The E ect of Universal Service Obligations on Public Pay-phone Use
In the European Union, public payphone regulation is provided for by the 2002 Universal Service Directive. aUnder Article 6 (1), the undertakings are required to provide a su cient number of payphones in public areas, in all parts of the territory, including areas where the market might not deliver these services, to meet the reasonable needs of end users. They also have to comply with several obligations related to prices and quality of service. In addition, calls to emergency services remain free of charge at payphones similarly to landlines and mobile phones. More speci c provisions, such as the minimum number of payphones to provide, depending on the population density, can additionally be imposed at the national level. Such coverage obligations have been de ned in ten out of the twenty-one EU Member States.4
However, with the ubiquity of the GSM (2G) network and the near ubiquity of the UMTS (3G) network, along with the development of competition in the telecommunication market, the costs of both xed and mobile phones have fallen signi cantly. Low-cost o ers, such as prepaid mobile tari s, along with social tari s based on income criteria have appeared. The acquisition of a telephony access has become more and more a ordable. In addition, with the increasing penetration rate of broadband, new ways of communications have emerged. National or international voice calls can now be made over the internet (VoIP) at an a ordable price or even at no cost.
As a result, the question arises as to whether market failures in the provision of communi-cation services still prevail in the European electronic communications market. Some European countries have examined how necessary public payphone services are perceived by the house-holds. For example, in 2013, based on article 6 (2) of the Universal Service Directive, the Belgian regulator decided to remove public payphones from the scope of universal service, considering that the mobile phone constituted a comparable and a ordable service, available in all parts of the territory. Universal service obligations do not a ect directly the demand for public payphones, but the supply-side (i.e., the provision of public payphones). As undertakings are compelled to provide a su cient number of public payphones, the demand is indirectly impacted. Therefore, the question as to whether universal service obligations have an impact on public payphone use is addressed by analysing how the demand evolves depending on the existence of universal service obligations.
To the extent that universal service obligations are binding for the rms, the supply of pay-phones, and hence their use, is expected to be more important with universal service obligations than without. However, this would not mean that universal service obligations are still neces-sary. Therefore, though a positive impact can be expected, it is also expected to be rather low. The average data show that public payphones are used predominantly by mobile phone owners when they cannot use their mobile phone, either because it is out of range, out of credit or out of battery. Yet, public payphone use is expected to be independent of mobile phone ownership. In addition, as public payphones are expected to be used principally by low-income households or households living in rural areas, a higher (more positive) impact of universal service obligations on these household categories is also expected.
To test the e ect of universal service obligations on public payphone use, the following linear model has been de ned:
P ayphone useit = 0 + 1regit + 2altertechit + 3pricesit + 4jobit (1) + 5areait + 6year + 7 it + 8 it + it:
where i is a household and t a time period. Household i chooses the number of calls to make from a public payphone in year t (P ayphone useit). Their choice is assumed to depend indirectly on the regulatory policy, (regit), on the alternative communication technologies owned, either xed or mobile (altertechit), and on the costs of using them, pricesit. Besides, the households choice is expected to be in uenced by the socio-professional group, jobit, and the area of living, areait. Two sets of control variables are added to the model: it controls for di erences in demographic characteristics, and it for cross-country di erences. Finally, year captures a potential time trend and it represents the error term.
Then, to estimate whether the e ect of universal service obligations varies depending on the alternative communication technologies owned, the socio-professional category of the household
and its geographic area, the following variant model has been de ned:
P ayphone useit = 0 + 1regit + 2altertechit + 3altertechit regit (2)
+ 4pricesit + 5jobsit + 6jobit regit + 7areait
+ 8areait + regit + 9year + 10 it + 11 it + it
As the e ect of universal service obligations (regit) is expected to depend on the value of other independent variables, three sets of interaction variables are added to Model (2). The regulation variable is interacted with the 4 di erent alternative communication technologies(altertechit regit), the 8 di erent job categories (jobit regit), and the 3 types of areas in which the household is settled (areait regit).5
The relationship between, on the one hand, public payphone use and on the other hand, the alternative technology, the job and the area of living is conditional to the presence or absence of universal service obligations. The coe cients of the interaction terms ( 3; 6; 8) represent the di erence in the e ect of the alternative technology owned, the job held and the area of living, respectively, between households living in countries where public payphone provision is regulated (regit = 1) and households living in countries where no regulation is imposed (regit = 0).
Finally, a counterfactual model predicts what would happen, in terms of public payphone use, if universal service obligations were absent. By setting universal service obligations to zero in countries where such obligations were imposed, the model predicts the average public payphone use in these countries if no regulation had been imposed. By comparing this prediction with the observed average public payphone use in the same population, the expected average regulation e ect is obtained.
Let (P ayphone use1it) denote the potential public payphone use by household i at time t if the household lives in a country where a regulatory policy is imposed, and (P ayphone use0it) denote the potential public payphone use absent any universal service obligations. For each household, the equation writes:
P ayphone useit = P ayphone use0it + usoit(P ayphone use1it P ayphone use0it): (3)
where P ayphone use1it stands for the households living in countries imposing universal service obligations and P ayphone use0it stands for the other households.
Let E[P ayphone use1it P ayphone use0it] denote the expected average regulation e ect, and E[P ayphone use1it P ayphone use0itjusoit = 1] the expected average regulation e ect for the households living in a country imposing universal service obligations. This expression can be rewritten:
E[P ayphone use1it P ayphone use0itjusoit = 1] = E[P ayphone use1itjusoit = 1] (4)
E[P ayphone use0itjusoit = 1]:
The rst term represents the average number of public payphone calls per year in the population living in countries imposing a regulation, which is observable. The second term is the average number of public payphone calls per year in the population living in countries imposing a regulation if no regulation were imposed, which is unobserved.
The models are estimated with ordinary least squares (OLS). Another option would have been to use a discrete choice model. The models are also estimated with an ordered logit using a categorical dependent variable on the frequency of use. The estimation results are similar to those obtained with OLS. However, the ve modalities of the dependent variable are unbalanced, as the fth modality representing the proportion of households, which never uses a public payphone, concentrates 87 percent of the households. The remaining 13 percent are shared between the four rst modalities. As a result, the models have been estimated with a simple logit model. A dummy variable has been created. It takes the value 1 if household i has used a payphone during year t and 0 otherwise. The same qualitative results have been obtained.6
In all speci cations, cluster (at country level) standard errors that are robust to heteroskedas- ticity are reported.
The main data on public payphone use have been extracted from 4 annual E-communication household surveys published by the European Commission from 2005 to 2009.7 These surveys have been conducted in 27 European countries and constitute micro-level pooled cross-sectional data on 106,989 households overall over the four years. For each country, there are approximately 1,000 observations per year (around 500 for small countries). The data provide information on public payphone use by indicating how often households use them. The data also include in-formation on the telecommunication equipment available in the household, such as the di erent telephony accesses or the existence of an internet connection, as well as information on individ-uals characteristics, such as age, gender, area of living and job.
This database has been completed with four other sources. First, the variables on the state of regulation in European countries have been constructed from regulatory policy documents published by NRAs and BEREC. When the information was not available or only in the national language, it has been obtained by contacting the NRA or in some cases the national operator in charge of providing universal service.
Second, data on the prices of mobile and xed telecommunication services have been ex-tracted from the Telecoms Price Developments Reports. These reports are produced on an an-nual basis for the European Commission Directorate General for Information Society by Teligen. They give the price of di erent composite baskets of telecommunications services, either xed or mobile, in the Member States from 1998 to 2010.8 Third, data related to EU countries such as the density of population, the unemployment rate and the gross domestic product (GDP) per capita come from Eurostat. This third set of data is measured at the NUTS 2 level, which is the basic regional level for the application of regional policies. Finally, data on mobile phones and xed phones subscriptions have been collected from the ITU databases for the EU Member States for the years 2000 to 2012.
Table A and Table B in Appendix A present the list of the variables used in the estimation and some summary statistics. Following the European Commission methodology, in order to take account of the di erences that could remain between East and West Germany, these regions have been separated in the analysis. The same is true for Great Britain and Northern Ireland. The following paragraphs describe the variables used in the empirical analysis.
The dependent variable, P ayphone useit, is a continuous variable representing the average number of public payphone calls made by household i in year t. This variable has been con-structed from a categorical variable measuring the frequency of use. The frequency of use has ve categories: once a week or more often; about once a month; less often than once a month; when away from home or travelling; never. The dependent variable has been built by computing for each category the number of payphone calls depending on the number of days, weeks and months in a year. Di erent values have been tested for each category, however it did not change the sign or the signi cance of the coe cients in the regression. The following values have been chosen: Once a week or more often: 60 calls per year (approximately every 6 days); About once a month: 17 calls per year (app. every 3 weeks); Less than once a month: 4.5 calls per year (app. every 2 to 5 months or every 2.7 months); When away from home or travelling: 2 calls per year (app. every 6 months); Never: 0 call per year.
The regulatory variables, regit, consist of two dummies. The rst one, usoit, takes the value 1 if universal service obligations for public payphone provision have been imposed in households i country in year t, and 0 otherwise. The second one, covit, takes the value 1 if universal services obligations imposed a minimum number of payphones to provide, depending on the density of population, in households i country in year t, i.e., coverage obligations and 0 otherwise.
As the alternative communication technologies can be seen as a substitute for public pay-phones, their impact on public payphone use has been estimated. Three dummy variables measure the availability of a telephony access in the household: moit takes the value 1 if house-hold i owns a mobile phone but no xed telephone in year t (mobile only), and 0 otherwise;f oit takes the value 1 if household i owns a xed telephone but no mobile phone in year t ( xed only), and 0 otherwise. Finally, f mit takes the value 1 if household i owns both a mobile phone and a xed telephone in year t ( xed and mobile), and 0 otherwise.
A dummy variable, internetit, takes the value 1 if household i has an Internet connection at home in year t, and 0 otherwise. Nowadays, the Internet represents a gateway providing a wide range of communication services. Voice calls can be achieved through VoIP services, such as Skype, and it is also possible to communicate via e-mail or via more elaborated communication services o ered by social networks. The vector of alternative communication technologies is then:
altertechit = moit; f oit; f mit; internetit: (5)
The variables on the alternative telephony technologies prices are de ned as follows. The variable f ixed rentalit represents the monthly PSTN rental charge (VAT included) in euro for household i in year t. A variable on the average price of a basket of xed-telephone calls in euro, f ixed priceit, has been added, as well as a variable on the average price of a basket of mobile calls in euro, mobile priceit. Obviously, these prices vary across countries, but not across households within a given country. The vector of prices is then: pricesit = f ixed rentalit; f ixed priceit; mobile priceit: (6)
Universal service is considered as a safety net for vulnerable end users (notably low-income households) and as a means to promote rural development, by ensuring that households located
in remote areas have access to communication services. Therefore, the impact of the socio pro-fessional group and of the type of area of living have been estimated. The jobit variable has 8 modalities, while the areait variable has 3 modalities: rural area or village, small or middle-sized town, and large town. There is no quantitative criterion to de ne the type of area. This is a qualitative variable based on the respondent perception.
The group of socio-demographic variables, it, includes a continuous variable as well as a dummy variable indicating the ageit and the gender, womanit, of household is respondent in year t. The vector of socio-demographic characteristics is then: it = ageit; womanit: (7)
The second group of variables, it, controls for cross-country di erences by including country characteristics, measured at the NUTS 2 level, namely, the density of population, densit, which may a ect network costs, the unemployment rate, unemplit, and the gross domestic product per capita (gdpit), which may a ect the demand for communications services. Variables on the national level of mobile and xed telephones subscription per 100 inhabitants, mobile subit and f ixed subit, are also included. To take account of di erences between Eastern and Western countries, a dummy variable, EU15, identi es the rst 15 countries in the EU. These control variables are commonly used in the academic literature dealing with the use of information technologies.9 The vector of country characteristics is then: it = densit; unemplit; gdpit; mobile subit; f ixed subit; EU15it: (8)
Finally, four dummy variables have been created for the years 2005, 2006, 2007 and 2009 to take into account a potential time trend in the use of public payphones.
The rst sub-section presents the empirical results for the baseline model (Model (1)), which estimates the impact of regulation on public payphone use. The second sub-section presents the empirical results for Model (2), which estimates whether imposing universal service obligations enhances the use of public payphones by vulnerable end users and contributes to regional and rural development.
Impact of Universal Service Obligations on Public Payphone Use
The following table presents the estimation results for Model (1), which estimates the general impact of universal service obligations (1a) and the speci c impact of coverage obligations (1b).
The estimation results show that the coe cient of universal service obligations is positive, but not statistically signi cant. Therefore, it cannot be concluded that universal service obli-gations have an impact on public payphone use. By contrast, column (1b) shows that coverage obligations have a positive and signi cant impact on payphone use. On average between 2005 and 2009, households living in countries imposing coverage obligations have made 0.606 public payphone calls more per year than households living in the other countries.
The coe cient of each alternative technology is negative and signi cant, as expected. The highest substitution e ect comes from the bundle of xed and mobile phones, meaning that households subscribing to both services are less inclined to use payphones. Households having a mobile phone only use public payphones slightly more than those having a xed telephone only. To understand this result, which may sound rather surprising, note that the main reason explaining public payphone use is the impossibility to use ones mobile phone.10
Similarly, households having an internet connection at home are less likely to use public payphones. This can result from the existence of two substitution e ects. The rst substitution e ect appears due to a lower marginal cost of an IP call compared to a public payphone call. The second substitution e ect operates when a member of a household is away from home and wants to call at his place. He can then decide to call through the IP network if he has internet at home. Hence, with the emergence of cheap or even free VoIP services, such as Skype, some households have started to substitute public payphones with the Internet.
Not surprisingly, public payphone use slightly increases with the price of mobile calls. On average, a one-euro increase in the price of mobile calls raises the number of payphone calls by 0.062 per year. Yet, the price of xed calls has an unexpected negative impact. This could be explained by the similar pricing scheme applied to call origination departing from the xed telephone network and from public payphones: higher prices for xed telephony may also imply higher prices for payphone calls, leading to a lower use of payphones.
Some socioeconomic variables are signi cant in explaining the heterogeneity across house-holds and countries in the use of payphones. Western European households tend to use public payphones more intensively than their Eastern neighbours. Public payphones use also decreases with the age of the household, but increases with the unemployment rate and to a lesser extent with the density of population. This latter result is con rmed by the relatively lower use of public payphones by households living in less urbanised areas compared to those living in large towns. GDP per capita, which is a proxy for income, is not signi cant. The income e ect seems to be better captured at the household level. Public payphone users are slightly more prevalent among low-income households, such as manual workers and particularly unemployed.
Impact of Universal Service Obligations on Vulnerable End Users and on Regional and Rural Development
Model (2) estimates whether the impact of universal service is di erent depending on the al-ternative communication technologies owned, the socio-professional group of the household and its area of living. The estimation results are provided in column (2a) when universal service obligations are imposed and in column (2b) when speci c coverage obligations have been de ned.
Table of contents :
I General Introduction
1 Regulation of Old Technologies
2 Impact of Regulation on Competition and Investment
3 Impact of Very High Speed Broadband Networks on Local Economic Growth
4 Research Questions and Results
4.1 Universal Service Obligations and Public Payphone Use: Is Regulation Still Necessary in the Era of Mobile Telephony?
4.2 Unbundling the Incumbent and Entry into Fiber: Evidence from France
4.3 Impact of Very High-Speed Broadband on Local Economic Growth: Empirical Evidence
5 Thesis Plan 7
II Universal Service and Public Payphone Provision
2 The Eect of Universal Service Obligations on Public Payphone Use
3 Econometric Model
4 The Data
5 Estimation Results
5.1 Impact of Universal Service Obligations on Public Payphone Use
5.2 Impact of Universal Service Obligations on Vulnerable End Users and on Regional and Rural Development
5.3 Counterfactual Analysis
III Entry into the Fiber Market
2 Literature Review
3 Broadband Market in France
4 The Data
5 Econometric Models
5.1 LLU Entry
5.2 Fiber Entry
6 Estimation Results
IV Very High-Speed Broadband Networks and Local Economic Growth
2 Literature Review
3 The Data
3.1 Dependent Variables
3.2 Very high-speed Broadband Networks
3.3 Explanatory Variables and Controls
4 Econometric Strategy
4.1 Empirical Approach
4.2 Potential Endogeneity
4.3 Empirical Models
5 Estimation Results
5.1 Robustness Checks
5.2 Matching Estimator
5.3 Panel Data Model with Fixed Eects and Dierence-in-Dierences for Deployments in 2013
V General Conclusion
1 Review of the Main Results
1.1 Universal Service Obligation and Public Payphone Use
1.2 Local Loop Unbundling and Entry into Fiber
1.3 Impact of Very High-Speed Broadband on Local Economic Growth
2 Future Projects