What is the magnitude of fixed–mobile call substitution? Empirical evidence from 16 European countries

Telecommunications Policy 38 (2014) 771–782

Contents lists available at ScienceDirect

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

What is the magnitude of fixed–mobile call substitution? Empirical evidence from 16 European countries Anne-Kathrin Barth n, Ulrich Heimeshoff Duesseldorf Institute for Competition Economics (DICE), Germany

a r t i c l e i n f o

abstract

Available online 21 June 2014

This paper investigates the degree of fixed–mobile call substitution (FMCS) within different European countries. We use quarterly data from 2004 to mid-2010 on 16 mainly Western European countries. By applying dynamic panel data techniques, we are able to estimate short- and long-run elasticities of the telecommunication usage prices on the fixed-line call demand. The own-price and cross-price elasticities found give strong empirical evidence for substitutional effects towards mobile services. In particular, the estimated cross-price elasticities of the mobile price on the fixed-line call demand are relatively large compared to other studies. & 2014 Elsevier Ltd. All rights reserved.

Keywords: Dynamic panel model Fixed–mobile substitution Telecommunication markets

1. Introduction After the implementation of GSM digital technology at the beginning of the 1990s, mobile devices became mass products, prices dropped and penetration rates dramatically increased (Gruber, 2005; Hausman, 2002). As well as the fact that the number of mobile subscribers has been larger than the number of fixed-line subscriptions since the early 2000s, we observe that fixed and mobile voice traffic volumes are converging. Whereas mobile call volumes are rising, fixed-line voice traffic volumes have continuously declined over the past decade. Fig. 1 exemplifies the development of the average monthly fixed and mobile voice traffic per subscriber for four different European countries from 2005 to 2010. We chose the UK because it has always been a leader in market liberalization and is one of the major economies in the European Union (EU). Furthermore, Germany is included as the largest economy in the EU. Austria and Finland are good examples of smaller member countries where especially in Finland as well as in other Nordic countries mobile telecommunications have been important much earlier than in other European countries. These countries provide a thorough overview of the convergence of fixed and mobile voice traffic in the EU. Obviously, the progress of convergence varies between these countries. For instance, in Austria and Finland mobile voice traffic has already exceeded the fixed-line traffic several years ago, and continues to do so. In other countries, such as Germany and the UK, however, fixed-line phones are still used more often to place calls than mobile devices. Fixed and mobile telecommunications markets are monitored by national regulatory authorities (Laffont & Tirole, 2000), but the degree of regulation is quite different. On the one hand, fixed markets are regulated quite heavily. On the other hand, mobile markets are regulated less restrictively, as they were more competitive from their inception (Haucap, 2003). However, recent observations lead to the question whether asymmetric regulation of fixed and mobile markets is still appropriate. If convergence of fixed and mobile traffic markets leads to substitution of both services, asymmetric regulation of the two markets is no longer suitable. Competitive pressure from one market due to the substitutability of services might restrict n

Corresponding author. E-mail addresses: [email protected] (A.-K. Barth), [email protected] (U. Heimeshoff).

http://dx.doi.org/10.1016/j.telpol.2014.04.009 0308-5961/& 2014 Elsevier Ltd. All rights reserved.

772

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

Austria

in min. 250 200 150 100 50 0

2005

2006

in min. 300

2007

2008

2009

2010

Germany

United Kingdom

in min. 350 300 250 200 150 100 50 0 2005

2006

in min. 300

250

2007

2008

2009

2010

Finland

250

200

200

150

150

100

100

50

50 0

0 2005

2006

2007

2008

2009

2005

2010

Fixed line - Minutes of usage per subscriber

2006

2007

2008

2009

2010

Mobile - Minutes of usage per subscriber Fig. 1.

market power in the other market so that an isolated view of single markets is no longer appropriate from a regulatory as well as an antitrust point of view. We expect convergence and, as a result, substitution between fixed and mobile services becoming stronger over time and hypothesize that one might arrive at a joint market applying standard market delineation techniques in the near future. If this hypothesis holds, a new market definition for regulation has to be found. The number of econometric studies analyzing the substitutional relationship between fixed and mobile networks is limited and their results are quite ambiguous. Most of the studies use data up to 2003. In contrast, studies using more recent data, including Briglauer, Schwarz, and Zulehner (2011), Grzybowski (2011) and Barth and Heimeshoff (2014), unanimously conclude that the two services are substitutes, at least in developed countries.1 These findings substantiate that fixed– mobile substitution already prevails. However, these studies differ on whether they just find significant cross price effects or apply formal market delineation techniques to define antitrust markets as for instance Briglauer et al. (2011) do. Consequently, the research focus has shifted from the question of whether the two technologies are substitutes or not, to the question of to what extent fixed and mobile services are substitutable, and if the magnitude found is strong enough to justify regulatory adjustments. Few studies exist that focus specifically on the usage of mobile phones instead of access, and all find different degrees of substitutability. Additionally, there is, to the best of our knowledge, no econometric paper analyzing fixed-to-mobile call substitution in a multiple-country setting. Therefore, based on the panel structure of our dataset the present paper controls for unobserved heterogeneity and avoids possible biases in estimated elasticities due to missing variables. As a result, our analysis sheds some new light on the relationship between fixed and mobile telecommunications services. We address fixed–mobile call substitution within 16 mainly Western European countries. Using quarterly data from 2004 to mid-2010, the paper analyzes to what extent fixed and mobile phone calls are substitutes. Our paper is structured as follows: Section 2 provides an overview of the empirical literature related to fixed–mobile substitution; in Section 3, the dataset and its descriptive statistics will be explained; Section 4 introduces our model specification and describes our estimation approach; and Section 5 explains our main results. Finally, Section 6 concludes.

2. Literature review Fixed–mobile substitution (FMS) can be analyzed on different levels: especially access and usage (ITU, 2010). Hence, empirical research on penetration models, as well as studies estimating access or calling demand, is relevant for the analysis of FMS (Vogelsang, 2010). To analyze the substitutability between products, usually own- and cross-price elasticities are estimated (Taylor, 1994). The following two subsections separately discuss the existing literature on the access and usage level.2 1 2

Note that Barth and Heimeshoff (2014) focus on access substitution. The literature review is based on the corresponding section in our previous paper (Barth & Heimeshoff, 2014).

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

773

2.1. Access level Studies analyzing FMS on the access level (FMAS) show a rather mixed picture. Some studies find fixed and mobile services to be complements at the subscription level, others show evidence of substitutability. Although the analysis of fixed–mobile substitution (FMS) is mainly an empirical question, the related econometric literature is, at least for Europe, not very extensive. Studies exist for South Korea, the USA, Portugal, the UK, and some African and Eastern European countries. Additionally, recent papers address FMS in India, Austria, China, Spain and the OECD countries.3 Based on panel data for the period 1991–1998 for eight South Korean provinces, Sung and Lee (2002) show that a 1% increase in the number of mobile phones results in a 0.1–0.2% reduction of fixed-line connections. They conclude that fixed and mobile telephones are substitutes on Korean telecommunications markets. Precisely, the number of mobile subscribers is positively related to the number of fixed-line disconnections, but negatively related to the number of new fixed-line connections, which suggests net substitution between fixed and mobile services. Estimating a hierarchical Bayes model for discrete choice data and using South Korean survey data for 2007, Rhee and Park (2011) find some evidence for separate fixedline and mobile telephony markets in South Korea. However, the authors predict that the two markets will converge in the near future as the mobile price premium continues to decrease. Rodini, Ward, and Woroch (2003), Ingraham and Sidak (2004), and Ward and Woroch (2010) provide evidence of the existence of substitutability between fixed and mobile networks in the USA by using the same US survey data for the time period 1999–2001. Rodini et al. (2003) analyze the substitutability between fixed and mobile access in the US modeling consumers’ wireless and second fixed-line subscription decision using logit regressions. They estimate own and cross-price elasticities finding substitution effects. Ward and Woroch (2004) show comparable effects estimating an Almost Ideal Demand System-Model (Deaton & Muellbauer, 1980). They conclude that mobile services are substitutes for fixed-line telephony at the traffic level, but not at the access level. However, note that they find a moderate degree of substitutability and further empirical evidence is needed to support the substitution hypotheses. Ward and Woroch (2010) estimate cross-price elasticities between fixed and mobile subscriptions by making use of US low-income subsidy programs (Lifeline Assistance) which cause large changes in fixed-line prices. Although they use the identical U.S. survey data, the elasticities found are larger than those for second lines reported in Rodini et al. (2003). Due to stronger price variation by incorporating subsidy programs, they are able to estimate the demand relationships versus the subscription decisions in Rodini et al. (2003). Caves (2011) estimates single equation models as well as simultaneous equation models for fixed and mobile demands using U.S. state-level panel data from 2001 to 2007. The author studies wireless- and wireline-access demand and finds that a 1% decrease in wireless prices results in a 1.2–1.3% decrease in the demand for fixed-line services. Gruber and Verboven (2001) deduce from their study, comprising data of 140 countries from 1981 to 1995, that the diffusion of mobile phones tends to be larger in countries with higher fixed network penetration. They conclude that the two technologies are complements. Barros and Cadima (2000) analyze time series data on fixed and mobile access in Portugal from 1981 to 1999. They find a negative effect of mobile phone diffusion on fixed-line penetration rates but none vice versa. Madden and Coble-Neal (2004) study FMS in 56 countries between 1995 and 2000 in a dynamic demand model and assess significant substitution effects between mobile and fixed-line subscription rates. They report a price elasticity of 0.12 and a cross-price elasticity with regard to the mobile price of
0.05 for fixed telephony. Heimeshoff (2008) studies FMS on the access level and estimates cross-price elasticities for 30 OECD countries between 1990 and 2003 by using instrumental variable regressions. Possible endogeneity problems are solved by instrumenting fixed and mobile prices and market structure with instrumental variables related to costs and policy indicators. This study provides evidence of one-way substitution, where mobile telephony can be a substitute to fixed-line services but not vice versa. Hamilton (2003) uses annual data from 1985 to 1997 for 23 African countries. The study shows that fixed and mobile phones in many African countries are still no substitutes. In many African countries usage of mobile phones does not reduce fixed-line usage but is primarily an improvement in social status. Compared to studies concentrating on developed countries, these results are not surprising, because in countries that lack an extensive fixed-line infrastructure, like many African and other low developed countries, mobile phone usage is often a result of a lack of supply of fixed-line services. Mobile phones are often the only means of access to a telephone in these countries. Using South African survey data from 1998 to 2001, Hodge (2005) studies how differences in tariff structures between fixed and mobile services have accounted for the popularity of cellular technology. Hodge finds that in low income households mobile phones are perceived as substitutes, while in high income households the two services are treated as complements. Vagliasindi, Güney, and Taubman (2006) show substitutional relationships between fixed and mobile services for Eastern European countries based on cross-section data for 2002. Garbacz and Thompson (2007) analyze FMS in 53 less developed countries finding asymmetric substitutional effects. Fixed connections tend to be substitutes in the mobile market, whereas mobile phones might be complements in the fixed-line market. Overall, investigating substitutional effects between fixed and mobile services in transition countries is always difficult as low quality of fixed networks in these countries often does not allow fixed–mobile substitution. Instead, mobile phones are often the only possibility to receive access to telecommunications. Narayana et al. (2010) analyze FMS on the subscription level in India using cross sectional survey data for 2003. He includes subscription prices as well as usage prices as explanatory variables in his regressions and finds that both prices are correlated and that the usage price has, in comparison to the subscription price, a much larger and more significant effect on

3

There is also one paper comparing FMS in different countries (Mao, Tsai, & Chen, 2008).

774

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

mobile and fixed-line subscriptions. Narayana finds much stronger substitutional effects in both directions than other studies using cross-sectional data for 2003. Ward and Zheng (2012) analyze fixed–mobile access substitution for 31 Chinese provinces using data from 1998 to 2007. Using Arellano–Bond type linear dynamic panel models, the authors estimate short- and long-run elasticities of fixed and mobile prices on fixed and mobile subscriptions, respectively and find cross-price elasticities between 0.20 and 0.28 in the short-run and 0.39 and 0.56 in the long-run. In comparison to other empirical studies they find rather strong substitutional effects. Endogeneity problems are dealt with by using further lags of their subscription measure as well as price related measures like the average Herfindahl index and the fraction of state ownership in fixed and mobile telecommunications for neighboring nearest provinces as instruments. Next to the papers focusing on voice substitution, Srinuan, Srinuan, and Bohlin (2012) examine Swedish survey data for 2009 and investigate whether fixed and mobile broadband services are complements or substitutes. They find that fixed and mobile broadband services are substitutes in most geographical parts of Sweden. Applying logit techniques and including prices for different internet technologies (DSL, cable, LAN/Fibre and mobile internet), the authors find that DSL and mobile broadband are more sensitive to price changes than cable and LAN/Fibre whereas the degree of substitutability varies from area to area. Based on quarterly Spanish household survey data for 2004-2009 and logistic regressions, Suarez and GarciaMarinoso (2013) estimate the percentage of households that engage in fixed-to-mobile-access-substitution in Spain and its main drivers. They find substitutability (0.02–0.79% per quarter) to be rather small. Main drivers for fixed-to-mobile-accesssubstitution are the availability of internet and mobile services previous to the substitution decision, socio-demographic characteristics of households, such as age, and the degree and types of expenditures on fixed services. Barth and Heimeshoff (2014) estimate the effect of several variables, particularly prices, on the stocks of fixed and mobile subscriptions. Applying dynamic panel approaches and using data of the EU27 from 2003 to 2009, the results indicate modest substitution effects towards mobile telecommunication networks. Additionally, there are some studies of European regulators which also discuss the issue of fixed–mobile substitution. Griffiths and Dobardziev (2003) conclude for the Netherlands that there exists some degree of substitutability already and this process will proceed as mobile call prices will continue to fall. Summing up, studies only exist for South Korea, Portugal, the USA, and some African and Eastern European countries. In addition, recent papers address FMS in India, the OECD and the European Union. Furthermore, the results are not as clear as expected. A possible reason might be that the estimation of cross-price elasticities is typically less robust than estimated own-price effects (Bonfrer, Berndt, & Silk, 2006). However, the results give some evidence that fixed and mobile services are already perceived to be substitutes in developed countries, but not (yet) in low-income countries. This finding is not surprising as in many African and other less developed countries: an extensive fixed-line infrastructure is missing. Thus, mobile phones are often the only possibility of having access to telecommunication services. This result is confirmed by Mao et al. (2008) showing that substitution of subscriptions is more common in less developed countries, whereas in high developed countries the so-called traffic substitution is more likely. However, another reason for the different findings could be that the majority of empirical studies focus on time periods before 2003. It is likely that the substitution effects of fixed and mobile networks are much stronger nowadays, e.g., due to further price reductions in mobile markets. Barth and Heimeshoff (2014) aim to fill this research gap by using data up to 2009. They find moderate and highly significant one-way access substitution in favor of mobile networks, but we expect the substitution effects to be much larger on the traffic level. The next subsection presents the existing research related to the traffic level.

2.2. Traffic Level The findings on the traffic level are much clearer compared to the results of the literature analyzing access substitution. All studies focus on developed countries and find substitutional effects on the traffic level.4 Horvath and Maldoom (2002) study survey data of over 7000 British telephone users (repeated cross sections in three waves: 1999, 2000, 2001) in a simultaneous equations model, and additionally estimate some probit regressions. They find that using mobile phones significantly decreases fixed-line usage. Their findings support the conclusion that fixed and mobile phones are substitutes in British telecommunications markets. Analyzing monthly traffic and revenue data from 1997 to 2002 for South Korea, Yoon and Song (2003) show that fixed and mobile calls are substitutes and fixed–mobile convergence can be observed in South Korea. Sung (2003) reports that mobile calls are substitutes for fixed-line toll calls by using Korean regional panel data from 1993 to 1997. Using traffic data from 1996 to 2002 for South Korean telecommunications markets, Ahn, Lee, and Kim (2004) concur with these results. Ingraham and Sidak (2004) analyze the effects of long-distance fixed-line call prices on mobile demand and report a small, but highly significant cross-price-elasticity of þ0.02 adopting Ordinary Least Squares and Instrumental Variable regressions. Ward and Woroch (2004) make use of the US bill-harvesting data and report comparable effects, applying the Almost Ideal Demand System-Model (Deaton & Muellbauer, 1980). They conclude that mobile services are substitutes for fixed-line usage at the traffic level, but not at the access level. However, the effect is only 4 For an analysis of differences of short- and long-run elasticities between pre- and post-paid markets in Turkey see Karacuka, Haucap, and Heimeshoff (2011).

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

775

of moderate strength. Adopting least squares and 2SLS regression based again on the similar US survey data, Ingraham and Sidak (2004) analyze the effect of long-distance fixed-line call prices on mobile demand and report a small, but highly significant cross-elasticity of þ0.02. Briglauer et al. (2011) calculate short- and long-run cross-price elasticities for fixed-line domestic calling in response to mobile price changes in Austria for 2002–2007. They use monthly data on call minutes and take average revenues per minute as price data. While they observe small and sometimes insignificant estimates for short-run elasticities, their results for longrun cross-price elasticities suggest strong substitution effects. Wengler and Schäfer (2003) evaluate the findings of a telephone survey for Germany consisting of 1691 persons (first wave), 2014 persons (second wave) and 101 persons (third wave) collected in March and April 2003. They only observe a very moderate tendency for fixed–mobile substitution in Germany in 2003 and most of the survey participants argue that they do not substitute between their fixed line and mobile phones. To conclude, there are only a few studies analyzing FMS on the traffic level. Additionally, all papers, with the exception of Briglauer et al. (2011), again use quite old data. Overall, it can be concluded that there is, to the best of our knowledge, no empirical study on the traffic level incorporating multiple countries. Using cross-sectional data instead of panel data is disadvantageous as it is not possible to control for unobserved heterogeneity so that results are likely to be biased. Thus, we extend this strand of the literature by using recent panel data from 2004 to mid-2010 on 16 mainly Western European countries5 on a quarterly basis. The following sections provide an overview of the dataset and the applied econometric approach of our empirical study.

3. Data Our dataset consists of the following resources: data from the Telecoms Market Matrices of Analysys Mason for the outgoing national fixed-line traffic, usage prices and the number of prepaid customers (Analysys Mason, 2011a, 2011b). Information on penetration rates and GDP is found in Merrill Lynch's Wireless and Wireline Matrices (Merrill Lynch, 2011a, 2011b). Additionally, data on mobile-only customers comes from the “Eurobarometer: E-Communications Household Surveys”. We also incorporate data on fixed-to-mobile and fixed-to-fixed termination rates out of the “Progress Reports on the Single European Electronic Communication Market” (EU Commission, 2010a, 2010b). Both the surveys and reports have been provided by the Directorate-General Information Society of the EU Commission. Furthermore, we use the OECD statistics for demographic information and BEREC's MTR Snapshot for data on mobile termination rates (BEREC, 2010; OECD, 2011). Table 1 illustrates the descriptive statistics for all variables used in our analysis.6 trafficfix describes the total amount of national outgoing fixed-line voice traffic (in mio.). pfix and pmob represent the prices of fixed and mobile network calls per minute, respectively. These prices are constructed by dividing the total voice revenues of all operators in a specific country by the total minutes of usage. penwireless and penwireline refer to the penetration rates of mobile and fixed-line networks in each specific country, respectively. mtr describes the mobile termination rates, ftf the fixed-to-fixed termination rates and ftm the fixed-to-mobile termination rates. The control variable gdp stands for the GDP (in bn. Euros). The variable percmobonly depicts the percentage of households having mobile, but no fixed-line access. percprepaid describes the percentage of prepaid contracts and percunder40 the percentage of the population aged under 40. pop measures the population in a specific country (in mio.). trend is a linear time trend. The time trend can be interpreted as a continuous improvement in the service quality, an increase in the availability of services, and an enhanced network performance as well as decreasing prices (Grzybowski, 2005). We also incorporate seasonal dummies dQ2–dQ4 and quarterly time dummies d2–d26 into our regressions. For a robustness check, we include a linear time trend and seasonal dummies into our regressions. All price variables (pfix, pmob, mtr, ftf, ftm, and gdp), the population size pop and the fixed-line traffic volume trafficfix are measured in logarithms in order to interpret their coefficients as elasticities. Furthermore, all price variables are measured in USD adjusted by purchasing power parities to ensure in international comparison.

4. Model specification Our empirical model is based on the Houthakker–Taylor model (also see Houthakker and Taylor, 1970; Dewenter & Haucap, 2008). Following Taylor (1994),7 we assume that an individual subscriber's demand for telephone calls (q) depends on the price of a call (π), the price of a substitute (p), the number of the network subscribers (N) and the income of the consumer (μ). Additionally, the demand is driven by K
4 other factors (xk;t ) with k A ½5; K which include the number of the subscribers in other networks, the age of the subscriber and the type of the contract. Let qnt denotes the desired number of 5

The countries in our study are summarized in Table A1 in the appendix. Additionally, the definition of the variables used can be found in Table A2 in the appendix. 7 Taylor (1994) expects the individual subscriber's demand to also depend on the price of access to a telephone network. Unfortunately, we do not have information on access prices in our dataset. However, Briglauer et al. (2011) find that access prices are not significant in their demand estimation for calls. Furthermore, they conclude that their results are robust applying three different specifications which include (1) access plus call prices (2) only call prices or (3) average prices. Hence, it is reasonable to assume that lacking access prices will not cause significant biases in our estimation. 6

776

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

Table 1 Descriptive statistics. Variable

Obs.

Mean

Std. Dev.

Min

Max

Fixed line traffic Fixed line price Mobile price Wireless penetration Wireline penetration Mobile termination rates Fixed-to-fixed termination rates Fixed-to-mobile termination rates GDP Percentage mobile-only customers Percentage prepaid customers Percentage of population under 40 Population Trend

388 388 388 388 388 388 352 356 388 388 388 388 388 388

11 191.74 0.07 0.21 1.13 0.38 0.13 0.01 0.14 791.25 0.28 0.49 0.50 28.28 14.29

12 317.79 0.03 0.07 0.21 0.11 0.06 0.01 0.06 782.52 0.18 0.18 0.02 25.38 7.14

390.80 0.01 0.09 0.64 0.19 0.04 0.00 0.04 145.75 0.00 0.07 0.43 5.22 1

44 919.00 0.15 0.46 1.86 0.64 0.31 0.02 0.35 2829.56 0.81 0.91 0.54 82.87 26

calls during period t for given prices, level of subscribers, income and other variables. Thus, we postulate k¼K

qnt ¼ α0 þα1 π t þ α2 pt þ α3 Nt þ α4 μt þ ∑ αk xk;t :

ð1Þ

k¼5

Now, qt denotes the actual number of calls made during the period. It is assumed that whenever q and qn diverge, a proportion of θ is eliminated within each period. In particular qt
qt
1 ¼ θðqnt
qt
1 Þ;

ð2Þ

where 0 oθ r1.8 After some rearrangement, we obtain k¼K

qt ¼ α0 θ þ ð1
θÞqt
1 þ α1 θπ t þ α2 θpt þα3 θNt þα4 θμt þ ∑ αk θxk;t : k¼5

ð3Þ

From Eq. (3), we infer that the one-period effect of a marginal change in variable i on q is equal to αi θ. Thus, the shortand long-run derivatives of q with respect to the variable i are equal to αi θ and αi, respectively. Looking at the full system of subscribers, we postulate for the aggregate demand for calls: Q t
Q t
1 ¼ ψðQ nt
Q t
1 Þ: Similar to Eq. (1), we assume k¼K

Q nt ¼ α0 þ α1 π t þ α2 pt þ α3 Nt þα4 Y t þ ∑ αk xk;t ; k¼5

where Yt denotes the aggregated income. Consequently, we formulate k¼K

Q t ¼ α0 ψ þð1
ψÞQ t
1 þα1 ψπ t þ α2 ψpt þα3 ψNt þ α4 ψY t þ ∑ αk ψxk;t : k¼5

Taking the panel structure of our data into account, the following equation studies the effects of certain variables on the national outgoing fixed-line voice traffic. Again, the one-period effect of a marginal change of variable i on the national fixed-line voice traffic is equal to αi ψ. Thus, the short- and long-run elasticities of the variable i are equal to αi ψ and αi, respectively: trafficfixit ¼ α0 ψ þ ð1
ψÞtrafficfixit
1 þ α1 ψpfixit þ α2 ψpmobit þ α3 ψpenwirelineit k¼K

þα4 ψgdpit þ ∑ αk ψxk;it þ ϵit k¼5

We expect trafficfixt
1 to have a positive influence on the current fixed-line voice traffic volume for the simple reason that if the voice volumes were higher in the last quarter, they will be higher today due to consumer habits. One reason might be that many telecommunication contracts run for 1 or 2 years and include, for instance, free calls to specific networks. This could probably influence consumers’ calling behavior and therefore the usage only changes partially within the contract duration. Hence, if consumers react with some time lag, long-run elasticities are expected to be higher than the estimated short-run elasticities (Dewenter & Haucap, 2008). 8 This assumption is derived from the theoretical model. Note that it is not necessarily essential for our specification. For our purpose only the persistence of calling behavior is crucial. We thank an anonymous referee for pointing this out.

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

777

We assume that the fixed-line usage depends on current fixed-line prices. We expect the own prices to have a negative impact on the fixed-line usage, meaning that an increase in the own price leads to a decrease in the voice traffic volumes. In order to find substitutional effects, the current mobile price pmobit must have a positive effect on trafficfixit . Network effects measured by penworelineit and gdp are assumed to have a positive influence. The term xk;it includes all additional explanatory variables such as the wireless penetration rate and population size. Additionally, the percentage of the population aged under 40, the percentage of mobile-only and of prepaid customers are included into our regression. εit is an error term and the αs are parameters to be estimated. Due to the structure of our panel dataset, we estimate a dynamic fixed effects panel model using the Newey–West procedure to avoid distortions in standard errors due to autocorrelation and heteroscedasticity (Wooldridge, 2010, pp. 310–315). To prevent spurious regressions, we apply panel unit root tests for all variables in our data set. We find that only the variables penwireless, penwireline, ftf, gdp and percmobonly are non-stationary and all integrated of order one. Our dependent variable trafficfixit on the left-hand side is Ið0Þ, hence, cointegration relationships cannot be present in our dataset and there can be no spurious regression problems.9 As will be discussed in the next section, we also take possible endogeneity problems into account by using instrumental variable techniques. 5. Empirical results Avoiding possible endogeneity problems, we instrument the first lag of our dependent variable, penetration rates and usage prices. Hence, we use further lags of the variables as well as termination rates as instruments. Mobile termination rates are an important (variable) cost factor for the operators, which occur particularly for off-net calls. The national regulatory authority in each country determines the termination charges, which can therefore be considered as exogenous. This assumption can be criticized as the decision of the regulator may be affected by other factors such as changes in volumes. Nevertheless, termination rates are the only cost shifter that directly influences the variable costs and can be observed by econometricians. By applying overidentification tests, we test for the exogeneity of our instruments and we cannot reject the null hypothesis stating exogeneity of our instruments. We lag all termination rates by one-quarter since variations in termination rates are not directly passed onto customers (Briglauer et al., 2011, p. 13). Table 2 illustrates our results using a linear time trend (column 2) or quarterly time dummies (column 3).10 For both regressions, we identify statistically significant effects at the 5% or higher significance level from lagged national outgoing fixed-line traffic (trafficfixit
1 ), current fixed-line prices (pfixit ), the current mobile price (pmobit ), the percentage of mobile-only users (percmobonlyit ) and the percentage of prepaid customers (percprepaidit ). All significant variables have the expected signs. Regarding the regression using a linear time trend (second column), we find that the lag of the national outgoing fixed-line traffic volume has a large positive effect (þ 0.71) on the current fixed-line traffic volume, which is significant at a 1% significance level. The own-price elasticity is negative, as expected. In the short run, a 1% increase in the fixed-line price leads to a 0.14% decrease in the fixed-line traffic volume, whereas the fixed-line traffic volume declines by 0:1378=ðψ ¼ 0:2938Þ  0:47% in the long run. The cross-price elasticity is positive: a decrease in the current mobile price (pmobit ) causes a decrease in the fixed-line traffic volume. In the short run, a 1% reduction of the mobile price indicates a 0.13% decline in the fixed-line traffic volume. In the long run, the cross-price elasticity is given by 0:1250=ðψ ¼ 0:2938Þ  0:43. One should note that this finding is a strong indicator of fixed–mobile substitution on the traffic level, especially in the long run. In addition, the shares of the population using only mobile services and/or prepaid contracts have the expected negative effect on current fixed-line traffic. The results are robust when using time dummies instead of a linear trend (column 3). Overall, our findings provide evidence for short- and long-run call substitutions from fixed to mobile services. The effects found are larger than in other studies but in line with recent studies as Ward and Zheng (2012). Additionally, we apply the Sargan/ Hansen's j test of exogeneity of instruments. With p-values of 0.57 and 0.67, we cannot reject the null hypothesis of exogeneity stating the validity of our specifications. Our results do not suggest a joint market from an antitrust perspective because the estimated long-run elasticity of fixed-line services does not exceed standard critical thresholds used in market delineation tests (Briglauer et al., 2011; Vogelsang, 2010). The following section concludes and provides some discussion of policy implications. 6. Conclusion What are the implications of our results for telecommunications regulation? Do they imply that changes in market definition due to increasing competitive pressure on fixed-line services by mobile services are necessary? Following Vogelsang (2010) the relevant question is whether mobile operators already constrain the local market dominance of incumbent fixed carriers and not only if conditions of formal market delineation tests are met. It is not necessary to have a joint antitrust market for mobile competitors to restrain fixed-line operators by competitive pressure (see Vogelsang, 2010). For call markets this condition of a joint antitrust market seems to be fulfilled in some advanced telecommunications 9

For further information see Hamilton (1994). The corresponding test statistics can be found in Table A3 of the appendix. The first stage F-statistics and the corresponding p-values can be found in Table A4 in the appendix. Furthermore, the pairwise correlations between all variables used are summarized in Table A6. We also estimate our model without penetration rates. The results can be found in Table A5 in the appendix. 10

778

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

Table 2 Empirical results. Variable

With time trend

With time dummies 0.6587nnn (0.0995)
0.1661nn (0.0661) 0.1256nnn (0.0483) 0.1566n (0.0928) 0.0843 (0.2092)
0.1407 (0.0935)
0.3559nnn (0.1085)
0.5212nnn (0.1813)
1.2126 (1.4815)
0.4778 (0.6148)

Time dummies

0.7062 (0.0921)
0.1378nn (0.0624) 0.1250nn (0.0520) 0.1137 (0.0939) 0.1755 (0.2328)
0.0321 (0.0708)
0.3036nnn (0.1043)
0.3922nn (0.1766)
2.2370 (1.6979)
0.0424 (0.5719)
0.0331nnn (0.0054)
0.0626nnn (0.0078) 0.0639nnn (0.0134)
0.0059n (0.0033) no

ψ

0.2938

0.3413

0.9399 275 2.0270 0.5668

0.9465 275 1.5690 0.6664

Lag:Fixed line traffic Fixed line price Mobile price Wireless penetration Wireline penetration GDP Percentage mobile-only customers Percentage prepaid customers Percentage of population under 40 Population Quarter2 Quarter3 Quarter4 Trend

2

R N Hansen's j p-value

nnn

yes

Heteroscedasticity robust standard errors in parenthesis. n Statistically significant at the 10% level. nn Statistically significant at the 5% level. nnn Statistically significant at the 1% level.

markets, for example in Austria fixed and mobile services already belong to the same market as Briglauer et al. (2011) show using a SSNIP test setting.11 However, there is little doubt that competitive pressure from mobile to fixed markets exists, but if this pressure is sufficient to constrain fixed line operators’ market power is still an open question. Thinking of deregulation of, e.g., the local loop may be too early in some countries but appropriate in others. The fast development of telecommunications markets makes regular critical assessment of the status quo of regulatory obligations necessary. The future development and regulation of telecommunications markets will remain an important field of research, particularly because of technological change which will be a key aspect for fixed–mobile substitution and the meaning of telecommunications for economic growth and development (Czernich, Falck, Kretschmer, & Woessmann, 2011; Munnell, 1992; Norton, 1992; Röller & Waveman, 2001). Thinking about the future suitable market definition for fixed and mobile markets and regulation of operators having significant market power in fixed-line markets, takes us to the following conclusions. Following our results, fixed and mobile markets still form separate markets in regulation from the perspective of market delineation. An operator with significant market power in fixed-line markets would still be subject to regulatory obligations. However, from an antitrust perspective today it is often suggested not solely to rely on market delineation but on an effect based approach. The basic idea is to examine market power on the basis of actual conduct and its effects on market outcomes (Peeperkorn & Viertiö, 2009). To implement this approach, one has to take into account all aspects which might constrain competitors in their behavior. Some authors even suggest to abandon market definition at all and just rely on the analysis of effects (Kaplow, 2013). A good example might be potential competition by entry. Fixed and mobile access markets are an interesting example for this approach because they do not belong to the same market from the viewpoint of classical market delineation but there are interrelationships between them that might restrict

11 SSNIP stands for Small but Significant Non-transitory Increase in Price. This test is a standard method for market delineation in antitrust analysis, for example in merger cases.

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

779

an operator in one certain market due to substitutability between fixed and mobile access. This fact can have significant influence on merger regulation in antitrust. Following Briglauer et al. (2011) for Austria, markets for fixed and mobile usage might even form a single market from an antitrust perspective. However, this is still not the case for the majority of European telecommunications markets. From our point of view neglecting competitive constraints by mobile markets would lead to wrong conclusions evaluating fixed-line operators’ market power and possibly implementing regulatory obligations which might be too restrictive. A more flexible approach to market definition in regulation could be a solution to this problem. If regulators had the opportunity to base their decisions on a case by case analysis, analyzing competitive restrains, they could take into account competition by mobile operators in their regulatory measures. Such framework would be closer to the effects-based approach suggested in antitrust analysis in the U.S. and the European Union today. It has the major advantage that more flexibility in market delineation would lead to better regulatory decisions based on actual competitive pressure than relying regulatory decisions on a rather static framework which defines markets ex ante without taking into account actual market behavior. Our study supports the general assumption that call substitution from fixed to mobile services is prevailing with time. Our results have an ample impact with regard to regulation. Although we show that fixed and mobile markets are converging and becoming closer substitutes, regulatory obligations in the two markets are still quite different. In conjunction with the estimation results the suitability of the definition of separate fixed and mobile markets in the current European regulatory framework may need to be reconsidered for future telecommunications regulation.

Acknowledgments We are grateful to Ralf Dewenter, Julia Graf, Justus Haucap, Miguel Vidal, three anonymous referees, and the participants of the DICE Brown Bag Seminar and the 2012 ITS Regional Conference in Vienna for helpful comments and suggestions. All remaining errors are solely the responsibility of the authors.

Appendix A

Table A1 Countries included in the empirical study. Country

Period

Austria Belgium Czech Republic Denmark Finland France Germany Greece

Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1

2005–Q2 2004–Q2 2005–Q2 2005–Q2 2004–Q2 2004–Q2 2004–Q2 2004–Q2

2010 2010 2010 2010 2010 2010 2010 2010

Country

Period

Hungary Italy Netherlands Poland Portugal Spain Sweden UK

Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1

2005–Q2 2005–Q2 2005–Q2 2005–Q2 2004–Q2 2004–Q2 2004–Q2 2004–Q2

2010 2010 2010 2010 2010 2010 2010 2010

Table A2 Definition of the variables used. Variable

Description of variables

trafficfix pfix

National outgoing fixed-line voice traffic volume (in mio.) Fixed-line price per minute calculated as fixed-line voice revenue (without interconnect payments) divided by total outgoing fixed-line traffic, given in USD PPP Mobile price per minute calculated as mobile voice revenue (without interconnect payments) divided by total outgoing mobile traffic, given in USD PPP Mobile penetration rate Fixed-line pentration rate Mobile termination rates, given in USD PPP Fixed-to-fixed termination rates, given in USD PPP Fixed-to-mobile termination rates, given in USD PPP Gross national product, given in USD PPP (in bn.) Percentage of the population using only mobile, but no fixed-line telephony Percentage of prepaid customers among all mobile subscribers (excludes customers who have not used their mobile account for more than three months) Percentage of the population aged under 40 Population (in mio.)

pmob penwireless penwireline mtr ftf ftm gdp percmobonly percprepaid percunder40 pop

780

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

Table A3 Maddala–Wu unit root tests. Unit root tests

Levels

First differences

trafficfix χ2 Prob 4 χ 2

164.1328 0.0000

438.5566 0.0000

pfix χ2 Prob 4 χ 2

101.7945 0.0000

338.4291 0.0000

pmob χ2 Prob 4 χ 2

69.9723 0.0001

371.488 0.0000

penwireless χ2 Prob 4 χ 2

14.1420 0.9973

199.4026 0.0000

penwireline χ2 Prob 4 χ 2

26.0979 0.7593

201.0629 0.0000

mtr χ2 Prob 4 χ 2

63.8101 0.0007

364.4442 0.0000

ftf χ2 Prob 4 χ 2

16.5843 0.9888

195.7924 0.0000

ftm χ2 Prob 4 χ 2

57.0201 0.0042

281.2384 0.0000

gdp χ2 Prob 4 χ 2

10.1392 0.9999

142.395 0.0000

percmobonly χ2 Prob 4 χ 2

41.7303 0.1165

304.3285 0.0000

percprepaid χ2 Prob 4 χ 2

58.9761 0.0025

220.2582 0.0000

percunder40 χ2 Prob 4 χ 2

182.6028 0.0000

431.2025 0.0000

76.1572 0.0000

579.3324 0.0000

pop χ2 Prob 4 χ 2

Table A4 Instrumental variables: first stage F-statistics and p-values. Variable

F-statistic

p-value

Regression with linear time trend (column 2) trafficfixit
1

74.83

0.0000

pfixit

38.03

0.0000

pmobit penwirelessit penwirelineit

37.32 30.49 76.85

0.0000 0.0000 0.0000

Regression with time dummies (column 3) trafficfixit
1

66.50

0.0000

pfixit

36.47

0.0000

pmobit penwirelessit penwirelineit

40.71 31.46 69.91

0.0000 0.0000 0.0000

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

781

Table A5 Empirical results without penetration rates. Variable

With time trend

trafficfixit
1

0.7927nnn

0.7621nnn

pfixit

(0.0665)
0.0920nn

(0.075)
0.1046nn

percmobonlyit

(0.0469) 0.0860nn (0.0417) 0.0138 (0.0678)
0.2561nnn

(0.0501) 0.0916nn (0.0423) 0.1034 (0.0897)
0.2782nnn

percprepaidit

(0.0940)
0.2510n

(0.0947)
0.3378nn

(0.1348)
2.6886nn (1.2567)
0.2320 (0.4972)
0.0348nnn (0.0055)
0.0595nnn (0.0081) 0.0748nnn (0.0104)
0.0042nn (0.0020)

(0.1433)
2.4674nn (1.1427)
0.8232 (0.5595)

ψ

0.2073

0.2379

R2 N Hansen's j p-value

0.9384 284 0.7779 0.3778

0.9435 284 0.0044 0.9474

pmobit gdpit

percunder40it popit dQ2 dQ3 dQ4 trend

With time dummies

Time dummies

yes

Heteroscedasticity robust standard errors in parenthesis. n Statistically significant at the 10% level. nn Statistically significant at the 5% level. nnn Statistically significant at the 1% level.

Table A6 Pairwise correlation. trafficfix

pfix

pmob

penwireless

trafficfix pfix pmob penwireless penwireline gdp percmobonly percprepaid percunder40 pop

1.0000 0.4863n 0.1033n
0.1090n 0.4522n 0.9788n
0.4930n 0.1704n
0.2938n 0.9561n penwireline

1.0000 0.4369n
0.0489
0.5514n
0.4640n 0.5492n 0.3853n 0.4582n
0.3646n gdp

1.0000
0.4651n
0.0249 0.0423
0.1783n 0.4435n 0.2834n 0.1078n percmobonly

1.0000
0.3896n
0.0808 0.2490n 0.1688n
0.4698n
0.1170n percprepaid

penwireline gdp percmobonly percprepaid percunder40 pop

1.0000 0.3648n
0.6993n
0.1107n 0.1142n 0.2954n

percunder40 pop n

Significant at 5% level or higher.

1.0000
0.4279n 0.1448n
0.2873n 0.9728n percunder40 1.0000
0.2014n

1.0000
0.0910 0.0311
0.3823n

1.0000 0.0349 0.2151n pop

1.0000

782

A.-K. Barth, U. Heimeshoff / Telecommunications Policy 38 (2014) 771–782

References Ahn, H., Lee, J., & Kim, Y. (2004). Estimation of a fixed–mobile substitution model in Korean voice telephony markets. Working paper, Seoul. Analysys Mason (2011a). Telecoms market matrix—Western Europe, London. Analysys Mason (2011b). Telecoms market matrix—Central and Eastern Europe, London. Barros, P. L., & Cadima, N. (2000). The impact of mobile phone diffusion on the fixed-link network. CEPR discussion paper, London. Barth, A.-K., & Heimeshoff, U. (2014). Does the growth of mobile markets cause the demise of fixed networks? Evidence from the European Union. Telecommunications Policy, forthcoming, http://dx.doi.org/10.1016/j.telpol.2014.03.003. BEREC (2010). MTR benchmark snapshot. BoR (10) 30 Rev1, Latvia. Bonfrer, A., Berndt, E. R. & Silk, A. (2006). Anomalies in estimates of cross-price elasticities for marketing mix models: Theory and empirical test. NBER working paper 12756, Cambridge, USA. Briglauer, W., Schwarz, A., & Zulehner, C. (2011). Is fixed–mobile substitution strong enough to de-regulate fixed voice telephony? Evidence from the Austrian markets. Journal of Regulatory Economics, 39, 50–67. Caves, K. W. (2011). Quantifying price-driven wireless substitution in telephony. Telecommunications Policy, 35, 984–998. Czernich, N., Falck, O., Kretschmer, T., & Woessmann, L. (2011). Broadband infrastructure and economic growth. Economic Journal, 121, 505–532. Deaton, A. S., & Muellbauer, J. (1980). An almost ideal demand system. American Economic Review, 70, 312–326. Dewenter, R., & Haucap, J. (2008). Demand elasticities for mobile telecommunications in Austria. Journal of Economics and Statistics (Jahrbuecher fuer Nationaloekonomie und Statistik), 228(1), 49–63. EU Commission (2010a). E-Communications household survey report, Brussels. EU Commission (2010b). Progress report on the single European electronic communications market 2009 (15th report) (Vol. 2), Brussels. Garbacz, C., & Thompson, H. G. (2007). Demand for telecommunication services in developing countries. Telecommunications Policy, 31, 276–289. Griffiths, N., & Dobardziev, A. (2003). Fixed mobile substitution in the Netherlands, market research and high level strategic analysis for OPTA. London: Ovum Telecoms Research. Gruber, H. (2005). The economics of mobile telecommunications. Cambridge: Cambridge University Press. Gruber, H., & Verboven, F. (2001). The evolution of markets under entry and standard regulation—The case of global mobile telecommunications. International Journal of Industrial Organization, 19, 1189–1212. Grzybowski, L. (2005). Regulation of mobile telephony across the european union: An empirical analysis. Journal of Regulatory Economics, 28, 47–67. Grzybowski, L. (2011). Fixed-to-mobile substitution in the European Union. Working paper. South Africa: University of Cape Town. Hamilton, J. (2003). Are main lines and mobile phones substitutes or complements? Evidence from africa. Telecommunications Policy, 27, 109–133. Hamilton, J. D. (1994). Time series analysis. Princeton, NJ: Princeton University Press. Haucap, J. (2003). The economics of mobile telephone regulation. Discussion paper, Hamburg. Hausman, J. (2002). Mobile telephone. In M. Cave, S. Majumdar, & I. Vogelsang (Eds.), Handbook of telecommunications economics (Vol. 1, pp. 564–604). Amsterdam: Elsevier. Heimeshoff, U. (2008). Fixed–mobile-substitution in oecd countries: An empirical analysis on the subscriber level. Nuremberg: Friedrich-Alexander-University Erlangen-Nuremberg. Hodge, J. (2005). Tariff structures and access substitution of mobile cellular for fixed line in South Africa. Telecommunications Policy, 29, 493–505. Horvath, R., & Maldoom, D. (2002). Fixed–mobile substitution: A simultaneous equation model with qualitative and limited dependent variables. Discussion paper. London: DotEcon. Houthakker, H. S., & Taylor, L. D. (1970). Consumer demand in the United States: Analyses and projections. Cambridge, MA: Harvard University Press. Ingraham, A. T., & Sidak, J. D. (2004). Do states tax wireless services inefficiently? Evidence on the price elasticity of demand. Virginia Tax Review, 24, 249–261. International Telecommunication Society (ITU) (2010). World telecommunication/ICT indicators database (13th ed.), Geneva. Karacuka, M., Haucap, J., & Heimeshoff, U. (2011). Competition in Turkish mobile telecommunications markets: Price elasticities and network effects. Telecommunications Policy, 35, 202–210. Kaplow, Louis (2013). Market delineation: Impossible and counterproductive. Antitrust Law Journal, 79, 361–380. Laffont, J.-J., & Tirole, J. (2000). Competition in telecommunications. Cambridge, MA: MIT Press. Madden, G., & Coble-Neal, G. (2004). Economic determinants of global mobile telephony growth. Information Economics and Policy, 16, 519–534. Mao, C.-K., Tsai, H.-C., & Chen, C.-M. (2008). FMS patterns: Penetration vs. traffic substitution in different groups of countries. Technological Forecasting and Social Change, 75, 356–384. Merrill Lynch (2011a). Global wireless matrix, USA. Merrill Lynch (2011b). Global wireline matrix, USA. Munnell, A. H. (1992). Infrastructure investment and economic growth. Journal of Economic Perspectives, 6, 189–198. Narayana, M. R. (2010). Substituability between mobile and fixed telephones: Evidence and implications for india. Review of Urban and Regional Development Studies, 22, 1–21. Norton, S. W. (1992). Transaction costs, telecommunications, and the microeconomics of macroeconomic growth. Economic Development and Cultural Change, 40, 175–196. OECD (2011). OECD statistics. Paris: OECD Publishing. Peeperkorn, L., & Viertiö, K. (2009). Implementing an effects-based approach to article 82. Competition Policy Newsletter, 1/2009, Brussels: European Commission. Rhee, H., & Park, M. (2011). Fixed-to-mobile call substitution and telephony market definition in Korea. Journal of Regulatory Economics, 40, 198–218. Rodini, M., Ward, M. R., & Woroch, G. A. (2003). Going mobile: Substitutability between fixed and mobile access. Telecommunications Policy, 27, 457–476. Röller, L.-H., & Waveman, L. (2001). Telecommunications infrastructure and economic development: A simultaneous approach. American Economic Review, 91, 909–923. Srinuan, P., Srinuan, C., & Bohlin, E. (2012). Fixed and mobile broadband substitution in sweden. Telecommunications Policy, 36, 237–251. Suarez, D., & Garcia-Marinoso, B. (2013). Which are the drivers of fixed to mobile telephone access substitution? An empirical study of the spanish residential market. Telecommunications Policy, 37, 282–291. Sung, N. (2003). Fixed–mobile call substitution: Evidence from Korean long-distance markets. Working paper. Seoul: University of Seoul. Sung, N., & Lee, Y.-H. (2002). Substitution between mobile and fixed telephones in Korea. Review of Industrial Organization, 20, 367–374. Taylor, L. D. (1994). Telecommunications demand in theory and practice. Amsterdam: Kluwer Academic Publishers. Vagliasindi, M., Güney, I., & Taubman, C. (2006). Fixed and mobile competition in transition economies. Telecommunications Policy, 30, 349–367. Vogelsang, I. (2010). The relationship between mobile and fixed line communications. A survey. Information Economics and Policy, 22, 4–17. Ward, M. R., & Woroch, G. A. (2004). Usage substitution between mobile telephone and fixed-line in the US. Working paper, Arlington. Ward, M. R., & Woroch, G. A. (2010). The effect of prices on fixed and mobile telephone penetration: Using price subsidies as natural experiment. Information Economics and Policy, 22, 18–32. Ward, M. R., & Zheng, S. (2012). Mobile and fixed substitution for telephone service in China. Telecommunications Policy, 36, 301–310. Wengler, M. O., & Schäfer, R. G. (2003). Substitutionsbeziehungen zwischen Festnetz und Mobilfunk: Empirische Evidenz für Deutschland und ein Survey internationaler Studien. Bad Honnef: wik Diskussionsbeitrag Nr. 249. Wooldridge, J. M. (2010). Econometric analysis of cross section and panel data (2nd ed.). Cambridge, MA: MIT Press. Yoon, C.-H., & Song, Y.-W. (2003). Telecom development in Korea: Substitution and integration of fixed–mobile services and regulatory implications. Communications & Strategies, 5, 257–270.