BRUEGEL POLICY CONTRIBUTION

ISSUE 2015/13 JULY 2015 ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS

MARIO MARINIELLO AND FRANCESCO SALEMI

Highlights

• Mobile telecommunications markets are an important part of the European Com- mission’s strategy for the completion of the European Union Digital Single Market. The use of mobile telecommunications – particularly mobile data access – is gro- wing and becoming an increasingly important input for the economy. • The EU currently does not have a unified mobile telecommunications market. The EU compares favourably to the United States in terms of prices and connection speed, but lags behind in terms of coverage of high-speed 4G wireless connections. • Europe’s long-term goal should be to make data access easier by increasing high- speed wireless coverage while keeping prices down for users. An increase in cross-border competition could help to achieve that goal. • The Commission has two important levers to help stimulate cross-border supply: (a) ensuring competition in intra-country mobile markets in order to provide an incentive for operators to expand into other jurisdictions, and (b) reducing mobile Telephone operators’ costs of expansion into multiple EU countries. The further development +32 2 227 4210 of policies on international roaming and radio spectrum management will be central [email protected] to this effort. www.bruegel.org Mario Mariniello ([email protected]) is a Research Fellow at Bruegel. Francesco Salemi ([email protected]) is a Research Assistant at Bruegel. The authors wish to thank Serafino Abate, Antonios Drossos, Stephen Gardner, J. Scott Marcus and Guntram Wolff for helpful comments. Research assistance by Afrola Plaku is gratefully acknowledged. BRUEGEL POLICY CONTRIBUTION ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS 02

ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS

MARIO MARINIELLO AND FRANCESCO SALEMI, JULY 2015

1 INTRODUCTION 2013, pp92-93); the vast majority of mobile traffic will soon be generated by 4G connec- The completion of the Digital Single Market (DSM) tions (Cisco, 2015). is one of the top priorities for the European Com- • Mobile broadband, or wireless internet access, mission under Jean-Claude Juncker. On 6 May could soon become a valid substitute for wired 2015, the Commission published a strategy out- broadband access for most typical internet lining how it intends to achieve that goal (Euro- uses3, in particular in low population density pean Commission, 2015). According to the areas where building fixed infrastructure might strategy, the completion of the DSM “could con- not be economically sustainable. tribute €415 billion per year to [the EU] economy • Mobile broadband technologies are developing and create 3.8 million jobs”1. rapidly and although there is still uncertainty about the details of the next generation (5G) A major plank of the strategy is addressing frag- wireless standard, the allocation and assign- mentation in the telecoms sector: access avail- ment of dedicated spectrum bands might start ability, quality and prices vary significantly across as early as 2020. the continent, with telecoms markets defined by national borders. Users’ access conditions are The fundamental question is how European largely determined by their place of residence. The mobile markets can be improved for the benefit of Commission's initial strategy document does not users. The often-heard answer is that barriers to yet offer any concrete solutions to this, but indi- cross-border competition should be gradually dis- 1. http://ec.europa.eu/priori- ties/digital-single-market/. cates areas for potential future intervention. mantled in order to move towards a pan-European market for mobile services. 2. Lam and Shiu (2010), for example, estimate that the In this Policy Contribution we specifically look at growth in mobile penetra- EU mobile telecoms markets and analyse poten- Pan-European networks imply lower production tion rates significantly tial concrete measures that could contribute to the and possibly network deployment costs, resulting affected total factor produc- Commission’s digital strategy goals of improving from economies of scale. This should imply lower tivity growth in a number of countries between 1995 end-users’ access conditions and addressing EU prices in the short-term and more investment in and 2004. They also found market fragmentation through the development of the long-term, leading to increased high-speed a two-way relationship cross-border supply of services. There is no appar- mobile broadband coverage. between mobile penetration ent structural reason why the supply of mobile rates and real GDP growth in services should stop at EU member states’ Avoiding the multiplication of networks would also these countries between 1997 and 2006. national borders. For the provision of mobile serv- reduce ‘double mark-up’ effects: when more than 3. Grzybowskiy and Ver- ices, wireless infrastructure is needed. We focus one network is needed to provide a service, for boven (2014) note that, on this for a number of reasons: example in the case of international calls, there is especially in recent years, a natural tendency to higher prices. Each network mobile broadband in EU • The diffusion of mobile telecommunication has owner chooses how much to charge for terminat- markets has been per- been shown to be a significant factor in improv- ing calls on its network and wants to maximise ceived as a potential substi- 2 tute for fixed broadband. ing productivity . only its own profits without considering the nega- The UK telecoms regulator • Mobile data consumption is growing rapidly tive effect that such choices could impose on the Ofcom found that there is a because of the fast take-up of smartphones profits of other network owners. The higher the growing positive gap and tablets (even though a large part of this price, the lower the demand will be for a comple- between mobile data rev- enue and fixed broadband traffic is being offloaded to Wi-Fi connections mentary good – in this case the call origination on revenue (Ofcom, 2014). at home or at work) (European Parliament, other networks, which are also needed to make BRUEGEL POLICY ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS CONTRIBUTION 03 the call. Cross-border networks operated by single for mobile telecommunications should be to allow operators would limit that phenomenon and ulti- differences in price and quality of service only if mately exert a downward pressure on tariffs. they relate strictly to differences in supply (ie Opening the borders would also mean increasing costs) and demand characteristics. In the long competitive pressure on national markets, with term, such an approach could be expected to lead users given access to a wider choice of operators. to converging tariffs across the continent, insofar as the progressive completion of the single market This would not necessarily mean that a uniform as a whole (not only the DSM) will imply an tariff for all EU users should emerge in such a increased convergence in the levels of purchas- market, nor that the Commission should impose ing power and production costs in EU countries. such a price. As long as significant structural dif- ferences between EU countries continue to exist, To identify how the Commission's goal of a single requiring uniform prices could harm customers market for mobile services might be achieved, we with a lower ability to pay, ie customers from first look at EU mobile markets in comparison with lower-income countries4. Figure 1 shows the aver- the US. We then examine how improved wireless age mobile operator revenue per user (ARPU, a access to data by final users, increased fast measure commonly used as a proxy for unit price mobile broadband coverage and lower prices could for mobile services) and the average hourly salary be stimulated by greater cross-border competition. per person in 2013 for each EU country (except We then show how policies on international roam- Austria). The correlation between the two variables ing and radio spectrum management could have is very high. It would be hard to imagine Bulgarian an impact on cross-border competition. customers paying the same mobile prices as cus- tomers from Luxembourg. 2 EU MOBILE TELECOMMUNICATIONS MARKETS

To reach its goal, the Commission should aim to Mobile market structure ensure that markets are competitive and exposed to a similar level of competition across the conti- There are about 40 mobile network operators nent. Customers from any country could be able (MNOs) in the EU. Many operate in just one or two to choose from a set of potentially EU-wide service countries. A restricted group of big international providers and possibly other suppliers with a local or regional focus (a scenario closer to that in the Table 1: Presence of mobile network operators US). In other words, the Commission's objective in EU countries Figure 1: Average hourly remuneration and MNO(s) Number of countries average revenue per user, EU countries, 2013 Vodafone 12 Deutsche Telekom 8 300 LU Orange, TeliaSonera 7 IE Hutchison 6 250 Tele2 5 CY UK SE 4. Differences in prices NL DK Telekom Austria, Telenor 4 200 BE would also be expected in SI ES DE FR Telefónica 3 MT more competitive markets, SK EU FI KPN, Belgacom, BITE, Elisa, OTE 150 2 because of differences in CZ EL (40% DT), PPF

ARPU (€) HU IT the cost of providing mobile HR PT Bouygues, Bulgaria Telecom, services in different 100 PL CYTA, DNA, Eircom, Everything countries. EE Everywhere (50% DT, 50% BG LT 50 Orange), Go, Iliad, Luxembourg 5. Even though none has a RO LV Online (LOL), Melita, MTN, NOS network that covers the Comunicações (formerly 1 entire land area or popula- 0 Optimus), Play, Polkomtel, tion of the US, each covers 0 5 10 15 20 25 30 35 40 45 Portugal Telecom, POST Average hourly remuneration (€) Luxembourg, RCS-RDS, SFR, more than 99 percent of the Source: Bruegel based on Eurostat and Digital Agenda TDC, Telecom Italia, Teledema, US population. See FCC Scoreboard. Note: Austria is missing because the ARPU value Telekom Slovenije, Tušmobil, (2014), p7. is not available. The correlation between the two variables is VimpelCom, Wind Hellas 6. FCC (2014), Table II.C.2, 0.81 (with 1 indicating a perfect positive correlation). Source: Bruegel based on Rewheel’s Digital Fuel Monitor. p16. BRUEGEL POLICY CONTRIBUTION ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS 04

companies (Vodafone, Deutsche Telecom, Telia- multi-regional operator (US Cellular) and several Sonera, Orange, Hutchison) have a larger Euro- regional and local providers. pean footprint, but nowhere near complete EU coverage (Table 1). By comparison, in the US there Some companies in Europe (eg Tele2, TeliaSonera, are four nationwide MNOs5 (AT&T, Verizon, Sprint Telenor) tend to concentrate on specific regions, and T-Mobile) which accounted for 95.3 percent of such as Nordic and eastern European countries. US mobile revenues in 20136. The US also has one Table 2 shows companies’ market shares across 7. While the financial per- formance of European tele- Table 2: MNO market shares (% of SIM) in EU countries, Q1 2014 coms operators should not AT BE FR DE IE IT LU NL PT ES UK be a primary policy concern Pre* Post** Pre* Post** in itself, it might be relevant Vodafone 29.5% 29.5% 40.9% 40.9% 26.4% 30.3% 39% 28.8% 25.2% to the extent that worse Telefonica 17% 37.8% 29.2% 42.3% 29.5% financial performance Deutsche Telekom 33.8% 32.6% 32.6% 26.1% might result in poorer serv- Everything Everywhere*** 33.6% ices to users, because, for Orange 27.7% 41.2% 10.6% 22.9% instance, of insufficient Hutchison 22.9% 9.3% 38.5% 11% 11.7% investment in the deploy- ment, maintenance or KPN 28.2% 20.8% 43.6% improvement of network Belgacom 44.1% 37% infrastructure or reduced POST Luxembourg 51.7% investment in new tech- Luxembourg Online (LOL) 0.7% nologies that would allow Telecom Italia 37.5% better utilisation of the VimpelCom 25.1% available resources (eg Telekom Austria 43.3% spectrum, base stations). SFR 32.3% 8. Bruegel calculation Bouygues 17.1% based on Eurostat and the Iliad 9.4% US Department of TeliaSonera 6% Commerce. Eircom 20.6% 20.6% Portugal Telecom 45.4% 9. The European Commis- NOS Comunicações 15.5% sion recently opened an in- depth investigation of a BG HR CZ DK EE FI GR HU LV LT PL RO SK SI SE proposed merger between Deutsche Telekom 46.6% 40.3% 44% 28.4% 33.9% TeliaSonera and Telenor in TeliaSonera 22.9% 45.8% 34.3% 41.6% 34.8% 46.2% Denmark – presented as a Telenor 34.7% 16.6% 32.2% 16.7% joint venture between the Tele2 13.5% 27.6% 40.3% 42.8% 26.1% Danish operations of the Vodafone 23.7% 30% 23.8% 31.4% two companies Telekom Austria 43.5% 40% 30.3% (http://europa.eu/rapid/pres Orange 27.4% 39.7% 43.2% s-release_IP-15- Elisa 26.6% 40.4% 4749_en.htm). Other deals Hutchison 9.7% 10.9% have been either formally OTE (40% DT) 49.2% 23.8% announced, pending formal PPF 36% 22.9% clearance from antitrust BITE 18.1% 20.4% authorities, eg the acquisi- tion of O2 (Telefónica) by DNA 25.3% Hutchison in UK TDC 50.8% (http://www.techweekeu- Play 18.7% rope.co.uk/mobility/4g/three Polkomtel 25.6% -o2-hutchison-whampoa- Wind Hellas 20.7% 167859), or are rumoured, Telekom Slovenije 58.2% eg the merger between Tušmobil 11.5% Hutchison and Wind in Italy RCS-RDS 5.1% (http://telecoms.com/42132 Teledema 2.1% 1/hutchison-and-vimpel- Bulgaria Telecom 21.8% com-in-talks-to-merge-3- Source: Bruegel based on Rewheel’s Digital Fuel Monitor. Note: Colours indicate market share rank in each country: red = largest market italia-and-wind/) that would share, orange = second largest, blue = third largest; green = fourth largest. * Pre and ** Post: data for Germany and Ireland shows the situa- further reduce the number of tion both before and after the Hutchison 3G UK/Telefónica Ireland and Telefónica Deutschland/E-Plus mergers. Data for each country might operators in these markets. not add up to 100% because of rounding. Malta and Cyprus are not included. *** 50% DT, 50% Orange. BRUEGEL POLICY ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS CONTRIBUTION 05

Europe. Because of different regulatory frame- denly adopted. In average real income per capita works, however, even companies that operate terms (in purchasing power standard), the ratio multiple networks in neighbouring countries oper- between the richest and the poorest US states is ate the network in each country on a stand-alone 1.73. In Europe, the ratio is 5.71 (or 2.91 excluding basis, so that it is fair to speak about EU national Luxembourg)8. markets rather than a unified EU mobile market in which users can buy mobile access from other In the last two decades, Europe and the US went operators active on the EU territory. The fragmen- through broadly similar market restructuring tation of the EU market – and the resulting smaller processes, with a series of merger and acquisition scale of operation – has been identified as one of deals that significantly increased the level of con- the factors behind the worse financial results of centration in the market after the entries of new European telecoms companies compared to their operators in the mid-1990s and the beginning of 7 US, Japanese and Korean counterparts . It should 2000s. Figure 2 shows the major merger events 10. In the US, after entries be noted however, that the absence of a homoge- in EU mobile markets from 2003 to 20159. Figure by operators made possible neous regulatory regime is not the only explana- 3 shows major merger events in the US mobile by the spectrum awards tion for differences in outcomes, such as user industry from 2004 to 201410. Mobile telecom- determined in the first access prices, in Europe. For example, supply and munication markets in EU member states and in series of large spectrum auctions (Broadband PCS) 11 demand conditions vary significantly across the US are now similarly concentrated . during the mid 1990s, and a Europe, much more than across the US, and dif- series of acquisition that ferences in prices might be justified even if a gave rise to operators with homogeneous regulatory framework was sud- nationwide footprints, two mergers in 2004 reduced Figure 2: Merger events in the EU, 2003-15 the number of nationwide operators from six to four, GERMANY: and successive mergers NETHERLANDS: Acquisition of E-Plus (KPN) KPN acqui res Telfort NETHERLANDS: by O2 (Telefonica) eliminated competition T-Mobile cleared with remedies AUSTRIA: acquires Orange from multi-regional opera- Mobilkom Austria AUSTRIA: UK: AUSTRIA: DENMARK: tors, eg MetroPCS acquired T-Mobile T-Mobile UK and Orange UK (Telekom Austria) H3G Austria’s European acquisition of me rge (cleareanc e with reme dies) by T-Mobile in 2013, Leap acquires 3G Mobile acquisition of Commission Tele.ring as Everything Everywhere (EE) Orange Austria opens in-depth Wireless acquired by AT&T cleared with cleared wi th reme dies investigation of reme dies TeliaSonera DK/ in 2014. This left US Cellular Telenor DK merger as the only multi-regional provider of mobile services 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 that has not yet been acquired by a nationwide DENMARK: provider. TeliaSonera DK GREECE: acquires Orange DK Vodafone abandons its attempt to merge with Wind Hellas because of concerns 11. According to the stan- GREECE: regulators would not approve it dard thresholds based on TGP IV and Apax, which jointly control IRELAND: the Herfindahl-Hirschman TIM Hellas, acquire Q-Telecommunications Acquisition of O2 (Telefonica Ireland) Index (HHI), which is com- by H3G cleared with remedies monly used to measure Source: Bruegel. market concentration. In Figure 3: Merger events in the US, 2004-14 the EU the average HHI (weighted by population) Acquisition of Softbank (Japan) acquires Cingular Wireless acquires Rural Cellular by Sprint Nextel and takes over Clearwire was 3216 in Q1 2014, indi- Alltel acquires AT&T Wireless and Verizon cleared with AT&T announces to create Sprint cating a high level of con- adopts the AT&T brand Midwest Wireless divestitures to AT&T its intention to acquire T-Mobile centration (Bruegel based AT&T acquires Acquisition of Centennial T-mobile merges AT&T acquires Alltel acquires Dobson by AT&T cleared with on Rewheel’s Digital Fuel AT&T abandons bid with Leap Wireless Western Wireless Communications divestitures to Verizon for T-Mobile MetroPCS Monitor); in the US the aver- age HHI weighted across 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Economic Areas was 3027

Acquisition of Alltel by Verizon Sprint gives up in 2013 (about a 40 percent FCC recommends T-Mobile acquires cleared with divestitures to takeover of non-approval of increase from an HHI of Sprint and Nextel merge SunCom Wireless AT&T and Atlantic Tele-Network T-mobile in face forming Sprint Nextel AT&T/T-Mobile deal of regulatory 2151 in 2003). Source for DOJ formally resistance US figures: FCC (2014), Sprint Nextel and Clearwire Corporation form a JV asks to block AT&T AT&T acquires combining their WiMax businesses into a takeover of T-Mobile Atlantic Tele-Network Chart II.C.1, p17; for 2003 new company, named Clearwire wireless business figures, FCC (2011), Table 9, Source: Bruegel. p47. BRUEGEL POLICY CONTRIBUTION ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS 06

Mobile end-users’ prices esting to note that price differences cannot be explained by higher download speeds. The Organisation for Economic Cooperation and Development publishes every two years data on Network investment prices for mobile services. The most recent data was collected between August and September Generally speaking, the telecommunications 2012 (OECD, 2013). Several EU countries had industry is characterised by large fixed costs lower prices than the US. This was the case in par- related to the acquisition of spectrum licenses and ticular in those countries where there exists at the roll-out of networks with sufficient geographic least one ‘challenger’ that does not compete in coverage and capacity (bandwidth), and by small other EU markets with the very same operators variable costs of providing actual services. In active in that country (eg Finland, Estonia and absolute and per capita terms, the US invests Poland) or where Hutchinson was present (eg Aus- more than EU countries in telecommunications tria, Denmark, Sweden and the UK). Before it network infrastructure. EU operators have in the embarked on a series of acquisitions, Hutchinson past indicated that their apparent investment was a ‘challenger’ in all countries where it oper- underperformance is a consequence of smaller ated, being consistently the smallest operator revenue streams compared to US operators. Lower with market shares well below 20 percent, and EU mobile telecoms revenues are likely to be often just around 10 percent. The difference in partly explained by the greater maturity of EU mar- prices between US and Europe is particularly kets: between 2005 and 2013, mobile cellular acute when considering services that include subscriptions per 100 people (the penetration data. Figure 4 shows download speed and prices rate) increased from 96 to 125 percent in Europe of wireless broadband in OECD countries. It is inter- and from 68 to 96 percent in the US (Bruegel Figure 4: OECD wireless broadband basket, Sept 2012, Tablet 250MB (top) and Tablet 2GB (bottom)

$35 90 $33.20

80.0 80.0 Speed Mbit/s Total $ PPP $30 80 70 $25

$20.60 60

$20 50 Mbps $16.49 42.2 42.0 42.0 42.0 $15.05 40.3 $14.99 $13.05

$13.16 40 $12.72 $13.27

$ PPP $15 $11.75 $12.02 $11.68 $11.62 $11.86 $12.29 $11.95 $11.02 $10.67 $10.95 $10.91 $9.63 21.6 30 $8.82 $8.28 $8.52 $8.51 $8.15 $7.84 $10 21.6 $6.84 $6.51 $6.31 14.4 $5.99 20 16.1 $5.67 $5.65 14.4 $5.04 $5 $4.81 7.2 7.2 7.2 7.2 7.2 7.2 7.2 7.2 7.2 6.0 10 5.0 5.0 3.6 3.6 3.6 3.6 2.8 2.2 2.0 2.0 2.0 1.0 $0 0.5 0 UK Italy Chile Israel Spain Korea Japan Turkey France Poland Ireland Austria Greece Mexico Iceland Finland Canada Estonia Norway Sweden Belgium Portugal Slovakia Slovenia Hungary OECD av. Australia Denmark Germany Czech Rep. Czech Switzerland Netherlands Luxembourg New Zealand United States

$40 45 42.0 42.0 42.2 42.0 40.3 Speed Mbit/s Total $ PPP $34.45 40

$35 $33.20 $33.00 $32.73 $30.97 $31.96

$30 $27.75 35

$25.48 30 $23.99 $25 $23.90 $22.64 $22.29 Mbps

$20.60 25 21.6 21.6 21.6 $17.70 $18.65 $18.58 $20 21.0 $17.75

$15.51 20 $ PPP $15.05 $14.79 16.0 $13.05 $12.94

$15 14.4 $12.63 $12.29 $12.55 14.4 $12.06 13.5 $12.02 $11.99

$11.83 15 $10.28 $10.47 $9.94 $11.12 10.8 $10 $9.04 7.2 7.2 $7.11 7.2 7.2 7.2 7.2 7.2 10 7.2 7.2 7.2 7.2 6.0 5.0 4.5 3.6 $5 3.6

2.8 5 2.0 2.0 1.0 $0 0.5 0 UK Italy Chile Israel Spain Korea Japan Turkey France Poland Ireland Austria Greece Mexico Iceland Finland Canada Estonia Norway Sweden Belgium Portugal Slovakia Slovenia Hungary OECD av. Australia Denmark Germany Czech Rep. Switzerland Netherlands Luxembourg New Zealand New United States Source: Bruegel based on OECD. The panels show (i) the price of the least costly options in OECD countries for baskets of wireless broadband which include total charges for 250MB (top) and 2GB (bottom) of data for tablet use per month in USD PPP (left axis) and (ii) the broadband speed of the contract in megabits per second (right axis). The correlations between prices and download speeds in the two figures are about 0.09 for the 250MB basket and 0.17 for the 2GB basket, suggesting little relationship between the two variables. BRUEGEL POLICY ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS CONTRIBUTION 07 based on World Bank). The increase in revenues in An analysis of OECD countries for the latest avail- the US is mainly due to the increase in penetration able year shows a high correlation between rates, similar to what European operators experi- mobile revenues and investment in per capita enced during the mid-2000s (Figures 5 and 6). terms (Figure 7). However, the correlation does not say much about the existence or direction of a Figure 5: Mobile revenue in the EU and US, € causal relationship. Both per capita investment billions, 2005-12 and revenue are highly correlated with countries’ 180 per capita GDP. Furthermore, different labour costs needed to build base stations or other network 160 EU infrastructure could significantly affect the mon- etary amounts invested in different countries, 140 US even for similar costs of equipment. Also, the use

120 of different accounting practices means that dif- ferent companies might compute the same finan-

100 cial items (eg revenues and capital investment) in substantially different ways, implying that 80 simple comparisons of these figures could be mis- 2005 2006 2007 2008 2009 2010 2011 2012 leading. Source: European Commission. Figure 6: Monthly wireless ARPU in the EU and In general terms, increased revenues can always US, €, 2005-12 be a result of lower competitive pressure. Compe- 60 tition stimulates investment by pushing compa- nies to invest and innovate as they seek other US 50 potential revenue sources (for an overview, see Motta, 2004; for an application to the telecoms

40 sector, see Nardotto et al, 2015). Competition EU might however reduce the incentive to invest if it

30 implies a reduction of expected profits after an investment is made12. The clearest example is with innovation: companies would not invest in 20

Q1 Q3 Q1 Q3 Q1 Q3 Q1 Q3 Q1 Q3 Q1 Q3 Q1 Q3 Q1 Q3 innovative projects without a patent system to 2006 2006 2008 2009 2010 2010 2011 2011 2012 2005 2005 2007 2007 2008 2009 2012 shield their inventions when their new products Source: European Commission. Figure 7: Public telecommunication investment per capita, telecommunications revenue per capita and GDP per capita in OECD countries in 2011

350 2500 Telecommunications($) per capita revenue 300 Public telecommunications investment per capita (left scale) GDP per capita (left scale) 2000 250 Telecommunications revenue per capita (right scale) 1500 200

150 1000

100 500 and GDP per capita ($ thousands) capita and GDP per 50

Public telecommunication investment per capita ($) investment telecommunication Public 0 0

UK 12. This is what is referred Italy Chile OECD Israel Spain Korea Japan Turkey France Poland Ireland Austria Greece Mexico Iceland Finland Canada Estonia Norway Sweden Belgium Portugal Slovakia Slovenia Hungary

Australia in the literature as non- Denmark Germany Czech Rep. Switzerland Netherlands Luxembourg New Zealand New

United States United monotonic (inverted-U) Source: Bruegel based on OECD. Note: public telecommunication investment per capita (left axis, in US$), telecommunication relationship between revenue per capita (right axis, in US$) and GDP per capita (left axis, in thousand US$). The correlation between revenue per competition and capita and investment per capita is 0.84, while GDP per capita has correlation of about 0.73 with both revenue per capita and investment/innovation investment per capita (with 1 indicating a perfect positive correlation). (Aghion et al, 2005). BRUEGEL POLICY CONTRIBUTION ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS 08

can be imitated so quickly and cheaply that they The coverage of Long Term Evolution (LTE) net- are not able to earn a sufficient return on their works – a standardised broadband wireless com- investment13. Hence, the contemporaneous cor- munication technology usually advertised as 4G – relation between revenue and investment con- is greater in the US than in the EU. In 2014, US cov- veys little information about the link between the erage reached 98.5 percent of the population, two variables, if not accompanied with information compared to 79.41 percent of households in the on the profitability of future new investment by EU (Figure 8). One explanation for such a differ- operators. ence is the delay by many European countries in assigning the radio spectrum necessary to pro- Mobile broadband connection (LTE/4G) vide 4G services over LTE technology. As Figure 9 coverage and speed shows, the US started to assign spectrum much earlier than EU countries. The first US auction took It should be noted that investment in itself might place in 2006 while the first auction in Europe took not be that relevant if that does not translate into place in 2008 in Sweden. For those EU countries infrastructure that directly benefits users. A meas- that were faster in auctioning off spectrum, a sim- ure of total expenditure on investment does not ilar, or greater, level of coverage to the US can be convey complete information on the benefits deliv- observed, while the only two countries with cov- ered. For example, territories with different struc- erage still below 50 percent, Cyprus and Bulgaria, tural features are likely to require different levels of have yet to assign spectrum in the 800 MHz band. investment. Less densely populated areas might Also, some US operators (eg Verizon) had a greater require higher investment. This does not mean that incentive to quickly adopt LTE because their net- users from those areas are better off than users works, unlike those of their European counter- from areas where investment levels are lower parts, were running on technologies that provided because the population is geographically more a significantly lower level of service. This also concentrated. For that reason, it is very important to forced the other operators in the US to quickly measure the outcome of an investment rather than respond and deploy LTE networks. the investment expenditure in itself. In terms of connection quality, download speed is The core benefit of new infrastructure is increased generally faster in Europe than in the US. Figure coverage, speed and reliability of communication. 10 shows that in several EU countries average In that context, the ability of users to access fast download speeds and peak download speeds are mobile telecommunications everywhere in Europe faster than speeds in the US, and that the per- now and in the future should be the main concern centage of consumers with a connection that is for EU policymakers, rather than the amount of faster than 4 megabits per second (Mbps) in investment expenditure. many European countries is higher than in the US.

Figure 8: LTE coverage in 2014 100% 99% 99% 100% 99% 96% 94% 92% 92% 92% 90%

90% 87% 84% 80% 79% 79% 77% 76%

80% 75% 73% 70% 68% 67% 67%

70% 65%

13. However, note that this 60% 58%

60% 56% is not the only effect a 52% 50% patent system might have

40% 36% on incentives to innovate. 30% Especially in cases in which 20% the boundaries of the 10%

patents are fuzzy, incum- 0% 0% US EU bents can use patents they UK Italy Malta Spain Latvia Greece France Cyprus Austria Poland Croatia Ireland Sweden Estonia Finland Portugal hold to block potential rivals Belgium Hungary Bulgaria Slovenia Slovakia Denmark Romania Germany Lithuania Czech Rep. Czech Netherlands or to extract rents from Luxembourg innovative firms, reducing Source: Bruegel. Note: LTE coverage for European countries is taken from the Digital Agenda Scoreboard and shows the percentage their incentives to invest in of households living in areas covered by LTE in 2014. The US data shows the percentage of the population living in areas covered innovative projects. by LTE in January 2014, and is from the FCC (2014), Table III.A.2, p31. BRUEGEL POLICY ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS CONTRIBUTION 09

3 IMPROVING EU MOBILE MARKETS up on LTE thanks to the progressive deployment of networks following spectrum assignments that The previous section showed that the area in were late relative to the US. This finding suggests which the EU has the most catching up to do is that the European Commission should implement mobile coverage of 4G connections. This is true a strategy that helps to increase the coverage and even taking into account that Europe is catching penetration of high-speed mobile broadband while Figure 9: Major recent US spectrum auctions and major EU 4G spectrum auctions

Spain: Ireland: United States: Netherlands: Multiband auction Multiband auction Auction-66 2.6 GHz auction Belgium: Sweden: Advanced Wireless 800 MHz auction Services (AWS-1) Austria: 1800 MHz Sweden: 2.6 GHz auction auction UK: 2.6 GHz auction Denmark: France: 4G auction 900/1800 MHz auction 800 MHz auction

2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 France: Sweden: 2.6 GHz auction 800 MHz Italy: Multiband auction De nmark: Finland: auction 800 MHz 2.6 GHz auction Austria: United States: Auction-97 Belgium: auction Multiband 2.6 GHz auction Advanced Wireless Denmark: auction Services (AWS-3) United States: Portugal: Netherlands: Finland: Auction-73 2.5 GHz auction Multiband auction Germany: Multiband 800 MHz 700 MHz band Spain: auction auction Multiband auction Re-auction Source: Bruegel. Figure 10: Average speed, Average peak speed and percentage of connections above 4 Mbps, 2014

18 16 Average mobile connection speed in megabits per second (Mbps) 14 16.0 12

10 9.0 8.3 7.9 7.3 6.8

8 6.7 5.9 5.9 5.3 5.0 4.9 4.8 4.6 6 4.8 4.5 4.5 4.2 4.2 3.9 3.5 4 3.2 2.4 2 0 US UK Italy Spain Japan France Poland Croatia Ireland Austria N’lands Sweden Belgium Slovakia Slovenia Hungary Australia Romania Denmark Germany Lithuania Czech Rep. Czech South Korea 120 100 Average peak connection speed in megabits per second (Mbps) 105.3 80 74.7

60 46.1 45.9 40.3 40.1 37.6 37.4 36.6

40 33.9 31.7 28.1 28.6 28.5 24.1 22.2 21.1 18.6 18.4 16.6 16.1 20 15.6 10.1 0 US UK Italy Spain Japan France Poland Croatia Ireland Austria N’lands Sweden Belgium Slovakia Slovenia Hungary Australia Romania Denmark Germany Lithuania Czech Rep. South Korea 100% 90% 78.3%

77.8% % of mobile connections above 4 megabits per second (Mbps)

80% 74.8% 71.3% 69.5% 66.3%

70% 91.8% 90.0% 60% 59.3% 52.3% 49.3% 49.3% 45.5% 44.0% 50% 44.5% 42.5% 42.3%

40% 33.0% 30% 28.3% 20.0% 20% 19.8% 12.5% 10% 0.9% 0% US UK Italy Spain Japan France Poland Croatia Ireland Austria N’lands Sweden Belgium Slovakia Slovenia Hungary Australia Romania Denmark Germany Lithuania Czech Rep. Czech South Korea Source: Akamai. Note: Yearly averages were computed using quarterly data published by Akamai in its State of the Internet quarterly reports (see https://www.stateoftheinternet.com). When data in some quarters for some countries was missing (data for Romania was available only for Q1 and Q2; data for Croatia and South Korea was not available in Q4), the average for those countries was computed using the data in the available quarters. Similar rankings, with several European countries outperforming the US, can be obtained using other speed measurements (eg Ookla/NetIndex and OpenSignal). BRUEGEL POLICY CONTRIBUTION ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS 10

guaranteeing access at affordable prices, particu- risk that significant domestic rents will be lost: larly in view of the adoption in the long-term of expanding supply across borders means impos- new technologies such as 5G, on which the delays ing a threat to the rents of other operators in other experienced with 4G should not be repeated. national markets, potentially leading to retaliation. The bigger the domestic profits, the more an oper- A particular emphasis should be placed on data ator has to lose from a cross-border service supply access, rather than voice and text. The recent ‘war’ between EU-wide operators. trend in mobile telecoms has been the increasing importance of data traffic: while voice traffic was Following from this, we consider two policies that basically flat from Q1 2008 to Q4 2014, data traf- could have a positive effect on cross-border com- fic increased 54 percent year on year and now petition by increasing the relative profitability of a greatly exceeds voice traffic (Akamai, 2015) service provided on an EU-wide basis, compared (even though a large part of this traffic is offloaded to a service provided to domestic markets only: to Wi-Fi connections at home or at work; see Euro- (1) international roaming and (2) radio spectrum pean Parliament, 2013, pp92-93). Data is management. expected to represent an even a larger share of traffic in the future, especially as voice and text 3.1 International roaming messages are themselves going to be data pack- ets running over shared links. This is already the To a great extent the debate about the conver- case with services like WhatsApp, an alternative gence on a single European tariff has overlapped to SMS, and Voice over IP (VoIP) services such as with the debate around international roaming Skype, which provide an imperfect alternative to charges. In July 2015, the European Parliament traditional telephone call services. and Council agreed in principle on a draft regula- tion that would eliminate roaming charges within A significant increase in cross-border competition the EU14. The new rules would reduce roaming sur- would help Europe to pursue the objectives of charges on national tariffs to €0.05 per minute for greater coverage and affordable access: the voice calls, €0.02 per SMS, and €0.05 per MB of expansion of cross-border networks and services data downloaded from April 2016. From June would reduce investment costs and increase prof- 2017, surcharges would be eliminated. ‘Fair use’ itability while preserving the incentive to supply limits would be implemented to prevent exploita- fast and reliable mobile services at affordable tion of arbitrage possibilities through permanent prices, for the reasons discussed in the introduc- roaming (ie customers using SIM cards from low- tory section. price countries for domestic use).

To stimulate cross-border competition the Com- Under the regulation, prices would still be differ- mission, the European Parliament and the Coun- ent in different countries, but the same price cil of the EU should use their regulatory powers to would have to be charged whether customers con- make it relatively more attractive to operate cross- nect to their provider’s home network or a network border networks instead of focusing on domestic in another country15. By comparison, within the markets. Aghion et al (2005) explains the theory framework currently in force, international roam- that we apply to this context: companies are ing services tend to be expensive, especially com- attracted by the prospect of higher profits. Lower pared to similar domestic services, and the price 14. http://www.consilium. costs of entry into cross-border markets and difference is not primarily due to differences in the europa.eu/en/press/press- releases/2015/07/08-roam increased competitive pressure in domestic mar- underlying costs. There are three main drivers that ing-charges/. kets should create an incentive to ‘escape’ domes- make international roaming more expensive than 15. At the time of writing, tic competition and seek profits across borders. domestic mobile access. First, customers are on the provisions have still to Conversely, increased domestic profits because average different (occasional roamers might have be formally adopted. The of a reduction of competitive pressure might higher purchasing power than the average domes- current text leaves open render new investment relatively less attractive. tic user and their need to use mobile communica- questions about how the This is particularly true if new investment to tion services might differ when travelling, for regulation will effectively be implemented. expand network reach across borders involves a example). Second, roaming requires access to BRUEGEL POLICY ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS CONTRIBUTION 11 multiple networks owned by different operators to those for customers living in the country they resulting in ‘double mark-up’ effects (ie the are visiting – sometimes referred to as ‘roam like increase in price due to lack of coordination a local’ (RLAL, see Marcus et al, 2013). This would between suppliers of two complementary goods, avoid arbitrage effects without the need for a ‘fair in this case access to separate networks) and, in use’ limitation clause, while potentially signifi- the case of international roaming, inefficient bar- cantly reducing roaming prices. RLAL conditions gaining processes16. Third and most importantly, can be achieved through the design of a regula- international roaming services are expensive tory framework so that competition in the roaming because they are normally sold in bundles with market is stimulated (while not preventing com- domestic services. Customers, who predomi- panies from charging different prices in different nantly use domestic services, tend to choose an member states, if the economic conditions so operator on the basis of its domestic offer – hence require). operators have little incentive to compete and reduce their tariffs in the international roaming The EU Roaming III regulation19 introduced a number market. Furthermore, at least at current prices, of structural measures, in force since July 2014, price reductions do not seem on average to stim- with this aim: MNOs were required to unbundle 16. Since roaming arrange- ments are usually recipro- ulate demand and generate higher usage of roam- roaming services from their domestic offerings and cal and operators try to ing services for voice and text (Marcus et al, give the option to “alternative roaming providers” to have a balanced flow of traf- 2013). offer these services to their customers. fic between themselves, operators do not choose their roaming partners only For this reason, the European Commission has The implementation of these ‘decoupling’ meas- on the basis of the whole- previously introduced price caps at the wholesale ures has however been unsuccessful. Very few sale price offered, since in and retail level and measures aimed at increasing alternative operators have so far entered the this framework roaming is transparency and avoiding bill shocks. The stated market and it is unlikely that others will enter in not only a cost but also a ultimate goal of the Commission is a ‘roam like at the near future. This might be because of the lack source of revenue, and larger operators might be home’ (RLAH) scenario, in which prices of mobile of commitment on the part of the Commission to preferred as partners even services do not change “just because [con- enforce the measures. Even before the entry into though another small one sumers] have crossed an invisible internal border force of the decoupling measures, the Commis- was offering a better whole- that is supposed to have disappeared” (Kroes, sion proposed other regulatory measures in the sale price (Shortall, 2010). 2011). This in practice means equalising roaming context of the Telecom Single Market package20, 17. In order to avoid this and domestic charges in order to allow consumers overlapping with Roaming III. In particular, MNOs problem, the draft regula- tion allows operators to to replicate their typical domestic consumption were offered the possibility to escape the apply for an authorisation to patterns while travelling in other EU countries. unbundling requirement by entering into pan- add a surcharge to the European alliances with other network operators extent necessary to recover RLAH charges could leave operators facing poten- and making RLAH offers to their own customers. costs. tially drastic business challenges. For example, an Even though it was unlikely that any MNO would 18. This can happen when operator from a low-income country might have a have entered such alliances, the fact that they operators are prevented from price discriminating domestic retail price that is below the wholesale could use them to escape the structural decou- when supplying a service to price the operator would need to pay to get access pling measure was enough to destabilise alterna- two group of customers to a host network in a high-income country, leav- tive roaming operators’ business plans by making with significant differences ing the operators with negative margins on roam- their future profitability uncertain. in the structure of their ing services. The operator could then find it demand. For a discussion on the welfare effects of unsustainable to offer roaming services in the The new regulation's elimination of roaming sur- price discrimination, see EU17. Moreover, establishing a direct link between charges suggests that the EU institutions have Papandropoulos (2007). the price charged for domestic use and the price little intention of credibly pursuing the decoupling 19. http://eur- charged to international roamers could introduce solution. However, the RLAH provisions do not lex.europa.eu/LexUriServ/L distortions in domestic markets, which could lead address the structural problems that are behind exUriServ.do?uri=OJ:L:201 to price increases18. high roaming prices and might have unintended 2:172:0010:0035:EN:PDF. consequences for domestic prices and MNO com- 20. http://ec.europa.eu/ A better solution would be for users of interna- petitiveness in home and other EU member state information_society/news- room/cf/dae/document.cfm tional roaming services to face conditions similar markets (BEREC, 2014). The Commission should ?doc_id=2734. BRUEGEL POLICY CONTRIBUTION ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS 12

therefore commit to seriously enforce decoupling ders the creation of operators with a larger Euro- measures, especially in case the RLAH caps result pean footprint for several reasons. Auctions in dif- in serious market distortions. ferent countries are run at different times. When bidding in early auctions, bidders willing to oper- If successfully implemented, structural measures ate in multiple countries face aggregation risks aimed at stimulating competition in the roaming (the so-called ‘exposure problem’)21. Bidders that market would imply a reduction of operators’ rev- want to operate in multiple countries are likely to enues from hosting roamers on domestic net- calculate their bids for individual lots (ie the rights works. International roaming revenue accounts for to use a certain range of frequencies in a given 5-12 percent of EU MNO revenues – with margins geographic area to provide wireless communica- often higher than 60 percent (Wall Street Journal, tion services) on the basis of the value that the 2014). The effective implementation of structural whole bundle of licenses they want to obtain will measures would therefore not only reduce costs have if ultimately acquired. The bundle value is for travellers. It would also expose domestic net- likely to be higher than the sum of the single works to increased competition, reducing their licenses. For instance, an operator might find it domestic revenues. Moreover, opening roaming more profitable because of economies of scale to markets to competition would provide MNOs with hold licenses for both France and Spain compared a concrete incentive to become more competitive to the average profits that two operators would internationally, since the measures would create make if each held one of the licenses. Bidders the opportunity for profitable entry into other seeking licenses in multiple countries face the risk countries’ markets. Effective implementation of of paying too much in early auctions, if they fail to structural measures to open up international secure other licenses in later auctions whose syn- roaming markets could stimulate operators to ergies would have justified the higher price. expand their service across borders. Fragmentation in assignment procedures also 3.2 Radio spectrum management reduces the ability of bidders to switch to substi- tute lots in other countries if lots in one country Radio spectrum (hereafter spectrum) is a scarce become too expensive. Instead each assignment resource that is essential for MNOs to provide wire- procedure is a separate exercise with its own par- less communication services. The public man- ticipation costs, resulting in less flexibility to agement of spectrum can be divided into two switch between lots compared to single EU-wide phases: allocation and assignment. Allocation auctions. Furthermore, an operator would only be refers to decisions over the uses of given bands able to substitute expensive lots in one country of spectrum (eg for wireless communications, tel- with those in countries where auctions have yet 21. This risk is also present evision or radio broadcasting). Allocation in the EU to take place and might regret choices made in for bidders who are willing is done by member states within a framework of early auctions if guesses on the prices in later auc- to buy lots only in one country if they are allowed international coordination and harmonisation, tions turn out to be wrong. to buy multiple lots and designed to counter cross-border interference. package bidding (ie the Harmonised supra-national allocation of spectrum Separate assignment procedures therefore possibility to submit single also brings other benefits, such as enabling wire- require bidders that desire to obtain licenses in bids for packages of lots) is less device manufacturers to produce the same more than one country to work on the basis of not allowed, but is in general mitigated by the device for use in many countries with associated guesswork about the outcomes of future auctions, possibility to switch to economies of scale, and enabling users to use the which tends to make bidding strategies in substitute lots if one of the same device in other countries. Assignment refers sequences of auctions more complex and might lots they were bidding for to the award of rights to use a portion of a specific push some bidders to bid more conservatively. became too expensive and by the possibility to band of spectrum. In the EU, the right to use spec- withdraw (usually with a trum for commercial purposes is currently To reduce costs for operators and incentivise the penalty or some other assigned on a national basis by member states, deployment of networks with a larger European mechanism that ensures most commonly through auctions. footprint, there should be a move towards EU-level bids represent assignment of spectrum, with a suitable transition commitments from participants) their bid. Fragmentation in the assignment of spectrum hin- period to take into account the variable features BRUEGEL POLICY ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS CONTRIBUTION 13 of each country’s market. For example, countries a bundle of lots) are allowed. An option would be to might differ in terms of their uses of spectrum allocate revenues on the basis of the expected bands, demand for spectrum22 or license periods. profitability of licenses25. In practice, member 22. For instance, in some With that constraint in mind, the harmonisation states would need to delegate the actual sale of countries there was no process should proceed as quickly as possible, so the licenses to the Commission, working along- demand for spectrum in the that the Commission and member states are side national regulators, but would still retain con- 3.4-3.8 GHz band. See ready for the allocation of frequencies to be used trol over the use of spectrum, such as the http://ec.europa.eu/transpa by future electronic communications technolo- conditions attached to the license. rency/regdoc/rep/1/2014/E N/1-2014-536-EN-F1-1.pdf. gies, such as 5G23. 23. While it is not yet clear Centralised pan-European auctions have so far what precisely 5G mobile Ideally, the EU should implement a system similar been resisted by member states concerned about technologies would be, they to that in the US, where the Federal Communica- sovereignty issues linked to the loss of control are presumed to be tech- tions Commission (FCC) assigns licenses in dif- over spectrum and potential revenue losses. In nologies offering larger ferent geographical areas through a single the system we have outlined, member states throughput and better spec- tral efficiency and scalabil- 24 auction . This would reduce aggregation risks for would retain control over license conditions, while ity. It is likely that in order to bidders willing to purchase spectrum in multiple the Commission would just act as a delegate for do this, different technolo- EU countries by reducing the amount of guess- the sale of spectrum, meaning that loss of control gies and spectrum at very work needed for their bidding strategies. Since over market structure downstream and sover- high frequencies (microwaves) should be operators would have to participate only in one eignty considerations relating to the use of spec- used. These technologies auction offering the possibility to bid for licenses trum should not represent an issue. The main risk are supposed, according to in the EU or possibly the European Economic Area, would be the possibility of stalemates because of those participating in their participation costs would likely be significantly lack of agreement between member states on development, to hit markets reduced. Expenditure in terms of public resources timing, the details of auction design, harmonisa- between 2020 and 2025. See 5GPP (2014) and GSMA would also likely be lower than the cumulative tion of license conditions or the mechanism for Intelligence (2014). cost of separate auctions in each member state. splitting package bids. 24. The FCC determines the lots to be sold in these From a practical perspective, the Commission As for revenues, it is important to stress that an auctions in what is called should involve national regulatory authorities in efficient and competitive telecoms market rather the band plan, in which the the design of the auction, in order to take into than revenues should be the main objective of portion of spectrum to be sold through the auction is account the characteristics of individual national spectrum auctions. An allotment of spectrum divided in specific markets and to exploit national regulators' experi- rights that enables an efficient and competitive frequency ranges (‘blocks’) ence with auctions. Lots should be still defined wireless telecoms market is likely to generate and the US territory is nationally to avoid any sovereignty concern and benefits for consumers that are much greater than divided according to some geographic partition for because markets will remain regulated on a revenues accruing to public finances. Hazlett et al each block. national basis for the foreseeable future. This (2012) found that a conservative estimate of the 25. For instance, methods reform should be coupled with a more harmonised annual consumer surplus ($174 billion) based on a proportion of the regulatory framework across EU countries that generated by mobile services in the US in 2009 bidding units assigned to would allow MNOs expanding across Europe to substantially exceeds all auction revenues each lot or the information operate the networks in the different countries as collected by the FCC from 1994 to 2009 ($53 revealed by the bids sub- 26 mitted during the auction (if a single one. The auction format should be decided billion) . Considering that design choices aimed detailed enough) could be on a case-by-case basis to adapt it to the details of at increasing revenues are not costless and might used. However, further of the economic scenario and to benefit from have negative impacts on welfare in the study on mechanisms for future innovations in auction design. downstream market (eg reducing the number of splitting revenues is surely licenses to assign), it seems natural that revenues needed if Europe decides to move in the direction of a The revenues obtained from the lots in each (as a non-distortionary/lump-sum form of public centralised auction with member state should either be transferred to the funding) should be an objective only insofar as package bidding. respective member state or be subtracted from that objective is in line with efficient and 26. Furthermore, also the the sums they have to transfer to the Commission. competitive telecoms markets. estimated annual profits of mobile operators in 2009 One complex issue would be the split of revenues In any case, generally speaking, if efficiently ($151.7 billion) were much higher than the cumulative between countries when package bids (ie bids for designed, a centralised auction should not leave revenues. BRUEGEL POLICY CONTRIBUTION ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS 14

member states worse off in terms of auction rev- Of course, any potential benefit of coordinated enue. This would mostly depend on whether the spectrum assignment would be undermined if 27. In terms of individual total revenue from a centralised auction is at least there is no mechanism to preserve or increase revenues, while it is possi- ble that some of the rev- equivalent to the cumulative revenues yielded by competition in the mobile market, for instance by enues might decrease, national-level auctions. preventing incumbents from hoarding spectrum arbitrage between lots in in order to undermine rivals. Spectrum is both an different countries would It is impossible to know whether this would be the instrument in the hands of the Commission to tend to generate more com- case: the number of variables that affect auction reduce costs for operators seeking to expand their petitive prices (more reflec- tive of the true value of revenues and the complexity of players' bidding European footprints and a tool to introduce or spectrum). Centralised auc- behaviour make any estimate very speculative maintain, when needed, competition in national tions would represent a sort (see Milgrom, 2004, or Salant, 2014, on spectrum markets. Spectrum aggregation limits, such as set- of insurance for countries auction design). However, there are a number of asides and spectrum caps, have often been used that for some reason could otherwise be late in assign- reasons to think that a centralised EU auction to maintain or introduce competition in the ing their spectrum, for would not reduce aggregate revenue, if properly market, avoiding concentration of spectrum in the instance because of some designed27. hands of few operators. When implemented cor- problem freeing frequen- rectly, these measures helped to foster competi- cies from previous users A hypothetical EU auctioneer would have an inter- tive mobile markets29. (see Klemperer 2004, Chapter 5, pp164-166 for est in maximising the participation of operators in reasons why countries auc- the auctions for the licenses in each country in Furthermore, if the Commission believes that fast tioning later might see their order to maximise aggregate revenue and achieve deployment of high-speed mobile networks is fun- revenues reduced). a more efficient assignment. National auctioneers, damental, it could try to induce member states to 28. Among the reasons why however, would look only at their national rev- include more stringent roll-out conditions in the auctions are preferred to enues and assignments without considering, for spectrum licenses assigned through the cen- comparative hearings (also called ‘beauty contests’), in example, that imposing a higher reserve price that tralised auction. This however would translate into addition to being more discourages some bidders from purchasing spec- reduced revenues, since MNOs would internalise transparent methods, is trum in their country could also reduce the will- these costs in their bids. More importantly, less- that the regulator managing ingness of these same bidders to participate in stringent conditions would enable a more cost- the spectrum might not have the information neces- auctions in other countries with potentially com- efficient roll-out. sary to identify who would plementary licenses. This suggests that an EU be the best users for the auctioneer might be able to achieve higher total 4 CONCLUSIONS spectrum on sale. The auc- revenues and a more efficient assignment. tion allows the spectrum Further integration towards a ‘single mobile tele- management agency to let the operators reveal Furthermore, the reduced aggregation risk and the coms market’ in the EU is certainly desirable. How- through their bids which reduced amount of guesswork for bidders would ever, integration is not an end in itself; rather, it is among them is best suited reduce bidding uncertainty. That reduced cost (or important to clarify the goals that integration is to use a certain block of expected cost) and uncertainty could be reflected meant to achieve. For example, unless income spectrum. However, since the lots sold in these auc- in more confident bids, increasing the likelihood levels converge in the long term, the emergence tions are rights to use an of obtaining larger revenues. of mobile tariff plans so that users are charged the essential resource to oper- same price everywhere in Europe is neither obvi- ate in a market, spectrum Ultimately, however, if revenues were the main ous nor necessarily desirable. aggregation limits should objective of these auctions – and they should not be used in order to be sure that bids do not contain the be, as discussed above – these are still going to Action from the European Commission to improve value for an incumbent of a be minor details. The elements that are likely to the functioning of European mobile markets is less competitive or more have the greatest influence on auction revenues welcome. The priorities should be increasing cov- concentrated market, or are design features meant to attract bidders and erage of high-speed mobile connections, and any other incumbency advantage. If properly to discourage collusive and predatory behaviour measures to support the creation of ‘pan-Euro- implemented, spectrum or other strategic manipulation, ensuring effective pean’ networks – networks operated by compa- aggregation limits would and robust competition in the auction and an nies with a wider European footprint. Wider also ensure that revenue effective 'revelation mechanism' through which European networks tend to be more efficient: they and efficiency of the mobile information on bidders’ characteristics is dis- promote cross-border competition and reduce markets are not conflicting 28 with each other. closed . deployment and operational costs, ultimately BRUEGEL POLICY ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS CONTRIBUTION 15

stimulating investment while keeping access Commission could intervene. Policies in other 29. In particular, set-asides prices low. It is particularly important to ease areas might have equally relevant effects, though are very effective tools access to mobile data traffic. if inappropriately designed, those policies might when credible potential entrants are present. The undermine the effectiveness of the measures dis- UK UMTS auction provides The Commission should start in areas where it is cussed in this paper30. Potential areas for inter- one of the most successful easier to appreciate the benefit of intervention: vention include: cross-country regulatory examples of competitive international roaming and spectrum management. harmonisation, not necessarily limited to the mobile markets fostered by By intervening efficiently with clear policy meas- design and the enforcement of telecoms regula- set-asides: H3G, the com- pany which entered the UK ures in these areas, the Commission could obtain tion, but also areas such as consumer protection mobile market thanks to the two outcomes: (a) stimulation of competition and other policies to reduce the costs of cross- license set-aside for a new within national mobile markets; (b) reduction in border services delivered through mobile net- entrant, turned out to be a the cost of cross-country expansion of supply. A works; increased coherence of the taxation/VAT competitive and innovative force in the UK mobile compression of profits in domestic markets would framework; and the introduction of measures to market. See COMP/M.5650 increase the incentive to look for profits in inter- support demand, such as measures to increase T-MOBILE/ORANGE, paras. national markets, possibly through securing the security of mobile online transactions. These 49 and 107-108. See Cram- bigger scale and a potentially more efficient pro- areas are all listed in the Commission's DSM strat- ton (2013a, b) and Cave duction cost structure. A compression of costs for egy and measures are expected in the future. and Webb (2013) for other successful implementa- international operations would make such a strat- tions of spectrum aggrega- egy even more profitable. The Commission should also resist any pressure tion limits and how these to relax merger control and facilitate domestic may also stimulate compe- On roaming, the Commission should not abandon consolidation, as this would be likely to have neg- tition in the auction and increase revenues. the idea of seriously pushing structural measures ative short and long-term effects on consumers to increase competition in the international roam- and on the development of a DSM. If mergers that 30. The provisional rules on net neutrality recently ing market and to move towards a ‘roam like a reduce domestic competition in one or more agreed by the Council and local’ scenario. For spectrum, centralised auctions member states pass through the merger regula- Parliament could particu- implemented in a way resembling US auctions tion net without proper remedies, the benefits that larly affect the incentives of would make it easier for operators to expand their society enjoys from mobile communications and operators to provide fast and unconstrained data footprints, and the use of spectrum aggregation the speed of the development of the industry access. These rules provide limits would ensure that such a shift does not might be reduced, and operators would find in that in general providers of result in less competitive mobile markets. domestic markets a profitable alternative to cross- internet services are not border expansion. In that sense, the Commission allowed to discriminate data An important caveat is that international roaming should take care of properly enforcing merger on their networks, but give them the possibility to and spectrum management are only two out of a rules so that it does not undermine the effects of manage traffic in times of number of policy areas in which the European its pro-DSM policies. exceptional network con- gestion, to offer a differenti- ated treatment for REFERENCES ‘specialised services’ need- ing a quality of service that 5PPP (2014) 5G Vision – The 5G Infrastructure Public Private Partnership: the Next Generation of cannot be assured by stan- Communication Networks and Services dard internet access and to Aghion, P., Bloom, N., Blundell, R., Griffith, R. and Howitt, P. (2005) ‘Competition and Innovation: an provide sponsored services (zero rating). These excep- Inverted-U Relationship’, The Quarterly Journal of Economics, 120(2): 701-728 tions, if not carefully Akamai (2015) Akamai’s State of the Internet: Q4 2014 Report, available at https://www.state- designed, could give the oftheinternet.com/resources-connectivity-2014-q4-state-of-the-internet-report.html incentive to operators to BEREC (2014) International Roaming – Analysis of the Impacts of ‘Roam Like At Home’ (RLAH) offer ‘standard’ internet access offer with low Cave, M. and Webb, W. (2013) ‘Spectrum Limits and Auction Revenue: the European Experience’, speeds and/or low endow- available at http://apps.fcc.gov/ecfs/document/view?id=7520934210 ments in order to push Cisco (2015) Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2014– more application and con- 2019, available at http://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-net- tent providers to move into working-index-vni/white_paper_c11-520862.pdf specialized or sponsored services (with several Cramton, P. (2013a) ‘The Rationale for Spectrum Limits and Their Impact on Auction Outcomes’, potential negative effects available at http://www.cramton.umd.edu/papers2010-2014/cramton-spectrum-limits-ex-parte.pdf on the market). BRUEGEL POLICY CONTRIBUTION ADDRESSING FRAGMENTATION IN EU MOBILE TELECOMS MARKETS 16

Cramton, P. (2013b) ‘The Revenue Impact of Competition Policy in the FCC Incentive Auction’, avail- able at http://www.cramton.umd.edu/papers2010-2014/cramton-revenue-impact-of-competition- policy-in-incentive-auction.pdf European Commission (2015) A Digital Single Market Strategy for Europe, COM (2015) 192 European Parliament (2013) Entertainment x.0 to Boost Broadband Deployment, available at http://www.europarl.europa.eu/document/activities/cont/201310/20131017ATT72946/20131017A TT72946EN.pdf FCC (2011) Mobile Wireless Competition Report (15th Annual), Federal Communications Commission FCC (2014) Mobile Wireless Competition Report (17th Annual), Federal Communications Commission Grzybowskiy, L. and Verboven, F. (2014) ‘Substitution between Fixed-line and Mobile Access: the Role of Complementarities’, CEPR Discussion Paper DP10045 GSMA Intelligence (2014) Understanding 5G: Perspective on Future Technological Advancements in Mobile Hazlett, T. H., Muñoz, R. E. and Avanzini D. B. (2012) ‘What Really Matters in Spectrum Allocation Design’, Northwestern Journal of Technology and Intellectual Property, 10(3): 93-123 Klemperer, P. (2004) Auctions: Theory and Practice, Princeton University Press Kroes, Neelie (2011) ‘A better deal for roaming’, SPEECH/11/502, available at http://europa.eu/rapid/press-release_SPEECH-11-502_en.pdf Marcus, J. S., Philbeck, I. and Gries, C.-I. (2013) ‘Structural Solutions and the Evolution of Interna- tional Mobile Roaming (IMR) in Europe: Where are We Headed?’, SSRN Working Paper, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2342637 Milgrom, P. (2004) Putting Auction Theory to Work, Cambridge University Press Motta, M. (2004) Competition Policy: Theory and Practice, Cambridge University Press Nardotto, M., Valletti, T. and Verboven, F. (2015) ‘Unbundling the Incumbent: Evidence from UK Broadband’, Journal of the European Economic Association, 13(2): 330–362 OECD (2013), OECD Communications Outlook 2013 Ofcom (2014) International Communications Market Report, available at http://stakeholders.ofcom.org.uk/binaries/research/cmr/cmr14/icmr/ICMR_2014.pdf Papandropoulos, P. (2007) ‘How should price discrimination be dealt with by competition authori- ties?’ Revue des droits de la concurrence (3): 34-38 Salant, D. J. (2014) A Primer on Auction Design, Management, and Strategy, MIT Press Shortall, T. (2010) ‘A structural solution to roaming in Europe’, EUI Working Papers RSCAS 2010/62, Robert Schuman Centre for Advanced Studies, Florence School of Regulation Wall Street Journal (2014) ‘European Telecom Firms Retreat on Roaming Fees’, 21 February