Consumer Perception of Mobile Telephony Tariffs with Cost Caps

Jan Kr mer a
Karlsruhe Institute of Technology Karlsruhe, Germany Email: kraemer@kit.edu

Lukas Wiewiorra
Karlsruhe Institute of Technology Karlsruhe, Germany Email: wiewiorra@kit.edu

AbstractIn this article we provide a seminal academic investigation of mobile telephony consumers perception of the recently introduced cost cap tariff in comparison to corresponding payper-use and atrate calling plans. Previous studies have identied several psychological effects through which consumers are believed to be biased towards atrate plans. However, atrate plans also limit customers tariff exibility in months of low usage. Cost cap tariffs are a hybrid between pay-per-use and atrate plans and can offer cost insurance while maintaining exibility. In particular, we provide evidence from a survey among 214 German university students that exibility, cost insurance and the so-called taximeter effect are the main drivers behind cost cap tariff choice. Furthermore, we show that the insurance and taximeter effects are distinct if evaluated within a atrate or cost cap tariff framework. Finally, based on our results we also comment on the protability of cost cap plans from a strategic management perspective.

I. I NTRODUCTION Competition in the mobile telephony market is erce and the scope for product differentiation is limited - a combination that leads mobile telephony providers to engage in price differentiation instead. Indeed, the actual extend of price differentiation in the market is striking and customers are faced with a multitude of different tariff types [1][3]. Traditionally, mobile telephony providers have only offered two-part tariffs, which consist of xed monthly fee and a usage-based component [4]. Two-part tariffs have a long tradition in telephony [5]. Their popularity stems from the fact a large portion of the providers costs are xed costs related to the build up, maintenance and operation of the network infrastructure, which can be recovered by the xed component of the two-part tariff [6]. On the contrary, the marginal costs of calling, which are constituted through switching or interconnection fees, for example, are relatively small and can be recovered by the variable component of the two-part tariff. In fact, if the telephony provider holds a monopoly, which has been the case for the xed-line providers for most of the twentieth century, two-part tariffs can even achieve to extract customer rent completely, comparably to perfect price discrimination [7]. However, in the mobile telephony market, which has been liberalized much earlier than the xed telephony market, actual infrastructure-based competition has created a very competitive market environment. In an effort to differentiate their service offering, providers have experimented with new

and innovative tariff structures to the extend that the mobile telephony market is ooded with a seemingly endless variety of voice telephony tariffs today. Besides the classical two-part tariff, the tariff portfolio ranges from three-part tariffs that include a usage allowance or minimum turnover requirements to one-part tariffs, i.e. pure pay-per-use or atrates tariffs. In this paper we highlight the innovative cost cap tariff which has been introduced to the German mobile market in 2009 by o2 (Telefonica).1 The cost cap tariff is a new type of two-part tariff that constitutes a hybrid between a pay-per-use and a atrate tariff. It is a pure pay-per-use tariff until the total costs exceed a predened cost cap, at which the tariff effectively becomes a atrate. o2 claims that the cost cap tariff is a great success and has signicantly contributed to attracting new customers [8]. In this article we provide a seminal academic investigation of mobile telephony consumers perception of cost cap tariffs based on a survey among 214 German university students. In particular, it seems obvious that consumers will perceive the cost cap tariff as weakly dominant over both the pay-per-use tariff (because it offers a cost insurance in months of high usage) as well as the atrate plan (because it offers exibility in months of low usage) if the cost cap tariff offers the same per-minute prices as a pay-per-use tariff and if the cost cap is set at the price of a corresponding atrate plan. On the other hand, it is questionable to which degree customers of payper-use plans would in fact accept a higher per-minute price under a cost cap tariff in order to be insured against high monthly invoices.2 Likewise, previous studies (e.g. [9][14]) have provided strong empirical evidence that customers have a bias towards atrate plans due to psychological reasons. In the following we will also investigate to what extend the cost cap tariff is also capable to benet from these psychological effects compared to pure atrate tariffs while additionally capturing those customers that have a strong desire for exible pay-peruse tariffs. The remainder of this article is structured as follows: In Section II we introduce the psychological effects that have
1 The tariff, which is called o2o, has ofcially been released on May 5, 2009. 2 In the o2o cost cap tariff customers pay 15 Euro cents per minute, while comparable pay-per-use plans are currently available from as low as 7.5 cents per minute.

978-1-4244-7986-3/10/$26.00 2010 IEEE

been identied as cognitive causes of a atrate bias and discuss them in the context of cost cap tariffs. Section III derives our research questions and IV presents the set up and content of the survey. In Section V, nally, we will present and discuss the results of our statistical analysis and conclude in Section VI by commenting on the results with respect to customers perception of cost cap tariffs.

and perhaps more complex tariff.4 The costs associated with a atrate, however, are easily accessible and independent of ones own (possibly unknown) calling volume. c) Overestimation: Consumers often fail to anticipate their own usage of telecommunication services correctly [16]; either because of limited foresight and uncertainty in demand, or because of bounded rationality. In anticipating their current or future service usage, consumers have a tendency to overestimate their actual demand, which, in turn, will induce them to select a atrate plan more often. At the same time, users are often overly condent that their usage estimation is correct [16]. This makes the overestimation effect distinct from the insurance effect and leads consumers to repeatedly choose the wrong calling plan. d) Taximeter: The so-called taximeter effect is derived from the experience of using a taxi in a foreign city: The taxi ride can be perceived as unpleasant, because the running taximeter constantly visualizes the currently accumulated costs. In the telecommunications domain, the analogous effect can for example be experienced while waiting in the queue of a very costly hotline. Under a atrate plan, on the contrary, all costs are sunk ex ante and the marginal costs of consumption are zero. This prevents consumers from experiencing the taximeter effect, who may thus favor atrate plans. However, there is also a distinctive reason for customers not to choose a atrate plan. e) Flexibility: By choosing a atrate, consumers commit themselves to pay a xed amount for telephone usage in each billing period, independent of their actual usage. This potential discrepancy between actual usage and committed costs can result in two adverse effects, which we have summarized under the roof of exibility. First, subscribers may ex post regret their cost commitment under a atrate plan after having realized that a usage-based tariff would have generated lower costs after all. Pay-per-use tariffs, on the contrary, avoid such commitments and react exibly to (exceptionally low) usage. Secondly, the cost commitment may tempt atrate subscribers to excess telephone usage. In other words, consumers may seek to avoid the ex post regret that they have not chosen a less expensive pay-per-use plan and therefore exploit the atrate by telephoning more than reasonable. This effect is comparable to the so-called buffet effect by which customers of all-youcan-eat buffets eat more than they desire in an effort to justify the break even costs of an all you can eat buffet in comparison to the la carte menu [17]. Consequently, consumers with a strong aversion against ex post regret or the buffet effect are more likely to opt for a exible pay-per-use tariff and seek to

II. R EASONS FOR B IASED TARIFF C HOICE Economic theory assumes that consumers maximize their utility when selecting a mobile tariff, such that the chosen calling plan should (on average) lead to the lowest possible telephone bill, given correct anticipation of the average usage. However, several studies (e.g. [12], [14][16]) have provided strong evidence that customers actual tariff choice may deviate from the economic prediction due to motivational, psychological or cognitive causes that lie outside of the standard economic framework of fully rational market participants with perfect foresight. The most prominent example of such systematic deviations is the the so-called atrate bias, which has been shown in context with all-you-can-eat buffets [17], gym subscriptions [9], Internet access plans [10][12] and mobile calling plans [13], [14]. The atrate bias asserts that consumers often favor atrate tariffs, although a pay-per-use tariff would be cheaper with regard to their actual usage volume. From psychological perspective, the main difference between a pay-per-use (or cost cap tariff for that matter) and a atrate tariff is that telephone costs are sunk in a atrate plan and consequently customers need not worry about the costs of current or future usage.3 More specically, with respect to information and communication services, the literature has identied four distinctive effects that are supposed to drive the atrate bias: a) Insurance: The insurance effect aims at customers that want to insure themselves against exceptionally (and unexpectedly) high telephony bills. Customers may be risk averse towards the correctness of their estimation of current service usage or with respect to future changes in demand. Therefore these customers are willing to pay (on average) more for a atrate plan than what the costs of a corresponding payper-use tariff would be, because they are insured against these risks. In this vein the difference between the costs for a atrate and the average costs for a pay-per-use tariff can be interpreted as an insurance premium. Indeed, [11] show that the option value attached to atrate plans is independent of the actual usage. b) Convenience: As outlined before, consumers face a tremendous variety of alternative tariffs in the mobile telephony markets. The convenience effect is driven by those consumers who want to avoid the effort of nding a cheaper
3 This corresponding psychological concept is known as mental accounting and has been investigated in a number of different domains ( [18][20]).

4 Network providers sometimes even try to leverage the potential of convenient consumers and engage in so called foggy pricing. In its most extreme form, service providers deliberately offer dominated calling plans, because the additional costs of doing so are very low and there is a positive probability that convenient or uninformed consumers will actually choose this tariff. For a critical analysis of foggy pricing see [3].

avoid the cost commitment associated with a atrate plan.5 III. R ESEARCH Q UESTIONS The main research question we seek to investigate in this article is whether a cost cap tariff can also tap on one or some of these effects and if so, how strong the impact on customer choice will be. Due to its hybrid nature, we can presume that the cost cap tariff can capture some of the atrate bias components, namely the insurance and overestimation effects, as well as the pay-per-use component, namely the exibility effect. At this point, it remains open, however, how consumers will evaluate the cost cap tariff compared to a atrate plan with respect to the convenience and taximeter effect. Table I summarizes the potential impact of the identied effects on the three fundamental tariff schemes.
TABLE I P OTENTIAL EFFECTS ON THE CHOICE OF DIFFERENT TARIFF SCHEMES Pay-per-use (PPU) + Cost Cap (CC) + ? + ? + Flatrate (FR) + + + + -

a new way to test the taximeter effect independent of the insurance effect.

IV. S URVEY D ESIGN The survey is structured in four distinctive parts. In the rst part, participants are asked to decide between alternative tariffs in two series of hypothetical usage scenarios. More precisely, participants are presented with several ctitious usage scenarios in which they have to indicate which of two alternative calling plans they would choose. This methodological approach is sometimes called quasi-experimental [12]. The rst part of the survey is subdivided into two series of choice scenarios. In the rst series of four usage scenarios participants had to decide between a atrate plan and a pure pay-per-use tariff. Usage scenarios differed only by the induced minimum and maximum usage (in minutes), whereas the average usage was identical in all scenarios.8 Hence, the setup measures if different perceived extrema of usage impact tariff choice.9 Moreover, it is emphasized that we have chosen realistic choice scenarios, which are in line with average usage volumes of mobile telephony and actual market prices in Germany at the time of the survey. For the rst series of usage scenarios, respondents are randomly assigned to one of two settings. In the rst setting, the usage scenarios are such that the costs of the atrate plan are identical to the average monthly costs under the pay-peruse plan (45 e). Thus, we would expect to observe an equal distribution of participants between the two alternative calling plans. In the second setting, the atrate is priced slightly higher 49 e). With this modied version of the question it can be veried if a potential bias towards a atrate is also present under a pricing scheme with higher than average monthly costs. In a second series of four similar usage scenarios, all respondents were asked to choose between a cost cap tariff and a atrate tariff. The usage scenarios were designed such that it is possible to check for a atrate bias over a cost cap plan. In the second part of the survey participants are presented with several verbal phrases, who are then asked to indicate their consent on a ve point Likert scale. These phrases are so called indicators which are later aggregated to constructs that measure the effects in table I. Most of the indicators
8 Analogous to the survey design of [12], [21] minimum and maximum usage is varied in all possible combinations of 33. and 100% deviation 3% of the average costs per month. The corresponding min/max usage levels are 0/400, 200/400, 0/600 and 200/600 minutes. 9 The hypothetical questions were presented in the following manner: Please imagine independently from your real usage behavior that you are using your mobile phone on average 300 minutes per month to call into German networks and you have to pay the bill yourself. Your amount of usage uctuates between the months. Below you are presented with different values of minimum and maximum usage per month. Average usage is always 300 minutes per month. Please indicate in all four cases your choice between tariff 1 where you pay 15 Euro cents per minute and tariff 2 where you pay 45 Euros at.

Insurance Convenience Overestimation Taximeter Flexibility

In the seminal analysis, it is important to isolate the psychological effects that are exhibited by a certain tariff scheme from the underlying economic effects. Therefore, in the following the cost cap plan is compared to a pay-per-use plan with the same per-minute price and with a atrate plan that is priced at the level of the cost cap. In this way, the cost cap plan weakly dominates both the pay-per-use and the atrate plan and hence tariff biases can be easily identied.6 Furthermore, recall that the pay-per-use and the atrate calling plan are both a one-part tariff, which put together, constitute the two-part cost cap tariff. These tariffs are not directly comparable to three-part tariffs (that consist of a xed fee, an included allowance of minutes for which the marginal prize is zero, and a positive marginal price for additional minutes beyond the allowance), and will therefore not be considered here.7 In the following section we present the design of a survey that is thought to quantify tariff bias effects and their impact on consumers choice of pay-per-use, cost cap or atrate calling plans. In addition, in contrast to previous studies on the atrate bias, our analysis of the cost cap pricing scheme also offers
5 It shall be annotated that in different domains, where excessive usage is a desired outcome to the consumer, the lack of exibility may even constitute a atrate bias. With a atrate plan, customers can strategically commit themselves to more service usage for the same reasons as above. For example, people often subscribe to a long term atrate contract for health clubs in order to commit themselves to use the gym more often. In the literature this reverse interpretation of the exibility effect is called self-discipline effect (cp. e.g. [9]). 6 See Section VI for ideas on possible alternative choice scenarios that should be subject to future research. 7 Three-part tariffs are for example very popular in the US cellular phone services market.

have been adapted from [21] and [15].10 However, we derived our own indicators with respect to those constructs that were thought to measure the taximeter and insurance effect under a cost cap plan as well as the exibility effect. Additionally, in this part of the survey respondents were asked for their appraisal of their own usage behavior and costs. To exclude unwanted effects stemming from the order of the items we presented them in random order to all participants. In the third part of the survey, data on the respondents actual mobile telephony tariff characteristics (such as contract duration, tariff model, average monthly bill is collected. Participants were notied that from this point on, they were asked about their real actual mobile phone usage. In the last and fourth part of the survey, participants were confronted with a standard risk aversion test based on [22] in which respondents must choose between a safe and a risky lottery in ten settings. Participants were incentivized to answer truthfully in the risk aversion test by offering the chance for a gift coupon of up to 38.5 e value. The determination of the actual value of the coupon is perfectly aligned with the decisions in the risk aversion test and thus, the test was implemented as a non-hypothetical setting. More precisely, to determine the coupon value, one of the ten settings was chosen in which the respondents participated in the respective lottery of his choice. Lottery outcomes were also orientated on the average mobile phone bill of participants. Since our sample is very homogeneous we can on average neglect income effects inuencing the results. The survey was conducted in February 2010 through a web-based questionnaire among undergraduate students at the Karlsruhe Institute of Technology in Germany. Students were invited to the questionnaire via the mailing list of an introductory economics course. We stimulated participation by offering a chance to take part in a lottery of four 25 e gift coupons for a large online retail store after completing the survey. 96.7% of the participants decided to participate in the lottery. Additionally one gift coupon was awarded in the risk aversion test. Certainly, we do not claim that students of economics are a representative panel for the German or European mobile telephony subscriber. However, we can presume that students are more rational and more aware of their mobile tariff and the associated costs than the average subscriber. Thus, if we nd empirical evidence for systematic selection biases in our panel, it is likely that the actual effects are in fact even stronger. We will discuss this issue in more detail in the following sections. V. R ESULTS A total of 214 respondents (23.5% female, 76.5% male) with an average age of 20.5 years (S=1.47; min=16; max=27;
10 In particular, [21] adapted and validated well known scales from marketing science to measure these effects (latent variables in a statistical model) inuencing the decision in favor for a atrate tariff. Even though the original scale was constructed to test for the effects in connection with Internet access prices we modied the wording only if necessary.

median=20) completed our survey successfully.11 All of the respondents are mobile telephony customers. The mean of the reported average monthly telephone bill in our sample is 19.14 e (S=17.53), with a noticeable discrepancy between the mean average monthly bill of the 46.73% with prepaid tariffs (M=10.15 e; S=9.37) compared to those with postpaid tariffs (M=27.03 e; S=19.19). Thus, our sample contains about an equal split of users with low and comparably high mobile telephone usage. At rst, we investigate the degree to which the participants in our sample are biased towards choosing a atrate plan over a pay-per-use plan. Table II summarizes the results. Participants were randomly assigned to one of two settings with different atrate prices, in which they were asked to state their preference for the atrate or a pay-per-use plan in four usage scenarios. In the rst setting, which was assigned to 52.3% of the participants, the atrate was priced at 45 e the costs for the alternative pay-per-use plan under average usage. The remaining 47.7% of the respondents evaluated the second setting, in which the usage scenarios are identical, but the atrate is priced slightly above the average costs for the payper-use plan.12
TABLE II C HOICE OF FLATRATE OVER PAY- PER - USE PLAN Usage (minutes) Flatrate choice at Min Avg. Max 45 e 49 e 0 300 400 18.63% 4.46% 200 300 400 46.08% 25.00% 0 300 600 41.96% 37.50% 200 300 600 83.33% 66.96% compared to a pay-per-use plan at 15 cents/minute

In both settings, except for the scenario with high minimum and maximum usage, respondents favored a pay-per-use plan over the atrate. Previous studies have argued that in each scenario in which the atrate is priced at average costs unbiased participants would be indifferent between choosing the atrate and the pay-per-use tariff [12], [14]. Therefore, one would expected the atrate to be chosen in 50% of the cases. In a qualitatively identical setting, [12] and [14] nd that respondents chose the atrate in three out of four scenarios at average cost pricing and in two out of four scenarios at slightly above average cost pricing. However, in our panel we nd to the contrary, that participants prefer the pay-per-use tariff in three out of four scenarios in both settings. Thus, if there is a bias in our sample, it would rather run in favor of a pay-peruse tariff.
11 A total of 346 Students followed the link and 247 of them completed the survey. In a nal validity question 7 participants indicated to have answered incorrectly in the survey and were eliminated from the nal dataset. Moreover, in order to increase the internal validity of the responses, another 17 observations were omitted from the analysis due to obviously random and irrational choices in the risk aversion test. Since there are no categorical wrong answers in the remaining parts of the survey, we generalized the validity of the answers in the risk aversion test to the other survey questions. However, our reported results do not signicantly vary if these observations are included. 12 We decided to present a price of 49 e to them.

We offer two alternative explanations for this nding. First, as argued above, the respondents in our sample are students of economics and therefore (i) aware of rational choices and expected utility theory and (ii) sensitive to costs in general. Moreover, the participants in our survey are also very aware of the costs of using their mobile phone. In the items that capture cost control (cp. Table III) respondents agree to a high degree (M=4.21; S=0.72) on a 5-point Likert scale. Second, although the relation between the usage scenarios is identical to those in [12] and [14] we use actual market prices.13 In [14], for example, which also considers mobile telephony tariffs and builds on survey data from July 2006, the atrate is priced unrealistically low at 30 e and 33 e while the pay-per-use tariff costs 30 cents per minute.
TABLE III S ELF REPORTED COST AWARENESS Cost Control (Cronbachss = 0.71, M=4.21, S=0.72) I can accurately assess my own mobile telephony usage behavior (M=4.12; S=0.88; median=4) I know my calling plan and all costs involved very well. (M=4.43; S=0.90; median=5) I can accurately assess my own mobile telephony costs (M=4.07; S=0.92; median=4)

cap tariff. Even in those scenarios in which the cost cap tariff has no advantage, more than 80% of the participants would still opt for this tariff. This result reveals the great attractiveness cost cap tariffs seem to have on mobile telephony customers, even in those settings where the cost cap tariff is factually identical to the atrate plan.
TABLE V C ONSTRUCTS CAPTURING FLATRATE EFFECTS Taximeter (FR) (Cronbachss = 0.81, M=3.67, S=1.00) A atrate is good because I dont have to think about the costs. (M=3.73; S=1.29; median=4) I enjoy using the phone less if costs increase every minute of use (M=3.79; S=1.24; median=4) Only if I have a atrate I enjoy using the mobile phone (M=3.00; S=1.39; median=3) If I have a atrate I use my mobile phone in a more relaxed and unselfconscious manner. (M=4.16; S=1.11; median=5) Insurance (FR) (Cronbachss = 0.76, M=2.63, S=1.05) To be sure my costs of mobile telecommunication will never be higher than a certain predened amount Im willing to pay a little bit more than average (M=2.52; S=1.13; median=2) Even if a at rate is somewhat more expensive than a pay-peruse rate Im happy because my costs wont exceed the xed amount. (M=2.75; S=1.21; median=3) Flexibility (FR) (Cronbachss = 0.63, M=3.66, S=0.76) It bugs me when the atrate wasnt protable in a billing period. (M=3.66; S=1.11; median=4) It feels like a decit when another tariff would have been cheaper than the atrate in a billing period. (M=3.58; S=1.11; median=4) It matters to me to max out the atrate. (M=3.49; S=1.19; median=4) It feels like a gain when the atrate was cheaper than the next best tariff in a billing period. (M=3.92; S=1.00; median=4) Convenience (Cronbachss = 0.79, M=2.11, S=0.84) Figuring out which rate is better takes so long that it isnt worth the effort. (M=2.01; S=1.04; median=2) It is too much trouble to nd out the prices for mobile telecommunications (M=2.08; S=1.20; media=2) The money you can save by nding a better calling plan than your current one doesnt make up for the time and effort involved (M=2.22; S=1.02; mean=2) The time it takes to switch to a cheaper calling plan isnt worth the effort. (M=2.11; S=1.01; median=2)

Next, we consider the respondents atrate choice in comparison to cost cap tariffs in different usage scenarios (Table IV). In this setting, the cost cap tariff offers a cap at the amount of the atrate price. Therefore, the cost cap tariff weakly dominates the atrate tariff for those usage scenarios where the minimum usage may be below 300 minutes, i.e. the point at where the cost cap becomes binding. Thus, in the absence of tariff biases, one would expect that the atrate plan is not chosen in the rst two scenarios. In the remaining two usage scenarios, the atrate and the cost cap are factually identical because the cost cap is reached even under minimum usage. Thus, there is no advantage in selecting the cost cap tariff anymore. To the contrary, if respondents are averse towards the taximeter effect, they may even prefer the atrate under these conditions. Thus, ex ante one could presume that the atrate is chosen in at least 50% of the cases in the latter two scenarios.
TABLE IV C HOICE OF FLATRATE OVER COST CAP PLAN Usage (minutes) Flatrate choice at Min Avg. Max 45 e 200 400 600 10.28% 250 400 550 11.21% 300 400 500 16.82% 350 400 450 17.29% compared to a cost cap plan at 15 cents/minute and 45 e cost cap

The actual choices of respondents differ substantially from the ex ante presumptions and are strongly in favor of the cost
13 At the time of the survey the o2o tariff offers per-minute prices of 15 cents and a cost cap at 40 e.

Further insight on the reasons by which mobile customers perceive the relative attractiveness of pay-per-use, atrate or cost cap tariffs can be gained by evaluating the constructs which capture the taximeter, insurance, exibility and convenience effects. Furthermore, recall that cost cap tariff framework allows us to present customers with constructs that can capture their perception of the taximeter and insurance

effect independently. In contrast to a atrate framework, the cost cap tariff allows for cost insurance, while maintaining the taximeter property as long as the cost cap is not reached. Therefore, we present our results separately for those constructs that are evaluated within the atrate framework (Table V) and those that are evaluated within the cost cap framework (Table VI).14
TABLE VI C ONSTRUCTS CAPTURING COST CAP EFFECTS Taximeter (CC) (Cronbachss = 0.60, M=3.11, S=1.06) Under a cost cap tariff I enjoy the minutes before the cost cap becomes binding less. (M=2.88, S=1.29, median=3) Until the cost cap becomes binding I pay more attention to my usage behavior. (M=3.33; S=1.22; median=4) Insurance (CC) (Cronbachss = 0.72, M=3.47, S=0.97) With a cost cap I can use the mobile phone more jauntily. (M=3.54; S=1.06; median=4) I think less about my usage behavior in a tariff with cost cap, because my bill can never exceed a xed amount. (M=3.39; S=1.13; median=4)

TABLE VII L OGIT R EGRESSIONS ON F LATRATE C HOICE Flatrate (1) vs. Flatrate (1) vs. Pay-per-use (0) Cost Cap (0) Taximeter (FR) 0.109 -0.366 (0.105) (0.129) Insurance (FR) 0.327 0.310 (0.0983) (0.122) Flexibility (FR) -0.130 -0.210 (0.120) (0.147) Taximeter (CC) -0.0243 0.268 (0.0863) (0.109) Insurance (CC) -0.0747 -0.234 (0.0941) (0.119) Convenience -0.0519 -0.0461 (0.105) (0.136) Risk Aversion 0.0595 -0.212 (0.0480) (0.0587) Male 0.0909 -0.254 (0.199) (0.246) Age 0.0780 -0.0347 (0.0576) (0.0733) Prepaid -0.190 -0.167 (0.195) (0.247) Bill -0.00753 0.00451 (0.00585) (0.00620) Cost Control -0.116 -0.0812 (0.119) (0.146) Max Usage 0.00951 -0.00444 (0.000868) (0.00185) Min Usage 0.00735 (0.000855) Flatrate Price -0.213 (0.0422) Constant 2.382 4.337 (2.482) (2.130) Observations 852 852 Log likelihood -449.97 -320.28 Pseudo R2 22.15 % 6.04 % Correctly Classied 72.30 % 86.38% Standard errors in parentheses p < 0.05, p < 0.01, p < 0.001

Notice that the internal consistency of the constructs, as evaluated by Cronbachs scale reliability coefcient , is high throughout (M=0.72; S=0.086). This is particularly noticeable for the exibility and cost cap constructs, which have been used in this study for the rst time.15 Finally, we estimate the inuence of the above constructs on tariff choice in a nonlinear logit model. In the regression we also include the available variables on the usage scenario (min usage, max usage, atrate price), the individual telephony behavior (prepaid, bill, cost control), socio-demographics (male, age) and the individual risk attitude as explanatory variables in the regression model.16 Table VII shows the results of the logit regressions. In the rst model, which estimates the impact of the explanatory variables on the probability to choose a atrate plan over a pay-per-use plan, only two of the four signicant variables have a non-negligible effect. More precisely, the choice of a atrate plan over a pay-per-use plan is mainly driven by the insurance effect and the price of the atrate. An increase of the insurance effect by one unit will c.p increase the probability of atrate choice by 38.7%, whereas an increase of the price by 1 e will decrease the probability of choosing the atrate by 23.7%. The second model estimates the participants choice of the atrate over the cost cap plan with the atrate-equivalent cost
14 The convenience construct is in fact independent of the particular tariff framework, but listed under the atrate effects for conciseness and comparability to previous studies. 15 We followed the factor analysis methodology in order to eliminate and group the questions items according to the scale reliability coefcient. 16 Pairwise correlations (Pearson) of all explanatory variables used in the logit regressions were all below the threshold of 50%.

cap.17 In general, the explanatory power of this model, as evaluated by Log likelihood and Pseudo R2 , is lower compared to the rst model, because the atrate is seldomly chosen in our sample. Whether the respondents prefer the atrate or the cost cap plan depends foremost on the taximeter and insurance effects as well as on the degree of risk aversion. An increase of a participants evaluation of the taximeter effect under a cost cap plan by one unit makes him 30.6% more likely to choose the atrate. Interestingly, with respect to the atrate taximeter effect, the impact is opposite: Respondents who enjoy mobile telephony more under a atrate plan (as opposed to a pay-per-use plan) are in fact more likely to chose a cost cap plan over the atrate, if given the option. Moreover, this effect is strong, since an increase in the atrate taximeter effect by one unit increases the probability of chosing a cost cap plan by 44.1%. This analysis also reveals that the context is important for the evaluation of the taximeter effect. We will return to this point during the discussion of the results in Section VI.
17 The variables min usage and atrate price were omitted from the regression due to collinearity or invariance, respectively.

Likewise, also the two insurance effects have opposite impact on the respondents choice of the atrate plan. First, recall that the cost cap is set at the price of the atrate in all choice scenarios. Therefore, respondents are equally insured against high usage under both plans. Nevertheless, participants are 36% more likely to chose a atrate over a cost cap tariff if their assessment of the insurance effect of atrates (compared to pay-per-use plans) increases by one unit. In reverse, if asked for the insurance effect with explicit reference to the cost cap tariff, a unit increase will result in an increased probability of 26.3% towards choosing the cost cap plan over the atrate. Again, this highlights that consumers perceive the taximeter and insurance effect differently in the context of a atrate or a cost cap plan. Interestingly, also the risk preference has a strong effect on the choice of the two equally capped tariffs. An increase of respondents risk aversion by one unit18 increases the chances to choose the cost cap plan by 23.7%. This result is comprehensible, since the cost cap plan does not only insure against high usage volumes, but also against low usage. VI. D ISCUSSION AND C ONCLUSION In this article we have investigated mobile telephony customers perception of the recently introduced cost cap tariff in comparison to corresponding pay-per-use and atrate plans. Previous studies have identied several psychological effects which are believed to bias consumers choice towards atrate plans. However, atrate plans also limit customers tariff exibility in months of low usage. Cost cap tariffs are a hybrid between pay-per-use and atrate plans and can offer cost insurance while maintaining exibility. Our investigation of cost cap plans reveals three main insights. First, customers favor cost caps tariffs over atrates that are priced at the cost cap level, even in those usage scenarios where the cost cap is reached with certainty. Obviously cost cap tariffs seem to draw consumers attention, especially when exibility is important to them. In fact, in our sample respondents show a strong desire for exibility. In contrast to previous quasi-experimental studies on tariff-choice, our study cannot conrm a atrate bias. To the contrary, our results are in favor of a pay-per-use bias and in line with [1] and [23], who have derived their results empirically from the 1986 Louisville, Kentucky local telephone tariff dataset. One could argue, that the pronounced desire for exibilty in our sample is due to the fact that we have distributed the survey among students of economics, who are aware of costs in general but also primed to make rational choices. The sample is obviously not representative of the German population and also rather conservative in terms of biased choice. However, the same arguments hold true for the studies that have found a atrate bias in a similar group [14], [21]. Furthermore, in contrast to [14], we have confronted respondents with realistic choice scenarios in terms of usage and prices. Nevertheless it cannot
18 Risk aversion is measured by the number of safe choices from the ten non-hypothetical choice scenarios of the risk aversion test. Thus, the risk aversion measure can take any integer value from 1 to 10.

be excluded that our ndings are extenuated with a larger representative set of participants. Furthermore, it is annotated that it is not surprising that exibility is not a signicant factor in our regression analysis: Because all respondents have in our sample have conrmed a high desire for exibility (M=3.66/5) with small standard deviation (S=0.76), this construct is not a good discriminator in the regression. Secondly, we can conrm that the choice of cost caps tariffs is in fact also driven by the taximeter and insurance effects. The convenience effect, on the contrary, turns out to be insignicant in our dataset. However, this may again be due to a sample bias, since students do not tend to place emphasis on convenience (M=2.11/5) and are rather homogeneous in this perception (S=0.84). Our third insight is more subtle and concerns the way in which the taximeter and insurance effects inuence tariff choice. In particular, we nd an opposite impact of these effects if evaluated with respect to a atrate or a cost cap tariff: When given the opportunity to choose between a atrate and a comparable cost cap plan, those consumers tend to favor the cost cap who prefer a atrate (in lieu of a pay-per-use plan) due to the taximeter effect. On the contrary, there also seems to be a distinct taximeter effect under a cost cap plan that induces customers to divert away from the cost cap tariff and towards the atrate plan. Similarly, although participants are equally insured under the available atrate and cost cap plan, those customers that have revealed the insurance effect as a driver for choosing the atrate plan are indeed less likely to choose a cost cap plan instead. However, if those customers that feel insured under a cost cap tariff will also choose this tariff variant more likely. In summary, we nd that cost cap tariffs have combined desirable properties of pay-per-use and atrate plans and that customers are well aware of this. Our insights on cost cap tariffs have direct ramications for the strategic management of mobile telephony providers that seek to offer this tariff. If people are, as other studies suggest (e.g. [9], [10], [15], [19], [24], [25]), presumably willing to pay a higher than average monthly fee for a atrate, it is likely that this holds also true under a cost cap tariff. Thereby, cost cap tariffs are particularly interesting for those customers that have a strong desire to be insured against high telephony bills, but wish to maintain tariff exibility. Thus, on the one hand, the cost cap tariff is attractive to those customers that are currently under a atrate plan if their perception of the cost-cap-related taximeter effect is not too strong. Especially risk averse atrate customers may even be willing to opt for a cost cap plan although the cost cap is higher than the corresponding atrate price, because then they are insured in months of low usage. On the other hand, those users that have currently a pay-per-use plan, because they have a strong desire for a exible tariff or because they have a comparably low usage volume, could also be lurked into a cost cap tariff even if this means to accept higher perminute prices. In other words, these customers are willing to pay an indirect risk premium via the per-minute price in order to be insured against high telephone bills. If the cost cap does

not become binding, the mobile operator can especially prot from these customers. In fact, this latter reasoning seems to have been the main driver for o2 to introduce a cost cap tariff whose cost cap is set at the level of the cheapest available all-net atrate in the market. On the contrary, the per-minute price under this cost cap tariff is twice as high as the cheapest standard pay-per-use rate in the market. However, from a strategic management perspective, it is also very risky to offer a cost cap tariff. In contrast to a atrate, under a cost cap plan consumers with a comparably low usage cannot compensate for the costs of heavy users. But, the company may also speculate on the existence of a more subtle effect: Because customers are insured against high costs and because the cost cap tariff may effectively mitigate the taximeter effect with respect to a pay-per-use plan, a customer may in fact phone more than under a pay-per-use plan, but total costs yet remain below the cost cap. The marketing of o2 focuses on this aspect of the consumption behavior. Users should feel save with the so-called cost airbag and therefore do not have to think about their consumption behavior anymore. Therefore, the target group for this tariff scheme does not seem to be the typical heavy or actual atrate user, but the average user who does not want to commit himself to high monthly costs. We have merely presented a rst step into the investigation of cost cap tariffs, which have, to the best of our knowledge, previously not been assessed academically. While our study has focused more on the choice between cost cap and atrate plans, future work should especially consider the choice between cost cap and pay-per-use tariffs in more detail. Recall that a cost cap tariff dominates the pay-per-use and the atrate plan at equal per-minute prices and at a cost cap level equal to the atrate price. Thus, future research should systematically investigate the cost cap tariff bias in those scenarios where (i) the per-minute price is higher in the cost cap tariff than in an equivalent pay-per-use plan and (ii) the cost cap is higher than the corresponding atrate price. Furthermore, it seems interesting to further investigate the reasons why the insurance and taximeter effects are perceived differently under a cost cap and atrate tariff framework. Within the atrate framework, the insurance and taximeter effect are always coincidental, whereas the effects can be separated under a cost cap framework. The question remains if consumers experience the taximeter effect even if they are insured and how this affects their perception during telephony consumption as well as their potential tariff switching behavior. R EFERENCES
[1] E. J. Miravete, Choosing the Wrong Calling Plan? Ignorance and Learning, American Economic Review, vol. 93, pp. 297310, 2003. [2] , Are All Those Calling Plans Really Necessary? The Limited Gains From Complex Tariffs, 2007, cEPR Discussion Paper No. 4237. [3] , The Doubtful Protability of Foggy Pricing, 2009, cEPR Discussion Paper No. DP6295. [4] J. Hodge, Tariff structures and access substitution of mobile cellular for xed line in South Africa, Telecommunications Policy, vol. 29, no. 7, pp. 493505, 2005. [5] S. Littlechild, Two-part tariffs and consumption externalities, The Bell Journal of Economics, vol. 6, no. 2, pp. 661670, 1975.

[6] T. Valletti, The theory of access pricing and its linkage with investment incentives, Telecommunications Policy, vol. 27, no. 10-11, pp. 659675, 2003. [7] W. Oi, A Disneyland dilemma: Two-part tariffs for a Mickey Mouse monopoly, The Quarterly Journal of Economics, vol. 85, no. 1, pp. 7796, 1971. [8] V. Briegleb. (2009, 11) O2 mit mehr kunden und gewinn. Heise Newsmeldung. [Online]. Available: http://www.heise.de/mobil/meldung/O2mit-mehr-Kunden-und-Gewinn-857876.html [9] S. Della Vigna and U. Malmendier, Paying not to go to the gym, American Economic Review, vol. 96, no. 3, pp. 694719, 2006. [10] K. E. Train, M. Ben-akiva, and T. Atherton, Consumption Patterns and Self-Selecting Tariffs, Review of Economic and Statistics, vol. 1, pp. 6273, 1989. [11] D. J. Kridel, D. E. Lehman, and D. L. Weisman, Option Value, Telecommunication Demand, and Policy, Information Economics and Policy, vol. 5, pp. 12544, 1993. [12] A. Lambrecht and B. Skiera, Paying too much and being happy about it: Existence, causes and consequences of tariff-choice biases, Journal of Marketing Research, no. May, 2006. [13] H. Mitomo, T. Otsuka, and K. Nakaba, A Behavioral Economic Interpretation of the Preference for Flat Rates: A Case of Post-Paid Mobile Phone Services, in 18th European Regional ITS Conference, International Telecommunications Society, Istanbul. Springer, 2007. [14] T. J. Gerpott, Biased choice of a mobile telephony tari type: Exploring usage boundary perceptions as a cognitive cause in choosing between a use-based or a at rate plan, Telematics and Informatics, vol. 26, pp. 167179, 2009. [15] J. Nunes, A Cognitive Model of Peoples Usage Estimations, Journal of Marketing Research, vol. 37, pp. 397409, 2000. [16] M. D. Grubb, Selling to overcondent consumers, American Economic Review, vol. 99, no. 5, pp. 17701807, December 2009. [17] D. Just and B. Wansink, The Fixed Price Paradox: Conicting Effects of All-You-Can-Eat Pricing. [18] R. H. Thaler, Mental Accounting and Consumer Choice, Marketing Science, vol. 4, no. 3, pp. 199214, 1985. [19] D. Prelec and G. Loewenstein, The Red and the Black: Mental Accouting of Savings and Debt, Marketing Science, vol. 17, no. 1, pp. 428, 1998. [20] R. H. Thaler, Mental accounting matters, Journal of Behavioral Decision Making, vol. 12, no. 3, pp. 183206, September 1999. [21] A. Lambrecht and B. Skiera, Ursachen eines atrate-bias, systematisierung und messung der einussfaktoren, Schmalenbachs Zeitschrift fr betriebswirtschaftliche Forschung, vol. 58, pp. 588617, 2006. [22] C. Holt and S. Laury, Risk aversion and incentive effects, The American Economic Review, vol. 92, no. 5, pp. 16441655, 2002. [23] S. Narayanan, P. Chintagunta, and E. Miravete, The role of self selection, usage uncertainty and learning in the demand for local telephone service, Quantitative Marketing and Economics, vol. 5, no. 1, pp. 134, 2007. [24] K. E. Train, D. L. Mcfadden, and M. Ben-akiva, The Demand for Local Telephone Service: A Fully Discrete Model of Residential Calling Patterns and Service Choices, Rand Journal of Economics, vol. 18, no. 1, pp. 10923, 1987. [25] B. M. Mitchell and I. Vogelsang, Telecommunication Pricing: Theory and Practice. Cambridge University Press, 1991.