by the
consumer) were discarded in determining the initial month of a subscribers deliberate use of the MNOs MI access

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

363

offering. Again, for this variable it cannot be surely ruled out that it slightly underestimates the duration of a persons exact
length of MI use experience because the individual may have gained MI experience while being served by other
competitors of the focal MNO.
Information on the screen size (length of diagonal in cm) and handset-type age (quarters elapsed since the participants
phone model was launched in the market) of the participants phone was collected from the collaborating rms archives.
In line with the recommendation of Turner, Kitchenham, Brereton, Charters, and Budgen (2010, p. 471), an objective
system-captured (Sharma, Yetton, & Crawford, 2009, p. 479) MI use intensity measure was extracted from the
collaborating MNOs billing engine. Earlier MI use research implies that the monthly volume of up- and downloaded
IP-switched data trafc via an MNOs network is an adequate metric to capture MI use intensity at the individual consumer

level (Gerpott, 2010; Ling & Sundsy, 2010; Smura, Kivi, & Toyli,
2009). This operationalization approach has the benet
that other mobile data services such as SMS, MMS, or WAP-based applications which are distinct from direct MI access are
eliminated from the measurement domain. Stated differently, this metric precisely connes itself to MI as the currently
most important sub-area of mobile data services. In addition, a customers IP trafc volume reects both the frequency and
the duration of MI use cases (see reference, Ghose & Han, 2011). Finally, particularly in the present study individual MI
trafc gures are not distorted by differences in the tariff schemes of consumers (use-independent versus use-dependent)
because our sample was limited to customers with a at rate (see Section 3.1).
We obtained the accumulated IP trafc quantities (in kilo-/mega-/gigabyte [KB/MB/GB]) carried over the MNOs GSM-/
UMTS-radio infrastructures and recorded for the selected subscribers by the MNOs billing system for May and June 2011,
respectively. In the total sample the average MI use intensity was 256.6 MB (SD 445.2; median 92.1) in May 2011. The
June volume mean amounted to 228.2 MB (SD 364.5; median 91.6). May and June MI use values were highly correlated
(r 0.72; p r0.001). Consequently, we construed this studys MI use intensity measure by averaging the mobile IP trafc
which a consumer generated in May and in June 2011. The internal consistency reliability of this measure was excellent
(Cronbachs a 0.83). In the total sample the monthly MI use intensity in May and June 2011 averaged 242.4 MB
(SD 376.0; median 101.0).
In the few earlier studies containing user-level monthly MI trafc data from consumers in ve rst world countries in
2005/06 (Verkasalo, 2009, 2011) and from German subscribers in 2008 (Gerpott, 2011) the averages of this MI use
intensity metric ranged from 1.9 MB per month in 2005 to 30.3 MB per month in 2008. Thus, the present studys monthly
MI trafc mean is noticeably higher than comparable statistics reported in earlier investigations. This divergence may be
explained through decrements in mobile data tariffs, improvements in end user equipment and progress in radio network
performance in the recent past. These changes have led over time to strong MI use increases worldwide. More precisely,
according to Cisco (2011) global MI data trafc has more than doubled in 2011 compared to 2010. However, relative to
broadband access via xed networks (copper, coax, ber), the average monthly MI use intensity observed in the sample
during the study period in 2011 is still low. In Germany, the average monthly IP trafc volume per landline broadband
access added up to 11.6 GB in 2011 (Bundesnetzagentur, 2011, p. 42). Hence, in May and June 2011 the average consumer
in our sample reached just 2.1% of the monthly IP data volume generated by a typical xed network broadband user in
Germany in 2011.
4. Empirical results
4.1. Research questions 1 to 4
Although research questions 1, 3 and 4 are univariate questions, the six demographic, MNO relationship and use
behavior study variables addressed with them may be interrelated. Therefore, in a rst step multivariate analysis of
variance (MANOVA) was conducted to achieve some control over the experimentwide error rate (Hair, Black, Babin,
Anderson, & Tatham, 2006, p. 401). After eliminating subjects with missing values 8713 cases remained for the MANOVA.
Four statistical criteria appropriate for testing overall signicance of differences in several dependent measures (Roys
greatest characteristic root, Wilks l, Hotellings trace, Pillais trace) unanimously indicated that the total set differences in
the dependent variables by advanced phone type was statistically signicant at p r0.001. Hence, it is admissible to
calculate six separate univariate analysis of variance (ANOVA) with MI phone type as the independent factor. Finally,
pairwise post hoc comparisons were made with a-levels adjusted according to the Bonferroni inequality to control for
overall rate of falsely rejecting a true null hypothesis. These three steps were supplemented by a multiple discriminant
analysis to gain a better understanding of the underlying multivariate data structures.
According to Table 1, statistically signicant differences across the four studied advanced phone types were found both for
gender and age. However, the share of males with either an iPhone 4 or an Android device was just about 7 to 8 percentage
points higher than among consumers with an iPhone 3 or another MI-enabled appliance. The mean and median age of
subscribers with a handset powered by the Android OS tended to be lower than that in the remaining three subsamples,
particularly compared to the group using miscellaneous other MI-enabled mobile devices. The maximum pairwise age mean
and median group divergences did not exceed 2.8 and 4.0 years, respectively, i.e., regardless of their statistical signicance
(partly caused by the studys large sample size) age differences were rather limited in absolute terms.
Research question 2 was taken up by comparing the gender and age distributions of our total sample and each of the
four phone type groups against three benchmarks (see Table 2). The rst was the current distribution of the German adult

364

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

Table 1
Differences in demographic, MNO relationship and mobile Internet use variables between consumer groups with varying types of advanced mobile
phones.

iPhone 4
(2204 r nr 2213)

iPhone 3
(957 r nr 959)
Variablesb

SD

SD

Other MI enabled
phone
(3736 r nr 3739)

Android phone
(2403 r nr 2410)
M

SD

Demographic characteristics
1. Gender (1 male, 0 female)
59.6%a
66.3%b
65.7%b
2. Age (years)
33.3a [30.0a]
11.1 33.0a [30.0a]
11.6 32.2a [28.0b]
11.7
MNO relationship characteristics
c
a
a
b
b
b
b
3. Tenure with provider
74.6 [62.2 ]
54.6 64.9 [51.0 ]
56.6 64.6 [52.6 ]
51.9
4. Tariff aged
11.0a [10.8a]
5.8
7.2b [7.2b]
3.4
8.7c [7.1b]
5.3
5. Mobile Internet use
19.7a [13.0a]
15.0 18.5a, c [8.0b]
16.3 19.9a, d [13.0a] 16.1
experiencee
Mobile Internet use behavior
6. Use intensityf
305.4a [173.9a] 371.3 417.0b [266.7b] 441.9 302.3a [174.1a] 392.2

ANOVA F-ratioa
(signicance
level)

SD

58.5%a
35.0b [32.0c]

12.6

75.0a [70.6a]
11.0a [10.3a]
22.8b [17.0c]

51.6 28.31 (p r 0.001)


5.8 301.66 (p r 0.001)
16.0 34.85 (p r 0.001)

84.3c [9.6c]

52.77 (p r 0.001)
29.97 (p r 0.001)

236.9 470.40 (p r 0.001)

For gender the statistic shown is a w2-value with 3 degrees of freedom.


Figure in squared brackets below mean Median. In each row, means (medians) with diverging superscripts are signicantly different at pr 0.05
according to Tamhane T2 post-hoc-tests (non-parametic pairwise median tests). For gender pairwise w2-tests were run. In pairwise median and w2-tests
the a-level was adjusted to 0.008 (0.05/6) according to the Bonferroni inequality to maintain an overall error of 0.05 in (falsely) rejecting the null
hypothesis.
c
Number of months that passed since the customers enrollment with current mobile network operator (MNO).
d
Number of months that passed since initial subscription of/switch to current tariff plan.
e
Number of months elapsed since ones current MNO recorded the rst month with a mobile Internet data volume exceeding 1 Megabyte (MB).
f
Average monthly mobile Internet data volume (in MB) generated in May and June 2011.
b

Table 2
Age distribution of subsamples with different phone types, population in Germany and of two benchmark studies.
Age interval (years)a
Phone type
1. iPhone 3
2. iPhone 4
3. Android phone
4. Other MI-enabled phone
Total sample
Population in Germany
Eimeren & Frees studyb
Accenture studyc

1719 (%)

2029 (%)

3039 (%)

4049 (%)

5059 (%)

3.4
5.0
5.7
3.8
4.5

43.7
42.9
46.8
39.2
42.5

25.0
25.5
21.5
23.1
23.4

18.6
17.1
15.8
19.4
17.8

6.4
7.0
7.6
10.6
8.5

3.7
14.5
24.3

14.4
32.0
25.0

14.2
21.9
22.3

19.8
19.3
18.3

16.9
7.5
10.1d

Z 60(%)
2.9
2.5
2.6
3.9
3.3
31.0
4.8

n
959
2213
2410
3739
9321
69.2 million
360
876

a
Data for the four phone types subsamples refer to a persons age as of June 30, 2011. Data for the population in Germany refer to the age of people
living in Germany as of December 31, 2010.
b
Telephone survey of 1800 persons living in Germany aged at least 14 years. Date of data collection: March 10, 2011 to April 21, 2011. Random
sample representative of the German population with regard to age and gender.
c
Online survey of 3128 persons living in Germany and aged at least 14 years. Date of data collection: January 2011.
d
Participants aged at least 50 years were combined into one age group.

population across six age intervals. The second and third were age distributions derived from two surveys of Eimeren and
Frees (2011) and of Accenture (2011) on MI use in Germany in early 2011.
The share of males in the German population is currently around 49.1% (Statistisches Bundesamt, 2011, p. 28). This is
evidently lower than the 62.3% proportion of males in the present studys total sample (w2 650.60; df 1; p r0.001) and
in each of the four phone subsamples. However, relative to the proportions of males of 68.6% and 68.5% among MI adopters
found by Eimeren and Frees (2011) and Accenture (2011), respectively this samples share of males was signicantly
smaller (w2 170.13; df 1; p r0.001 and w2 164.47; df pr0.001, respectively). Noting that our data was collected
about half a year later than the data of the two benchmark studies, it may be concluded that the gender divide with regard
to MI subscription is narrowing fast.
According to the results in Table 2, MI users in the sample were evidently younger than the adult population in
Germany (Statistisches Bundesamt, 2011, p. 28). The difference in the age distribution of the study subjects and that of the
adult population in Germany was extremely signicant (w2 8419.72; df5; p r0.001). In a similar vein, the age
distributions of each of the four phone type groups differed clearly (p r0.001) from that of the adult German population.
Furthermore, although the age distribution divergences were less pronounced when contrasting the present subjects with
those of Eimeren and Frees (2011) and Accenture (2011), the share of persons aged 20 to 39 was still signicantly higher

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

365

(pr0.001) in the present total sample as well as in each of the four phone type groups than in the quoted two German
benchmark studies. Analog to the previous line of reasoning for gender this results pattern may be taken to point in the
direction that the MI adoption gap between young and old consumers is not closing swiftly.
Research question 3 concerned divergences in three MNO relationship characteristics between MI adopters with
varying advanced phone type. As shown in Table 1, each of the three variables differed signicantly depending on
consumers advanced phone types. The largest deviation was found for tariff age: iPhone 3 and other MI-enabled phone
customers had subscribed to their current rate plan earlier than consumers with an iPhone 4 or an Android model.
Similarly, users with an iPhone 4 or an Android device tended to diverge from the other two groups such that the former
had signed their initial contract with the MNO about 10 months later. In terms of MI use experience iPhone 4 subscribers
had the lowest mean and median values of all subsamples. Taken together the results patterns for the three MNO
relationship characteristics leave room for the following thesis: A considerable share of the iPhone 4 users and to a lesser
extent of the consumers with an Android appliance may have been motivated to choose their current data at price
scheme and even have been acquired by the collaborating MNO because the rate plan was bundled with the provision
of an iPhone 4 or another fashionable advanced device running on Android. In contrast, consumers equipped with
MI-enabled phones using OS other than iOS or Android appear to be have been acquired a considerable time ago and were not
sufciently intrinsically or extrinsically incentivized to exchange their device against one representing the latest state of the art.
Research questions 4 dealt with MI use intensity deviations across advanced phone categories. The analysis in Table 1
reveals that there were large mean and median differences in the monthly MI data volume in May and June 2011 across device
categories. iPhone 4 consumers generated more trafc than subscribers in any of the three other phone type groups. Similar MI
use intensity mean and median values were found for subscribers equipped with an iPhone 3 or an Android appliance. Subjects
with MI-enabled phones running not on iOS and not on Android had by far the lowest MI use intensity. Compared to MI use
intensity statistics reported in earlier work, which looked at individuals accessing the Internet via a handset type other than an
iPhone 3/4 or an Android device (Gerpott, 2011; Verkasalo, 2008a, 2008b), the data consumption values of individuals with an
iPhone 3, iPhone 4 and an Android appliance in the present investigation were remarkably higher. In contrast, MI use intensity
of the sampled customers with MI-enabled phones not powered by iOS or by Android closely resembled that detected in prior
research. This observation in combination with ndings of Ling and Sundsy (2010) could imply that a switch to an iPhone or
to an advanced Android appliance actually triggers positive changes in MI usage among consumers over time.
Since the results in Table 1 allow only limited insights with regard to the relative contribution of the six study variables
in distinguishing MNO customers with various phone types two supplementary discriminant analyses were run. The rst
included each of the four device groups and the second excluded the group of customers using miscellaneous handsets
with OS other than iOS or Android. The overall multivariate Wilks l criterion and its level of statistical signicance
(pr0.001) in the rst analysis (see Table 3) are identical with the MANOVA results reported in Section 4.1. In the fourgroup-analysis the rst canonical function accounted for 89.6% ((0.9800.806)/(10.806); see Wilks l values in Table 3)
of the variance explained. The second three-group-analysis also achieved a statistically signicant (p r0.001) overall
multivariate Wilks l with the rst function absorbing 85.7% of the variance explained ((0.9860.902)/(10.902); see
Wilks l values in Table 3). Hence, the capability of the rst function to separate users with different appliance categories
Table 3
Multivariate discriminant analysis of mobile Internet customers with different advanced phone types.
Function Ia
Discriminating variablesb
1.
2.
3.
4.
5.
6.

Gender
Age
Tenure with provider
Tariff age
Mobile Internet use experience
Use intensity

Function IIa

SW
114
082
072
 595
 224
746

SC
[  143]
[  118]
[084]
[908]
[009]
[331]

124
045
 137
 654
 227
749

SW
[  149]
[  001]
[177]
[927]
[105]
[370]

 218
037
474
715
 468
607

Function IIIa
SC

[  252]
[338]
[528]
[242]
[  489]
[821]

 169
082
266
658
 178
463

Canonical correlation
Wilks l
(p r )

421 [290]
806 [920]
(001) [(001)]

137 [120]
980 [986]
(001) [(001)]

041
998
(005)

Group centroids
iPhone 3 customers (n 957)
iPhone 4 customers (n 2203)
Android customers (n 2394)
Other phone type customers (n 3159)

 071 [541]
618 [  315]
184 [073]
 549

393 [154]
 031 [081]
 057 [  136]
 054

 006
044
 062
019

[  198]
[336]
[339]
[202]
[  217]
[612]

SW

SC

 394
757
283
 259
 092
 036

 327
847
541
 216
146
 224

a
SW Standardized weights. SC Structure correlations. Leading decimals are omitted. Example:  218  0.218. Percentage of correctly classied
cases for four [three] groups 46.8% [47.5%]. This hit ratio is 18.5 [8.8] percentage points above the proportional chance criterion of 28.3% [38.7%]
(see reference, Hair et al. 2006, p. 302).
b
For an explanation of the discriminating variables see Table 1.
c
Figures in squared brackets refer to results after exclusion of the group other phone/OS type customers.

366

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

by far exceeded that of the remaining canonical variate(s) both in the three- and in the four-group-analysis. The three
functions correctly assigned 46.8% of the 8713 cases in the analysis with four device groups. This is equal to a 65.4%
increase compared to the proportional chance criterion of 28.3% resulting from the unequal sizes of the four groups in the
present sample. In the three-group-analysis 47.5% of the 5554 subjects were righty classifed which corresponds to a 22.7%
gain compared to the proportional chance criterion of 38.7% resulting from the unequal sizes of the groups in the present
sample. Hence, only the rst discriminant analysis fullls the requirement that its predictive accuracy y should exceed
any criterion value by at least 25% (Hair et al., 2006, p. 306). Therefore, substantive interpretations of the contributions/
correlations of the study variables to/with the rst discriminant function capturing most of explained variance are
acceptable only for the analysis considering four phone types.
Table 3 reports standardized discriminant weights for both analyses. The absolute values of the weights represent a variables
relative contribution to a function. However, they may be distorted by correlations among the predictors. Consequently, Table 3
additionally shows the structure correlations (synonymous: discriminant loadings) of each variable with the functions, which are
less affected by multicollinearity. According to Hair et al. (2006, p. 312), loadings exceeding 7.40 are considered substantive for
interpretation purposes. Thus, the four-group-analysis results in Table 1 reveal that MI use intensity and tariff age are
substantively important variables in the rst discriminant function. As indicated by the four group centroids (see Table 3) this
rst function primarily separates consumers in the other phone group (low use intensity, high tariff age) and iPhone 4 group
(high use intensity, low tariff age) from iPhone 3 and Android users (moderate use intensity and tariff age).
In summary, the analysis in Table 3 suggests that use intensity followed by tariff age are the most important
characteristics in distinguishing iPhone 4 and other phone type consumers from the two other device groups. Gender, age
and the two other studied MNO relationship variables contribute little to separating the four appliance groups.
4.2. Research questions 5 and 6
Research questions 5 and 6, which concern the skewness of the distribution of MI use intensity in the sample were
taken up by appraising the skewness and variation coefcients of the MI use intensity metric both in the total sample and
in the four device category groups. Furthermore, participants were split into use intensity quartiles. In the total sample the
skewness statistic for MI use intensity was 3.26 and its coefcient of variation amounted to 1.55. An earlier study found
skewness and coefcient of variation values of 3.89 and 2.24, respectively in a small sample of 443 MI adopters in
Germany most of whom used laptops as their primary MI access devices (Gerpott, 2011). Against these yardsticks it may

Table 4
Mobile Internet use intensitya for consumers with different advanced phone types.
Advanced phone type
Statistic

iPhone 3

iPhone 4

Android phone

Other MI- enabled phone

n
M
SD
Skewness
Coefcient of variation (SD/M)
Sum/total

959
305.4
371.3
2.6
1.2
292888.3

2212
417.0
441.9
2.1
1.1
922356.9

2409
302.3
392.2
3.2
1.3
728303.5

3738
84.3
236.9
8.1
2.8
315082.2

Percentile 125 (n)


Class minimum
Class maximum
Class share of sample total MI use

(240)
0.1
69.0
2.6%

(554)
0.0b
115.9
3.6%

(621)
0.0b
68.9
2.3%

(981)
0.0b
0.5
0.0%c

Percentile 2650 (n)


Class minimum
Class maximum
Class share of sample total MI use

(240)
69.4
173.9
9.7%

(552)
116.1
266.4
10.9%

(587)
69.3
174.9
9.6%

(909)
0.5
10.0d
1.1%

Percentile 5175 (n)


Class minimum
Class maximum
Class share of sample total MI use

(240)
174.9
398.7
22.3%

(555)
267.0
565.9
23.4%

(600)
175.3
374.8
21.5%

(918)
10.0 d
63.9
8.6%

Percentile 76100 (n)


Class minimum
Class maximum
Class share of sample total MI use

(239)
404.9
2,829.1
65.4%

(551)
566.4
3,436.7
62.1%

(601)
376.6
4,173.4
66.6%

(930)
64.1
5,033.9
90.3%

Average monthly Megabyte (MB) mobile Internet trafc in May and June 2011.
Minimum two digit value 0.01 for iPhone 4 and other phone types, respectively; 0.02 for Android phone.
Two digit value 0.03%.
d
Two digit values for class maximum (minimum) of percentile 2650 (5175) 10.02 (10.04).
b
c

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

367

be hypothesized that MI use intensity is less unevenly spread among consumers accessing the Internet with a handset
than among mobile subscribers with a device resembling more a desktop computer than a phone.
The results in Table 4 show that the skewness index of the intensity measure was also strongly positive in each of the
four appliance type subgroups but varied considerably from 2.07 (iPhone 4) to 8.14 (other MI-enabled phone). Similarly,
subsample variation coefcients displayed a large range from 1.06 (iPhone 4) to 2.81 (other MI-enabled phone). MI use
intensity distributions were not too dissimilar between iPhone 3, iPhone 4 and Android phone consumers. As can be seen
from Tables 4, 65.4%, 62.1% and 66.6% of the average monthly MI trafc in May and June 2011 recorded for consumers with
an iPhone 3, iPhone 4 and Android phone, respectively was generated by those 25% of the subjects in a device group with
MI use intensity values exceeding the highest quartile cut-off value (see last line in Table 4). Thus, for each of the three
appliance subsamples, there existed a limited number of heavy users generating MI data volumes far above the subsample
median with extreme values extending to 2.8 GB, 3.4 GB and 4.2 GB per month in the iPhone 3, iPhone 4 and Android
phone group, respectively. The group with other MI-enabled phones displayed a MI use intensity distribution that differed
remarkably from that of the three other subsamples: In the other group 25% of its members caused 90.3% of the MI trafc
(see last line in Table 4). There were 47 subscribers in this group each of whom generated a monthly MI trafc volume of at
least 1 GB. Put differently, in the other subsample 1.3% very heavy users were responsible for 23.7% of the overall MI trafc
of this group. 54.3% of the very heavy MI users were equipped with a Symbian phone, the remainder was spread across
phone models with six other mobile OS (mainly Windows Phone, Bada).
All in all, the analysis indicates that a large share of consumers with a at data rate entailing substantial costs (around
30 h per month) exploited the MI usage options embedded in their tariff scheme only to a very limited extent. This
observation applies somewhat less to consumers with an iPhone 3/4 or an Android phone and more to those using another
MI-enabled phone type. However, across all appliance types under study there existed a small group of subscribers
generating a very large proportion of the samples overall MI trafc.
4.3. Research questions 7 to 9
Research questions 7 to 9 touch upon associations between ve objectively measurable personal attributes of MI adopters
and features of their handheld appliance (mobile OS, screen size, handset type age) on the one hand and their MI use intensity
on the other. Bivariate correlation and multivariate ordinary least squares (OLS) regression analyses were conducted to address
these questions. The ve personal attributes and three dummy-variables reecting the mobile OS of a participants handset and
the two other appliance features were simultaneously entered into the regressions as predictors of MI use intensity. Analyses
were run both for the full sample of 8712 participants and a reduced sample which dropped customers not equipped with an
Table 5
Regressions of mobile Internet use intensity.
Reduced samplea (n 5554)

Full sample (n 8712)


Independent variablesb
Demographic characteristics
1. Gender (1 male)
2. Age (years)
MNO relationship characteristics
3. Tenure with providere
4. Tariff agef
5. Mobile Internet use experienceg
Device characteristics
6. iPhone 3/iOS (1yes)
7. iPhone 4/iOS (1yes)
8. Android phone (1 yes)
9. Screen size (diagonal in cm)
10. Quarters since market release of device
Regression R2
Regression F-value (po )
a

po

VIFc

Simple r

po

VIF

Simple r

058d
 137

001
001

1.032
1.184

074
 185

054
 169

001
001

1.028
1.190

065
 214

 113
 036
093

001
001
001

1.415
1.201
1.250

 148
 137
007

 123
 009
106

001
526
001

1.432
1.227
1.244

 152
 074
051

152
302
157
111
 021

001
001
001
001
133

1.517
1.562
1.662
1.775
2.096

043
241
073
242
 176

121
170

001
001

3.155
1.277

105
 099

001
001

1.241
3.525

 048
133
 095
307
 103

173
181.95(001)

104
71.30 (001)

Exclusion of 3158 customers (36.2 % of total sample) with other MI-enabled phone.
In the full sample regression, users with a phone with other operating systems (OS) than iOS or Android ( other OS/phone type) were the reference
category. Changing the reference category to Android results in the following additional b-values (signicance levels): iPhone 3 b 0.043 (p r0.001),
iPhone 4 b 0.149 (p r 0.001). Switching the reference category to iPhone 3 yields the following additional b-value: iPhone 4 b 0.090 (pr 0.001). In the
reduced sample regression, users with an Android phone are the reference category. Changing the reference category to iPhone 3 results in the following
supplementary b-value: iPhone 4 b 0.013 (p r 0.668).
c
VIF Variance ination factor. VIF 1/(1  R2). R2 Coefcient of determination of a regression of a focal variable on all remaining predictors.
d
Except for VIF- and regression F-values leading decimals are omitted. Example:  137  0.137.
e
Number of months that passed since the customers enrollment with current mobile network operator (MNO).
f
Number of months that passed since initial subscription of/switch to current tariff plan.
g
Number of months that elapsed since ones current MNO recorded the rst month with mobile Internet data trafc exceeding 1 Megabyte (MB).
b

368

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

iPhone 3/4 or an Android phone. The sample diminishment was introduced to explore the extent to which the correlations
addressed in the research questions hold after excluding subscribers who work with phones powered by miscellaneous mobile
OS. Besides, in order to check the robustness of the OLS results ordinal regressions (OR) were computed, which other than the
OLS analysis do not assume that MI use intensity is normally distributed in the sample. In essence, the OLS and OR outcomes
did not deviate substantially. Therefore, we restrict the result presentation to the OLS regressions.
The predictors explained a signicant share of 17.3% and 10.4% of the MI use intensity variance in the full and in the
reduced sample, respectively (see R2 in Table 5). In the regression for the total sample the three mobile OS dummies
contributed most strongly to explaining variance in the criterion (effect size [f2] of the three dummies0.077). After
controlling for the other study variables, signicant (p o0.001) MI usage differences were found between customers with
diverging device types (see variables 68 in Table 5). More specically, a supplementary inspection of the unstandardized
regression weights of the device dummy variables (not reported in Table 5) revealed the following: iPhone 4 consumers
caused on average 79 MB monthly trafc more than iPhone 3 subscribers who in turn outperformed subjects with an
Android appliance by 52 MB per month. The latter groups monthly MI activity level was nally on average 134 MB higher
than that of MI adopters in the group with other advanced phone types. Age had the second largest impact on MI use
intensity in the full sample analysis (f2 0.020) with older consumers tending to generate less MI data volumes than their
younger counterparts. In the full sample regression the only additional predictor, which exceeded the f2-threshold of 0.01
(recommended by Cohen (1988, p. 79) for the identication of small effects) was customer tenure with an effect size of
0.011 and a negative -weight of 0.113. This implies that newly acquired customers cause more MI trafc than
subscribers who enrolled with their present MNO long ago. Each of the -estimates for screen size, MI use experience,
gender and tariff age was also statistically highly signicant (see Table 5), but the effect sizes of the predictors were
invariably below 0.01. The very small effect sizes imply that in the present full sample the impacts of these variables were
too small to be of practical importance (irrespective of their statistical signicance).
In the regression analysis for the reduced sample including only consumers with iPhones 3/4 and Android handsets the
effect size f2 of the two mobile OS/phone type dummies fell to 0.028. Android users were estimated to have a signicantly
lower MI trafc volume than both iPhone 3 and iPhone 4 users. In addition, the second analysis suggested that the
difference in average MI activity levels between the two groups with different generations of Apple phones (estimated at
11 MB per month) was not statistically signicant after partialling the remaining predictors (see footnote b in Table 5). The
results imply that the relatively high explanatory power of the mobile OS/phone-type dummies mainly stems from
consumers with other MI-enabled OS/phone types because this group generated much less MI trafc than consumers with
one of the remaining specic OS/phone types under study. The effect size of age amounted to 0.027 in the reduced sample
analysis compared to 0.020 in the rst regression. This indicates that age had a somewhat stronger negative impact on MI
use intensity among consumers with an iPhone or an Android device than among subscribers relying on other OS/phone
types in accessing the Internet. Again, the sole predictor to additionally pass an f2-level of 0.01 in the second regression,
was customer tenure (f2 0.012; 0.123). The effect sizes of screen size, MI use experience, handset type age and
gender were below this threshold. This means that they also failed to achieve practical relevance as MI use intensity
predictors in the reduced sample analysis.
5. Conclusions
The present research set out to empirically compare gender, age, three MNO relationship attributes and MI use
intensity for MI adopters in Germany who relied on different categories of advanced phones when accessing the Internet
over an MNOs radio network. It expands extant work at least in two ways. First, we applied an unobtrusive collection
procedure of real-life data. Second, we extracted an objectively measured metric from an MNOs billing engine to describe
MI use intensity and the value distribution of this variable in a large sample of 9321 MI adopters in a country, which has
rarely been studied in prior scholarly research projects.
Our study conrmed that in Germany in mid-2011 MI at rate subscribers who access the Internet via various
advanced handset types were still younger and more likely to be male than the adult population. The MI adoption divide
between young and old consumers was more pronounced than between males and females. Overall, the results suggest
that the diffusion of Internet access through advanced mobiles follows a pattern similar to that observed earlier for
Internet access over German xed line narrow- and broadband networks (Eimeren & Frees, 2011). During the study period
MI access subscription reached a diffusion status comparable to that of xed network Internet adoption among consumers
in Germany about 10 years ago.
MI adopters with an iPhone 3, iPhone 4, Android handset or another MI-enabled advanced mobile displayed only very
limited differences in terms of their gender and age distributions as well as two MNO relationship characteristics (customer
tenure, MI use experience). The only exception was the time span elapsed since consumers had subscribed to their current data
at price scheme (tariff age): MI adopters with an iPhone 3 or another MI-enabled advanced phone type had a signicantly
higher tariff age than their counterparts equipped with an iPhone 4 or Android device. This should not come as a surprise since
sales of the iPhone 4 had started later in the summer of 2010 (about two years after the iPhone 3 introduction) and a broad
range of powerful Android devices has been launched on the market only recently. The moderate gender and age differences
between iPhone 3 and 4 users and consumers running an Android handset or other MI-enabled phone indicate that on average
users of the second and third generation of Apple phones today are no longer a special breed of consumers with regard to their

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

369

demographics. Besides, the small divergences between iPhone 3 and iPhone 4 owners conrm observations of Arruda-Filho and
Lennon (2011) according to which the iPhone 4 was not perceived as a revolutionary offer attracting once more mainly the
same type of Apple enthusiasts who had previously bought an iPhone 3.
Clear divergences with regard to consumer MI use intensity were detected depending on a subscribers phone type.
After controlling for effects of gender, age, three MNO relationship as well as two other device characteristics consumers
with an iPhone 3 or iPhone 4 displayed a signicantly higher MI use intensity than subscribers equipped with an Android
appliance. The latter group of subjects in turn generated signicantly more MI trafc than customers relying on an
advanced phone with a mobile OS other than Apples iOS or Googles Android.
Our ndings do not only provide a previously unavailable picture of personal characteristics and MI usage activity
levels of German MNO subscribers with four different advanced phone types. They also have managerial ramications for
MNO. In practice, they imply that MNO should reect on the low adoption rate of advanced MI enabled phones among
older people, particularly those aged 50 and above. In the present sample just 11.8% of the MI adopters fell into this age
range compared to 47.9% of the adult population in Germany (see reference, Table 2). Hence, MNO could devote more
resources to develop messages for promotion campaigns and supplementary handset application programs, which meet
special needs of this age group (e.g., mobile medical or physical tness apps).
Extending Verkasalo (2008b) and Gerpott (2010) our study showed that MI use activity level was again highly diverse
among MI adopters across various advanced phone types. In the present sample, even the choice of a at MI rate plan did
not necessarily imply that an MNO customer always mutated into an excessive MI user. This holds in particular for
subscribers who do not use an iPhone 3/4 or an Android handheld. In fact, our analysis revealed that in the subsample of
consumers with other MI-enabled phones 50% (75%) of its members did not generate a monthly MI trafc volume of more
than 10 (64) MB (see reference, Table 4). This means that these customers exhausted their data at rate only to a very
small degree. From an MNO perspective, such a nding may be characterized as a mixed blessing. On the one hand it
implies that an operator has attracted a customer group paying a considerable monthly price for a at data plan, while at
the same time generating very sparse MI trafc. Hence, this group is likely to be very protable for an MNO at least in the
short run.
On the other hand, several studies (Iyengar, Ansari, & Gupta, 2007; Wong, 2010) suggest that an economic mist
between MNO subscribers rate plans and their mobile communication usage patterns erodes customer satisfaction and
promotes customer churn. Therefore, MNO should consider to take steps which narrow the gap between the relatively
high monthly data at rate and MI usage especially in the group of consumers equipped with an advanced phone type not
running on iOS or Android. Based on our ndings MNO could address this target group with subsidized offers of attractive
state-of-the-art Android phone models because ownership of this phone type could increase MI activity levels. In contrast,
an active migration policy for customers in the other phone type group, which entails the shift to an iPhone, should be less
protable for an MNO for two reasons. First, the hardware subsidy an MNO has to carry when bundling Android phones
with a data at rate contract is lower than the one resulting from marketing iPhones. Second, iPhone customers consume
signicantly more mobile data network capacity than subscribers operating an Android phone.
The strongly positive skewness of the MI use intensity metric found across all four mobile types suggests that fair use
policies recently introduced by various MNO are justied although such policies may be criticized by consumer advocates.
Such policies can take two main avenues (Harno, 2010). First, MI data transfer speed is reduced once a customers monthly
mobile IP trafc volume exceeds a threshold. Second, the price plan is not a true at rate but it entails a data volume
ceiling beyond which the consumer has to pay a use-dependent charge mostly per MB of additional trafc. Given that
there exists a small group of MI users with advanced mobile phones who consume a very large share of the data
transmission capacity available in enhanced 2G or 3/4G radio access networks a waiver of fair use rules would mean that
the very large majority of an MNOs MI subscribers are likely to suffer from a severe quality of service degradation caused
by a tiny number of power users. Thus, in line with Harno (2010) we conclude MNO have no other choice but to avoid the
occurrence of negative externalities triggered by this user group through the imposition of fair use standards.

6. Limitations
Like most empirical work, the present study is not without limitations from which directions for further scholarly efforts can
be derived. A rst shortcoming is that the sample consisted of data at rate subscribers of one MNO in Germany only. The
extent to which this affects the generalizability of the ndings to consumers or business customers with deviating usedependent price schemes of other MNO in Germany or outside of the German market is uncertain and requires further
investigation. A second shortcoming stems from the cross-sectional nature of the data collection mainly in May and June 2011.
This makes it difcult to draw rm conclusions with regard to changes in MI use behaviors after consumers had started to
operate a specic phone type. Therefore, additional research should gather individual level MI usage data at several points in
time to more convincingly corroborate the view that it is the phone type which drives MI use intensity and not predominantly a
(xed) level of MI usage activity causing the selection of specic types of mobiles by consumers.
A third deciency is that the customer records of the cooperating MNO contained only a few plain demographic but no
psychographic characteristics (e.g., self-assessment of MI literacy, perceptions of MI use motives, personal innovativeness).
It is obvious that a better understanding of mobile data services adoption can be achieved by segmenting MNO customers

370

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

both by demographic and psychographic variables. Hence, more work is required which explores both classes of personal
characteristics among MI consumers with distinct categories of advanced MI-enabled phones.
A fourth shortcoming relates to the system-captured operationalization of MI use intensity. Measuring use intensity on the
basis of monthly mobile IP trafc makes it difcult to infer the context surrounding customers MI uses. For instance, IP trafc
may be caused by down- or uploading software, music or games, by exchanging instant messages or emails, or by WWWbrowsing (Verkasalo, 2011). The lack of differentiation between various MI application types is particularly unfortunate
because researchers (e.g., Verkasalo et al., 2010) have demonstrated the advantages of exploring MI use at the level of the
individual application category instead of MI use in general. Furthermore, advanced handsets themselves generate some IP
trafc, which is not under direct user control (so-called idle trafc) and might be excluded from use intensity operationalizations. Thus, further investigations are desirable to develop more ne-grained measures of use intensities of various MI
application types, which are extractable from MNOs billing and customer administration systems.
Finally, our reliance on archival customer data and a system-captured use intensity measure has its strengths (e.g.,
avoidance of spurious correlations caused by common method bias). However, this approach also has its drawbacks. It
allows no profound insights into subjective customer experiences when accessing the Internet through an advanced
mobile. Such insights also require reactive qualitative and quantitative data collection strategies capturing subjective
(stated) customer views through well-established research instruments as for instance questionnaires, personal
interviews or focus group discussions. Hence, future scholarly work on MI adoption and use intensity should move
toward integrating MI-related customer perceptions with measures of specic mobile device features in explaining both
objectively revealed and subjectively claimed usage behaviors of MI adopters.
References
Accenture, K. K. (2011). Mobile Web Watch-Studie 2011Die Chancen der mobilen Evolution. Kronberg: Accenture.
Amethon. (2008). IPhones dominate Australian mobile data consumption according to rst comprehensive mobile Internet report. Retrieved from: /http://
www.amethon.com/content_common/pg-200809-australian-mobile-internet-insights-report-.seoS.
Armey, L., Vladar, A., & Pereira, F. (2011). Are phones fun? Usage and barriers to adoption for wireless data services among the younger generation of
Asians. International Journal of Electronic Business, 9(12), 825.
Arruda-Filho, E. J., Cabusas, J. A., & Dholakia, N. (2010). Social behavior and brand devotion among iPhone innovators. International Journal of Information
Management, 30(6), 475480.
Arruda-Filho, E. J., & Lennon, M. M. (2011). How iPhone innovators changed their consumption in iDay2: Hedonic post or brand devotion. International
Journal of Information Management, 31(6), 524532.
Bao, P., Pierce, J., Whittaker, S., & Zhai, S. (2011). Smart phone use by non-mobile business users. In: M. Bylund, O. Juhlin, & Y. Fernaeus (Eds.), Proceedings
of the 13th International Conference on Human Computer Interaction with Mobile Devices and Services (Mobile HCI 11) (pp. 445454). New York: ACM.
Bitkom (2012). Jeder dritte geht per handy oder tablet ins internet. Retrieved from: /http://www.bitkom.org/les/documents/BITKOM_Presseinfo_Me
diennutzung_03_04_2012.pdfS.

Elektrizitat,
Gas, Telekommunikation, Post und
Bundesnetzagentur (2011). Tatigkeitsbericht
2010/11 Telekommunikation. Bonn: Bundesnetzagentur fur
Eisenbahnen. Retrieved from: /http://www.bundesnetzagentur.de/SharedDocs/Downloads/DE/BNetzA/Presse/Berichte/2011/TaetigkeitsberichtTK
20102011pdf.pdf?__blob=publicationFileS.
Campbell, P. R., & Ahmed, F. (2011). An assessment of mobile OS-centric ecosystems. Journal of Theoretical and Applied Electronic Commerce Research, 6(2),
5062.
Chae, M., & Kim, J. (2004). Do size and structure matter to mobile users? An empirical study of the effects of screen size, information structure, and task
complexity on user activities with standard web phones. Behaviour & Information Technology, 23(3), 165181.
Choi, H., Kim, Y., & Kim, J. (2011). Driving factors of post adoption behavior in mobile data services. Journal of Business Research, 64(11), 12121217.
Cisco (2011). Cisco Visual Networking Index: Forecast and Methodology, 20102015. San Jose. Retrieved from: /http://www.cisco.com/en/US/solutions/
collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-481360.pdfS.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences ((2nd ed.). Hillsdale: Erlbaum.
Eimeren, van B., & Frees, B. (2011). Drei von vier Deutschen im NetzEin Ende des digitalen Grabens in Sicht? Media Perspektiven, 42(78), 334349.
Ericsson (2011). Trafc and market data report. Stockholm. Retrieved from: /http://hugin.info/1061/R/1561267/483187.pdfS.
Gerpott, T. J. (2010). Communication behaviors and perceptions of mobile internet adopters. Info, 12(4), 5473.
Gerpott, T. J. (2011). Determinants of self-report and system-captured measures of mobile Internet use intensity. Information Systems Frontiers, 13(4),
561578.
Ghose, A., & Han, S. P. (2011). An empirical analysis of user content generation and usage behavior on the mobile Internet. Management Science, 57(9),
16711691.
Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., & Tatham, R. L. (2006). Multivariate data analysis ((6th ed.). Upper Saddle River: Pearson Prentice Hall.
Harno, J. (2010). Impact of 3G and beyond technology development and pricing on mobile data service provisioning, usage and diffusion. Telematics and
Informatics, 27(3), 269282.
Haverila, M. J. (2011). Cell phone feature preferences and gender differences among college students. International Journal of Mobile Communications, 9(4),
401419.
Hong, S.-J., & Tam, K. Y. (2006). Understanding the adoption of multipurpose information appliances: The case of mobile data services. Information
Systems Research, 17(2), 162179.
Iyengar, R., Ansari, A., & Gupta, S. (2007). A model of consumer learning for service quality and usage. Journal of Marketing Research, 44(4), 529544.
Jiang, P. (2008). Adopting mobile internet: A demographic and usage perspective. International Journal of Electronic Business, 6(3), 232260.
Kenney, M., & Pon, B. (2011). Structuring the smartphone industry: Is the mobile Internet OS platform the key? Journal of Industry Competition and Trade,
11(3), 239261.
Kim, B. (2012). The diffusion of mobile data services and applications: Exploring the role of habit and its antecedents. Telecommunications Policy, 36(1), 6981.
Kim, B., & Oh, J. (2011). The difference of determinants of acceptance and continuance of mobile data services: A value perspective. Expert Systems with
Applications, 38(3), 17981804.
Kim, K.J., Sundar, S.S., & Park, E. (2011). The effects of screen-size and communication modality on psychology of mobile device users. In: D.S. Tan,
S. Amerski, B. Begole, W.A. Kellogg, & M. Tungare. (Eds.), Proceedings of the international conference on human factors in computing systems (CHI 2011),
Extended Abstract Volume (pp. 12071212). New York: ACM.
Kim, S., & Garrison, G. (2009). Investigating mobile wireless technology adoption: An extension of the technology acceptance model. Information Systems
Frontiers, 11(3), 323333.

T.J. Gerpott et al. / Telecommunications Policy 37 (2013) 357371

371

Kim, S. S., & Malhotra, N. K. (2005). Predicting system usage from intention and past use: Scale issues in the predictors. Decision Sciences, 36(1), 187196.
Legris, P., Ingham, J., & Collerette, P. (2003). Why do people use information technology? A critical review of the technology acceptance model.
Information & Management, 40(3), 191204.
Lin, F., & Ye, W. (2009). Operating system battle in the ecosystem of smartphone industry. In: V. Muhin, & Y. Zhiwei (Eds.), Proceedings of the rst
international symposium on information engineering and electronic commerce (IEEC 09) (pp. 617621). Los Alamitos: IEEE.

Ling, R., & Sundsy, P. R. (2010). The iPhone and mobile access to the internet. In: J. R. Hoich,
G. F. Kirchner, C. Linke, & I. Schlote (Eds.), mobile media and
the change of everyday life (pp. 213223). Frankfurt/M: Peter Lang.
Malykhina, E. (2007). Analysis: How smartphone platforms compare. Retrieved from: /http://www.informationweek.com/news/196902226S.
Matthews, T., Pierce, J., & Tang, J. (2009). No smart phone is an Island: The impact of places, situations, and other devices on smart phone use. IBM research
report RJ 10452 (A0909-003), September 11. San Jose.

Muller,
R. M., Kijl, B., & Martens, K. J. (2011). A comparison of inter-organizational business models of mobile app stores: There is more than open vs.
closed. Journal of Theoretical and Applied Electronic Commerce Research, 6(2), 6376.
ODoherty, K., Hill, S. R., Mackay, M. M., & McPherson, J. (2010). Mobile data service usage and preference: An investigation of Australian consumers.
International Journal of Mobile Communications, 8(1), 106127.
Okazaki, S., & Hirose, M. (2009). Does gender affect media choice in travel information search? On the use of mobile Internet. Tourism Management, 30(6),
794804.
Ovum, M. (2011). Smartphone forecast 201116 (March 2011). London: Ovum.
Peslak, A., Shannon, L.-J., & Ceccucci, W. (2011). An empirical study of cell phone and smartphone usage. Issues in Information Systems, 12(1), 407417.
Pousttchi, K., & Goeke, L. (2011). Determinants of customer acceptance for mobile data services: An empirical analysis with formative constructs.
International Journal of Electronic Business, 9(1-2), 2643.
Serif, T., & Ghinea, G. (2008). Mobile information access in the real world: A story of three wireless devices. Computers in Human Behavior, 24(4),
13851403.
Sharma, R., Yetton, P., & Crawford, J. (2009). Estimating the effect of common method variance: The methodmethod pair technique with an illustration
from TAM research. MIS Quarterly, 33(3), 473490.
Shepard, C., Rahmati, A., Tossell, C., Zhong, L., & Kortum, P. (2010). LiveLab: Measuring wireless networks and smartphone users in the eld. Performance
Evaluation Review, 38(3), 1520.

Smura, T., Kivi, A., & Toyli,


J. (2009). A framework for analysing the usage of mobile services. Info, 11(4), 5367.
Son, M., & Han, K. (2011). Beyond the technology adoption: Technology readiness effects on post-adoption behavior. Journal of Business Research, 64(11),
11781182.
Spehr, M. (2010). Schon wieder ein Kultobjekt aus Cupertino. In: Frankfurter Allgemeine Zeitung, 24 June 2010. Retrieved from: /http://www.faz.net/aktuell/
wirtschaft/unternehmen/apples-iphone-4-im-test-schon-wieder-ein-kultobjekt-aus-cupertino-1621336.html?selectedTab=comments&tabCounter=
3&showMarginalSlot=1S.

die Bundesrepublik Deutschland mit internationalen Ubersichten.


Statistisches Bundesamt (2011). Statistisches Jahrbuch 2011 fur
Wiesbaden:
Statistisches Bundesamt.
Strategy Analytics (2011). Global smartphone sales forecast by country: EMEA (September 2011 Datasheets). Boston: Strategy Analytics.
Sugai, P. (2007). Exploring the impact of handset upgrades on mobile content and service usage. International Journal of Mobile Communications, 5(3),
281299.
Szajna, B. (1996). Empirical evaluation of the revised technology acceptance model. Management Science, 42(1), 8592.
Tojib, D., & Tsarenko, Y. (2011). Post adoption modeling of advanced mobile service use. Journal of Business Research, in press (http://dx.doi.org/10.1016/
j.jbusres.2011.05.006).
Tseng, F.-M., & Lo, H.-Y. (2011). Antecedents of consumers intentions to upgrade their mobile phones. Telecommunications Policy, 35(1), 7486.
Turner, M., Kitchenham, B., Brereton, P., Charters, S., & Budgen, D. (2010). Does the technology acceptance model predict actual use? A systematic
literature review. Information and Software Technology, 52(5), 463479.
Verkasalo, H. (2008a). Dynamics of mobile service adoption. International Journal of E-Business Research, 4(3), 4063.
Verkasalo, H. (2008b). Handset-based measurement of mobile service demand and value. Info, 10(3), 5169.
Verkasalo, H. (2009). Analysis of mobile internet usage among early-adopters. Info, 11(4), 6892.
Verkasalo, H. (2011). An international study of smartphone usage. International Journal of Electronic Business, 9(1-2), 158181.
Verkasalo, H., Lopez-Nicolas, C., Molina-Castillo, F. J., & Bouwman, H. (2010). Analysis of users and non-users of smartphone applications. Telematics and
Informatics, 27(3), 242255.
West, J., & Mace, M. (2010). Browsing as the killer app: Explaining the rapid success of Apples iPhone. Telecommunications Policy, 34(45), 270286.
Wong, K. K. (2010). Fighting churn with rate plan right-sizing: a customer retention strategy for the wireless communications industry. Service Industries
Journal, 30(13), 22612271.

You might also like