Brian WilliamsonAn internet traffic taxwould harm Europe’sdigital transformationJuly 2022

About the AuthorsBrian Williamson works at the intersection of the ‘digital economy’ and policy. Brian has worked on thecontribution of connectivity and applications to economic and social outcomes and the policies likely tosupport the greatest possible contribution. Clients have included governments, regulators, telcos, andtech companies.DisclaimerThis is an independent report funded by the Computer & Communications Industry Association. Theopinions offered herein are purely those of the authors. They do not necessarily represent the views ofthe client, nor do they represent a corporate opinion of Communications Chambers.

Contents1.Executive summary . 12.Context . 43.Data growth is good for telcos . 5The Axon paper relies on a flawed estimate of traffic related costs5Costs attributable to data growth are low and declining5The revenue uplift from data growth is positive for telcos7What telcos say to investors differs from what some say to policy makers7The real problem – not enough data growth?8Conclusion94. Giving telcos money and taxing content and applications would not foster investment – itwould harm it . 105.6.7.Giving telcos money would not foster investment10Taxing internet traffic would reduce investment and innovation in applications and networks10Conclusion11The assertion of asymmetric bargaining power is a red herring . 12Prices for IP transit are falling alongside cost declines12BEREC (2017) found that the IP interconnection market generally works well12WIK (2022) found that the IP interconnection market generally works well13Experience in South Korean suggests that Europe should not move to sending network pays13Conclusion14Zooming out to understand the broader picture. 15Incentives15GDP and jobs16Productivity and prices16The tax base for broadband funding17Value chains17Europe’s Digital Decade18Net neutrality18The Green transition18Fairness19Reductio ad absurdum20Conclusion21The devil in the detail . 22Who might pay?22What should be the level of internet traffic tax?22Who would receive the funds?22What conditionality should apply?22Would an internet traffic tax conflict with other policies?23What should an impact assessment cover?23Conclusion24

1. Executive summaryA report by Axon (2022), commissioned by theEuropeanTelecommunicationsNetworkOperators’ Association (ETNO), proposes thatcertain content and application providers berequiredtonegotiateafeewithtelecommunications network operators. Axonclaims this would promote network investmentand be beneficial overall.This paper argues that there is no sound basis forimposing a fee that would harm rather thanpromote investment by reducing innovation anduse in relation to content and applications; andwould harm achievement of the EuropeanCommission’s digital transformation vision for2030.Data growth is good for telcosThe cost estimates cited in the Axon report areflawed as a basis for assessing traffic relatedcosts since they are not based on an assessmentof incremental traffic costs.The incremental costs of internet traffic arenegligible for fixed broadband access, low anddeclining for mobile access and low in transitmarkets where content and applicationproviders invest in network capacity e.g., insubsea fibre optic cables. The predominant IPmodel is settlement free peering.Incremental traffic related costs therefore rangefrom approximately zero for fixed access to lowand declining for transit, core and mobilenetworks.Not only are the unit costs of traffic decliningover time – otherwise past traffic growth couldnot have been accommodated – but trafficgrowth drives telco revenues.Traffic growth promotes fixed access retentionand fibre adoption, and traffic growth ismonetised via tiered mobile data tariffs wherebyconsumers pay more for more. Data growth isgood for telcos, not bad, which is what they sayto investors:“Surging demand for mobile data is theclear driver for future growth in thebusiness” Telefonica, 2020There is no basis for requiring content andapplication providers to pay an additional feerelated to traffic.What is proposed is in fact an internet traffic taxunrelated to the opportunity cost of trafficwhich may be negative and is certainlynegligible.Giving telcos money and taxing content andapplications would harm rather than fosterinvestmentIf you simply give money to telcos the demandand price for access is unchanged, so thebusiness case for investment would beunchanged. The money would go toshareholders, and even if strings were attachedadditionality would be hard to verify.If you tax content and applications you willreduce the development, adoption and use ofcontent and applications, on which the businesscase for network investment depends. Theincidence of an internet traffic tax would alsoextend far beyond its initial point of application,for example, impacting those using cloudservices. The net impact is therefore likely to benegative for investment.Further, an internet traffic tax would be inimicalto the European Commission’s goal of achievingdigital transformation, including the goals of75% of EU companies using traffic intensiveapplications such as cloud, AI and big data andthe growth of scale-ups across sectors fromfintech, to healthcare and the gaming andcreative sectors.Finally, data growth is not exponential but hasbeen declining (except for a spike in growth[1]

during ‘lock down’). Perhaps the question thatshould be asked is whether the real problem isinsufficient data growth, both as a reflection ofdigital transformation and as a spur to networkinvestment.The assertion of asymmetric bargaining poweris a red herringAxon asserts that the root cause of the problemtheir proposals seek to address is asymmetricbargaining power between major OTTs andtelecommunications network operators inrelation to IP traffic. Yet reviews by BEREC in2017 and WIK in 2022 for Bundesnetzagentur,do not support the Axon conjecture and find thatthe market is competitive and functioning well.A further review by BEREC into the sending partypays principle and ‘fair contribution’ is pending,but on the basis of existing reviews the Axonassertion of asymmetric bargaining power asgrounds for an internet traffic tax is a redherring.Evidence from South Korea suggests aninternet traffic tax would be harmfulWe have a natural experiment of what Axonproposes in South Korea, and the outcome is notpositive. WIK (2022) report that:“Market observers report a decline indiversity of online content and expect risingprices for end users for content, as well aslower network infrastructure investments.Quality for end users is declining.”Zooming out to understand the broaderpictureIt is essential to ‘zoom out’, as the EuropeanCommission do in their vision for 2030, andconsider how to foster digital transformationand the benefits that will flow from it.Yet Axon focuses on a hypothetical GDP gainfrom additional network investment but doesnot consider the GDP loss from a tax on internettraffic which would discourage the use anddevelopment of content and applications. Notonly would the negative impact on content andapplications negatively impact GDP directly, itwould also do so indirectly by reducing demandfor network investment.A further illustration is that Axon focusses on theenergy used by networks, but not the potentialfor applications to reduce energy usethroughout the economy, for example, viaonline collaboration. As noted in the EuropeanCommission Strategy for Data:“Data driven applications will benefitcitizens and businesses in many ways. Theycan improve sustainability and energyefficiency”The devil in the detailWhilst we conclude that an internet traffic tax isnot justified and would prove harmful, werecognise that the European Commissionexamination may go into detail regarding howsuch a mechanism might be implemented andwhat ramifications it might have across adjacentpolicy areas.Three issues are highlighted, namely ensuring acompetitive level playing field betweenincumbent and entrant investors in theallocation of any traffic tax revenues, andcompatibility of the proposed policy with globaltax reform and net neutrality principles.First, given the rise in network entrants able totap into infrastructure funds, ensuring neutralityof funding between entrants and incumbents isnot only an issue of fairness but also a materialconsideration in relation to efficiency.Telecoms regulation has focused on ensuringequivalence of access to incumbent networks forretailers, the issue at stake here is different.Namely, ensuring that vertically integratedincumbent retailers do not self-preference intheir choice of network inputs, therebyweakening an entrant’s investment case. Oneoption, adopted in New Zealand in relation tostate support for fibre investment, would be to[2]

make structural separation a condition forreceipt of funds.Second, given that the proposal is focused on‘big tech’, is not based on evidence of marketfailure or inefficiency, and aims to move awayfrom voluntary contracting, it will be viewed as atax. As such it may run counter to agreement notto introduce new digital taxes whilst proposalsfor reform of taxation of multinationalenterprises, including digital corporations, areunder development globally.Third, the proposed approach is not onlydiscriminatory between sources of traffic basedon the scale of service provider but may open-upscope for telcos to discriminate by enhancingtheir terminating monopoly power over accessto end-users. As has proved the case in SouthKorea, this can be expected to raise objectionson grounds the approach undermines netneutrality.ConclusionAn internet traffic tax is not justified on groundsof asymmetric bargaining power, would harmrather than promote network investment andwould hinder the achievement of digitisationgoals for Europe. It is incoherent to tax the verything you want more of, namely digitisation. Thesuggestion of an internet traffic tax shouldtherefore be rejected.[3]

2. ContextThis paper focuses on evaluating the propositionin a report by Axon1 commissioned by theEuropeanTelecommunicationsNetworkOperators’ Association (ETNO), that imposing afee on data traffic for some content andapplication providers and transferring themoney to telecommunications networkoperators would be both fair and beneficial.A similar proposition was put forward a decadeago and rejected at the time. Nevertheless, thequestion of whether some content andapplication providers should in effect be taxedandthemoneytransferredtotelecommunications network operators hasagain arisen and needs to be considered, takingaccount of available evidence.The Axon report puts forward a proposition butdoes not answer any of the questions that needto be decided to evaluate that proposition.The supporting report on costs by FrontierEconomics likewise, on examination, does notaddress the question Axon suggests is relevant,namely what are the incremental costs of datagrowth? Even the answer to the question wouldbe partial because data growth also involvesbenefits, both in terms of what application andcontent use enable but also in terms of theimpact of induced demand on telcos revenuesand the business case for network investment.This paper discusses evidence in relation to thebalance of costs and benefits for telcos fromdata growth, the question of whether there is aproblematic power imbalance in relation topeering and transit markets and the overallimpact of an internet traffic tax on digitaltransformation – an express goal of theEuropean Commission. We conclude, in each ofthese areas, that the evidence viewed in theround runs contrary to the assertions in the Axonreport.The European Commission may however wish togo further than an in principal examination ofthe proposed internet traffic tax and considerhow it might work in practice. This papertherefore includes some closing thoughts on thequestions and evidence relevant to such anassessment.Europe’s internet ecosystem: socio-economic benefits of a fairer balance between tech giants and telecom operators,May 2022.[4]

3. Data growth is good for telcosThe Axon report argues, in essence, that datagrowth is bad for telcos. In fact, data growth isgood for telcos given that the incremental costsof data are negligible for fixed access and lowand falling for mobile access, and because datagrowth drives demand and revenues for telcos.Whilst some (but by no means all) telcos haveargued that data growth imposes costs, that thisis unfair and that they should receive paymentfrom application providers, they have alsopointed to data growth as a driver of revenuegrowth when communicating with investors.The Axon paper relies on a flawedestimate of traffic related costsAxon utilises a report by Frontier Economics forDeutsche Telekom, Orange, Telefonica andVodafone as a basis for their claims regardingtraffic related costs.2However, the footnotes to the FrontierEconomics report point clearly to the fact that itdoes not provide incremental internet trafficrelated costs, namely:“This will include costs which are variablewith respect to traffic, but also some costswhich are required to deliver any traffic butwhich do not vary with the level of trafficcarried.” Footnote 5“Given that we use accounting informationfrom the operators, we rely on theoperators to provide us with their splitbetween ‘traffic sensitive’ and ‘subscribersensitive’ costs.” Footnote 6“We recognise that a bottom-up exercisecould in theory produce improvedincremental cost estimates but this felloutside the scope of our work.” Footnote 9“These estimates are illustrative of therelevant costs and cannot be construed asindicative of a hypothetical amount ofrecovery by Telcos from OTTs.” Footnote 14Frontier Economics do not vouch for the costsplit provided by operators, recognise that analternative ‘bottom-up’ modelling approach is inprinciple superior and caution against utilisingtheir estimates as indicative of cost recoveryfrom OTTs.The Frontier Economics estimates do notprovide a basis for any of the claims Axon makein relation to costs.In any case, opportunity cost rather than costshould have been considered by Axon, namelythe net impact of traffic growth considering theimpact on network operators costs andrevenues.Below we first consider incremental costs beforeconsidering the revenue uplift associated withinternet traffic growth.Costs attributable to data growth arelow and decliningIncremental costs for fixed access networksFor fixed broadband access, for the last mile‘capillary network’, incremental data relatedcosts are for practical purposes zero i.e. existingnetworks including copper networks could carryvastly more data than they do (peak bandwidthdemand may support an upgrade from copper tofibre but data growth per se does not).3Frontier Economics, Estimating OTT Traffic- related Costs on European Telecommunications Networks, March ope/?utm campaign Telco&utm content twitter&utm medium social&utm source twitter3 Even a 10 Mbps link could carry 26 TB per month, well in excess of typical existing data traffic per household.2[5]

The fact that capillary access incremental trafficrelated costs are essentially zero and core andtransit costs are very low underpins the marketshift to fixed access tariff packages that arepredominantly unlimited in relation to data use.Figure 1: Mobile unit cost and revenue estimatesIncremental costs for mobile access networksTo meet traffic growth in mobile networks, onceexisting cellular sites become congested, moresites (cell splitting), and/or more spectrumand/or more efficient technology (transitions tohigher ‘G’s) is required i.e., there is anincremental cost. 0.60However, mobile operators are free to reflectincremental costs in their tariff structures (sayvia tiered data plans) and the incremental costsassociated with data growth are in any case lowand falling.In 2010 Ericsson estimated the costs ofaccommodating mobile data over 2G/3Gnetworks at less than 1/GB.4 More recentestimates are lower still, reflecting rapidproductivity growth in relation to mobilenetworks.Ericsson (2020) estimated the cost per gigabyte(CPGB) and revenue per gigabyte (RPGB) formobile broadband and fixed wireless access.5The estimates are shown in Figure 1. The costsfor mobile broadband are as low as 0.1/GB.Mobile unit cost and revenue es1matesCost/GB (worst case) 1.20 1.06Cost/GB (best case) 1.18Revenue/GB 1.00 0.80 0.40 0.20 0.00 0.32 0.08Mobile 0.16 0.04Fixed wireless accessSource: CommunicaFons Chambers, EricssonEricson estimate mobile data traffic of 11.3Exabytes (an Exabyte is 10 18 bytes or 1 billionGigabytes) per month for Western, Central andEastern Europe for 2022.6 Assuming trafficrelated costs of approximately 0.1 per GB thisamounts to 0.1 bn per month or a little over 1bn per year.This estimate is an order of magnitude lowerthan the cost estimate reported be Axon of 1322 bn. Further, traffic growth is also a revenuedriver for mobile, and the net ‘opportunity cost’of traffic growth may be negligible or negative(which would be consistent with positivecomment by telcos regarding traffic growth toinvestors).Incremental costs for core and transitnetworksThere are some costs associated with trafficgrowth in core and transit networks common tofixed and mobile access, but these are low.Further, prices for IP transit continue to fallrapidly:7“Across a range of markets, 10 GigE pricesfell 18% compounded annually from Q22018 to Q2 2021. A comparable sample of100 GigE port prices fell 30% over the sameperiod.”Greger Blennerud (Ericsson), Mobile broadband – busting the myth of the scissor effect, usting-the-myth-of-the-scissor-effect5 David Wait (Ericsson), Understanding the Economics of 5G Deployments, June mics-of-5g-deployments6 Ericsson, June 2022 data. ility-report7 sit-price-trends4[6]

In any case tech companies invest in networks,for example, in subsea fibre-optic cables toensure capacity is available to meet growth andreduce international connectivity costs for thebroader ecosystem.8 The predominant IP modelis settlement free peering. Any costs, net orrevenues and investment, may therefore beambiguous and are in any case likely to bemodest.ConclusionThe costs of data growth are low, ranging fromzero for fixed capillary access to around 0.1/GBfor mobile broadband. Core and transitnetworks costs are very low and declining,content and application providers invest intransit networks and the predominant IP modelis settlement free peering.The revenue uplift from data growthis positive for telcosAs BEREC noted when this issue arose in 2014:9“Ultimately, it is the success of theCAPs which lies at the heart of the recentincreases in demand for broadband access(i.e. for the ISPs very own access services).”Yet Axon focuses on costs and does notconsider the revenue uplift from data growth.Fixed accessFixed access tariffs are typically unlimited withrespect to data consumption reflecting the zerocost of data carriage in the last mile accessnetwork and low costs in the core network.However, fixed access providers benefitindirectly from data growth since data growthreduces the risk of mobile substitution and,given that data growth is correlated with peakbandwidth demand, is part of the pitch toupgrade to fibre. Illustrative is the March 2022letter from EE advising customers of a priceincrease which commented as follows:“Since 2018, broadband usage hasincreased by 90%, so having a connectionyou can count on matters - a lot. That's whywe're bringing EE Full Fibre, cutting-edgebroadband that can handle anything withblistering speeds of 900Mbps, to morepeople than ever.”Telcos benefit from data growth in relation tofixed access since data growth involves very lowcosts (thus the shift to unlimited tariffs) butpromotes fixed access customer retention and,via the correlation with peak demand, promotesfibre adoption.Mobile accessGiven that mobile access does involveincremental costs, and because mobile networksmay have localised congestion, mobile operatorsoffer monthly mobile data tiers at different pricepoints.Data growth therefore benefits telcos in relationto mobile access since data growth encouragescustomers to migrate to higher priced packageswith larger data allowances.What telcos say to investors differsfrom what some say to policy makersWhat telcos say to investors, and what some ofthem say to policy makers, are starkly different.The structurally separated fixed operator Chorusin New Zealand makes a virtue of applicationsincluding Netflix and data growth:“The unrelenting growth in demand fordata, the increasing reliance on both highspeed download and upload performance,as well as the emerging awareness of fibre8 The Economist, Tech giants are building their own undersea fibre-optic networks, October -optic-networks9 BEREC's comments on the ETNO proposal for ITU/WCIT or similar initiatives along these lines, November 2012. register/subject g-these-lines[7]

broadband’s contribution to sustainability,are all underlying trends that support ourbusiness.”10Mobile operators monetise data traffic growtheven more directly, and this is reflected in theircommunication with investors:“Vodafone VOD.L, the world's secondlargest mobile operator, has raised its fullyear earnings forecast for the first time inrecent history, as customers switch to usingmoremobiledataontheirsmartphones ”11Increased demand during Covid 19 is viewed ashaving brought forward digitalisation anddemand, and this is seen as positive:12“The world has changed because of thepandemic," Chief Executive Nick Read toldreporters on Tuesday."We see a compelling opportunity for highgrowth given the step change we've seentowards a digital society over the past year.Importantly, this growth opportunity existsin both Europe and Africa."Telefonica are also positive about data growth:13“Surging demand for mobile data is theclear driver for future growth in thebusiness”Suppressing data growthprove harmful overallwouldUltimately it is the benefits consumers andcitizens derive from services and applicationsthat matter and suppressing data growthassociated with applications can be expected toresult in harm overall.As the European Commission has noted:14“Data driven applications will benefitcitizens and businesses in many ways. Theycan: improve health carecreate safer and cleaner transportsystemsgenerate new products andservicesreduce the costs of public servicesimprove sustainability and energyefficiency”Discouraging the development and use of datadriven applications via an internet traffic taxwould undermine all of the above potentialbenefits of data driven applications.The real problem – not enough datagrowth?Absent rapid data growth, one might expect adecline in revenues, as telcos unit costs continueto decline due to technology advances andproductivity gains:15“Advances in these technologies have beenvery rapid in the past 25 years and continueat blistering rates to this day. Withoutcontinued increases in internet technologyChorus Annual Report 2021. Page 12., Vodafone lifts profit view as customers ditch wifi for mobile, 14 November roup-results-idUSKBN1DE0PI12 Reuters, Vodafone ramps up investment to capture growth opportunity, 21 May ts-12-drop-full-year-earnings-2021-05-18/13 Telefonica, Mobile data, how is Telefónica Europe capturing this growth opportunity? Deutsche Bank European TMTConference, September 2020. tes/5/2021/10/100910 european TMT conference.pdf14 European Commission, A European Strategy for Data. strategy-data15 David M. Byrne and Carol A. Corrado, The Increasing Deflationary Impact of Consumer Digital Access Services, July 15,2020. er-digitalaccess-services-20200715.htm1011[8]

and capacity from 2010 to 2015, the worldcould not have achieved the reported 29percent per year increase in IP traffic andnearly 78 percent per year increase inwireless data traffic that it did during thisperiod ”IP transit costs and prices have also fallenrapidly, whilst the capacity/speed of fixedbroadband access has increased from dial-upspeeds of 56 kbps to fibre in the 100 Mbps to 1Gbps range today.Figure 2: Mobile data growth is not exponential– it is declining60Western European mobile data - past & forecast 80%GB per monthAnnual growth rate504030201002018 2019 2020 2021 2022 2023 2024 2025 2026 202770%60%50%40%30%20%10%0%Source: CommunicaFons Chambers, Ericsson (June, 2022)Viewed from this perspective continued growthin data demand is essential to maintain existingtelco revenues and to monetise new investment.Fixed broadband traffic growth is also slowing.17Whilst the response to Covid-19 tended to liftdata consumption overall, the trend for bothfixed and mobile access has been declininggrowth i.e., data growth is not exponential.Data growth is good for telcos, in addition to thebenefits associated with digitisation and use ofdata driven applications throughout Europe.Mobile data growth is forecast to continue todecline, for example, Ericsson project thatgrowth per smartphone in Western Europe willcontinue to decline16, see Figure 2 (a similardecline is forecast for Central and EasternEurope).ConclusionArguably the real challenge telcos face isdeclining data growth, rather than excessivedata growth. Declining data growth may alsosignal a slowdown in adoption and use of datadriven applications.From these perspectives content andapplications development, and the data flowsthis involves, should be fostered rather thantaxed.Ericsson, June 2022 data. ility-reportKenny, Patterns of fixed traffic growth, 2021. -internettraffic-growth-2021.html1617[9]

4. Giving telcos money and taxing content and applicationswould not foster investment – it would harm itIt is argued by Axon that taxing content andapplication providers and giving the money totelcos would promote network investment,particularly in very high-capacity fibre and 5Gnetworks.To illustrate, a key part of the business case for5G is that it can accommodate data growth viamore spectrally efficient technology and thescope to utilise additional spectrum bands. AsEricsson put it:19Giving telcos money would not fosterinvestment“Growth in mobile traffic is among theforemost economic drivers of nextgeneration wireless networks.”Simply giving telcos money would not changethe price or demand for network access18 andwould not therefore impact investment directly.Rather, value would simply be transferred totelco shareholders.If data growth is reduced by a tax on data, thebusiness case for 5G investment would in turn bediminished.One could seek via additional regulation toensure that any additional money was invested.But doing so is likely to be challenging in termsof verifying additionality.Taxing internet traffic would reduceinvestment and innovation inapplications and networksIn 2014 a mobile internet traffic tax of 150forints per gigabyte (around 0.50) wasproposed in Hungary.20 The proposed tax wasmodelled21 and would have had a negativeimpact on data growth, network investment andconsumer welfare.22 The impacts of the tax ondata growth and economic welfare arereproduced in Figures 3 and 4. The estimatedimpacts were substantial.Whilst giving money to telcos would not fosterinvestment, taxing internet traffic woulddiscourage the use and development ofapplications on which network investmentdepends.Unless telcos lowered the price of broadband and/or data which might be expected were access genuinely a two-sidedmarket.19 David Wait (Ericsson), Understanding the Economics of 5G Deployments, June mics-of-5g-deployments20 Reuters, Hungary plans new tax on Internet traffic, public calls for rally, 22 October net-tax-idINKCN0IB1AO2014102221 A tax has the same impact irrespective of who pays, namely an internet traffic tax would have the same impact andultimate burden irrespective of whether telco’s, consumers or application providers were taxed.

growth in core and transit networks common to fixed and mobile access, but these are low. Further, prices for IP transit continue to fall rapidly:7 "Across a range of markets, 10 GigE prices fell 18% compounded annually from Q2 2018 to Q2 2021. A comparable sample of 100 GigE port prices fell 30% over the same period." 1.06 0.16