If you are reading this, you are already within a narrow cross-section of people whose eyes have not rolled up into their heads at seeing the term “spectrum auctions” in a headline. The rules of how wireless spectrum is made available to service providers are arcane and, at various times, demand some knowledge of physics, engineering, economics, game theory, law, political science and probably a couple of other fields I haven’t thought of. It is frankly quite daunting. But access to wireless spectrum is one of the chief determinants of affordable access to communication and the current system of making it available is failing to stimulate competition and is a barrier to affordable access in rural areas.

My previous article, Rethinking Affordable Access, goes into some detail about how access growth is slowing due to the challenges to delivering affordable access in more sparsely populated rural areas where income levels are generally lower than in urban areas. In the article, I propose the creation of an enabling regulatory and financing environment for small-scale operators, both non-profit and for-profit, to tackle rural access challenges with new technologies and new sustainability approaches.

One of the biggest stumbling blocks to making this happen is lack of access to wireless spectrum. There are other important access technologies like fibre optic cables and satellite services (I will come back to them in a later article) but the most important technology for the rural last mile, particularly in economically poor regions, is wireless communication.

Wireless spectrum comes in two types, licensed and license-exempt. License-exempt spectrum for broadband access is typically WiFi technology. License-exempt means that WiFi does not rely a license, issued by the regulator, granting exclusivity to a particular frequency. Instead it relies on compliance with requirements to operate at low power levels and on device protocols designed to facilitate co-existence with other WiFi equipment. This has been spectacularly successful. WiFi is a multi billion dollar a year industry that has led to WiFi being embedded in everything from watches to refrigerators to shipping containers. But WiFi has limitations. The low power requirements mean that the signal doesn’t go that far. If you want to roll-out WiFi as a rural access technology, you need a lot of hotspots to create coverage. Also, the nature of WiFi is that it doesn’t penetrate structures particularly well, so hotspot use often demands travelling closer to the access point, which has implications for marginalised groups/individuals.

Licensed spectrum, such as is used by mobile operators for 2G, 3G, 4G, etc services, does not have these limitations. The exclusive nature of the spectrum licenses for mobile services mean that much higher power output levels are permitted. Thus a single mobile base station can cover a radius of several kilometres. The actual range is dependent on a variety of factors from the radio frequency used, terrain, tower height, and type of radio technology.

In African countries, when the first modern mobile networks were getting off the ground on the continent (around 1994), GSM spectrum licenses were generally given to anyone prepared to commit to building out a national network. There was no shortage of spectrum available. There were annual fees associated with licenses but no up-front, pay-to-get-in-the-door fee.

At the same time, in the United States, the world of spectrum assignments was changing. Demand was beginning to outstrip the administrative availability of spectrum and the regulator determined that market forces should determine the assignment of high-demand spectrum. Thus the spectrum auction was born. The underlying philosophy, as articulated back in 1959 by economist Ronald Coase, was that a property-rights model would create the same efficiencies as seen in real-estate markets where property is sold according to the value expected to be derived from it. Spectrum auctions were intended to “discover” the appropriate price of a particular spectrum frequency.

African countries were slower to adopt auctions as a mechanism for assigning spectrum. Nigeria pioneered auctions beginning in 2002 but it wasn’t until more than a decade later that other countries on the continent began to follow suit. Since 2013, there have been a number of spectrum auctions in African countries and it is worth looking at them collectively to try to understand their impact in the last 5-6 years.

The orange bars in the graph below looks at the price paid per GHz per person in each auction. It seems reasonable to divide spectrum fees by population as a means of comparing countries. Strictly speaking Kenya did not hold an auction but as operators did pay an up-front fee for a spectrum license, I believe it is comparable. Also Mozambique in 2013 had an auction but did not issues any licenses as there were no bids. The data there represents the announced reserve price. The blue bars represent GDP (PPP) per capita. The goal of the graph is to explore whether there is a correlation between the price asked/paid for spectrum and the relative economic wealth in the country.

Graph of spectrum auction fees in African countries contrasted with GDP per capita

If, as Coase predicted, spectrum auctions are the key to efficient outcomes, one might expect to see some correlation here. There isn’t any. However, the single biggest factor affecting the spectrum auction outcomes is not what operators are willing to pay for spectrum but rather the reserve prices set by governments. Reserve prices are the minimum opening bid in a spectrum auction. In the vast majority of cases, where they bid at all, operators bid the minimum amount.

For low (and not so low) income countries, spectrum auctions represent an economic windfall to the treasury that is almost impossible to resist for governments to resist. Telecommunications is a lucrative business and governments see auctions as a way of ensuring they get their share. If operators are willing to pay the price, it is a strategy that is hard to argue with for cash-strapped governments.

HOWEVER, there are two negative outcomes of spectrum auctions that do not receive the attention they deserve.

A Firewall to Competition

Prices paid for an LTE spectrum license in the last few years in Sub-Saharan Africa range in the tens of millions of dollars, with Tanzania at the low-end selling 20MHz of 700MHz spectrum for $10M in 2018 and Ghana at the high-end selling 20MHz of 800MHz spectrum for$67.5M in 2017. In both cases, those were the reserve prices set by the government. The willingness of operators to pay millions of dollars just for the privilege to using an LTE frequency is a testament to how much money they expect to make from their network. Although, if you are not the dominant operator, it is still a risky bet.

The problem here is that auctions have now become the single biggest barrier to market entry and hence to competition. We live in a world where everything related to telecommunications infrastructure is dropping in price. From fibre optics to radios to base stations; everything is getting cheaper. The volume of demand, mass market manufacturing and a shift towards software-defined radios has changed the telecommunications landscape the point where Facebook’s Open Cellular initiative is promising an LTE base-station for under $1000. Not only is the equipment much cheaper but it is typically consuming much less power as well meaning that network OPEX as well as CAPEX costs are lower.

Map of terrestrial fibre optic networks in Africa - https://afterfibre.nsrc.org

The growth of undersea and terrestrial fibre optic infrastructure in the last ten years has changed backhaul as well. A single fibre-optic backhaul network can serve multiple high-demand operators without breaking a sweat. This means that a new operator often has no initial need to build their own backhaul network.

Thus for a new operator, backhaul costs are purely OPEX and can scale with demand. Radio equipment is in the tens of thousands of dollars. In theory, this should be an environment that is ideal for small-scale operators whether startups or cooperatives. Oh, except for millions of dollars you need for that spectrum license. Kenya has recognised this issue to some degree and is now offering a 10 year period for operators to pay off the $25M for a new 700MHz license. That’s a step in the right direction but it still makes starting a telecommunications company a massive risky bet.

If I were to consider starting up as a network operator in Kenya, I would be obliged to make the following comparison. For $25M, I could have a spectrum license for the 700MHz band but that $25M gets me nothing except the right to use that frequency. For that same $25M, I could trench about 5000km of fibre across the country. Fibre has a lifespan of at least three (if not 4or 5) times that of wireless technology and is relatively infinite in capacity. I could build WiFi infrastructure for a hundred dollars an access point all along that network. Every smartphone has WiFi these days. Suddenly that $25M spectrum license is not looking that great to me.

The upfront charge for spectrum is a massive disincentive to the innovation that comes from small businesses. It is no secret that big operators have a kind of love/hate relationship with spectrum auctions. They want them to be high enough to dissuade the competition but not too high.

A Disincentive to Rural Access

The problem of spectrum auctions is worse than just shutting out new market entrants. High prices paid at auction for spectrum don’t just firewall off competition they are an active disincentive to rural roll-out. Rural markets are a challenge to operators. People are more spread out, incomes are lower, terrain is often more challenging, and complementary infrastructure like power networks are often shaky or non-existent. A spectrum license costing tens of millions of dollars added to the balance sheet puts extra pressure on operators to generate a return. Inevitably there is going to be a focus on deploying infrastructure into the areas where the most revenue is going to be generated at the lowest cost.

Regulators attempt to address this problem by including rural ‘coverage obligations’ with the license in an attempt to oblige successful bidders to roll out access into rural areas. This hasn’t worked out amazingly well. Some operators find it easier to simply pay coverage obligation penalties because it is cheaper than complying with the obligations. Even where operators do establish rural coverage, their access technology, pricing, and offerings may not be well-designed for rural populations resulting in lower uptake.

It is ironic that while operators are clamouring for access to spectrum to be able to better serve urban environments with broadband, hundreds of megahertz of IMT spectrum lies unused in rural areas. This is true in most countries, not just Sub-Saharan Africa. Last year at the Dynamic Spectrum Alliance Global Summit, a member of the UK House of Lords, David Willetts, spoke about spectrum utilisation in rural UK. He presented data from the regulator indicating that, 5 years after auction of 2.6GHz spectrum, it remained 98% unused across the country.

There is a practical alternative

There ought to be a way to make unused spectrum in rural areas available to solve the challenge of affordable access. Happily a mechanism exists. It’s called “secondary use” and it is what Television White Space (TVWS) or dynamic spectrum technology is based on. Secondary use spectrum allows for the use of spectrum by someone who isn’t the primary license holder as long as their use does not interfere with the primary license holder. With TVWS, this is managed by a geo-location database which maps the television broadcaster signals and allows any TVWS device user to determine what television channels are empty in a particular area at any one time and consequently are available for use.

This is not a new concept. Advocacy for TVWS has been going on for some time. In the United States, the principles of dynamic spectrum management are being extended into the 3.5GHz band with the Citizens Broadband Radio Service. Sadly, almost everywhere else in the world, the 3.5GHz band is being auctioned for 5G.

There is an opportunity to adapt secondary spectrum use to LTE frequencies. Allow small-scale operators that are willing to deliver service into rural areas access to licensed but unused LTE spectrum. That might be unassigned spectrum or it might be assigned spectrum that is unoccupied.

To address the inevitable push-back from operators who have paid millions of dollars for their spectrum licenses, offer them a rebate on their universal service fund contributions. For every X kilometres of rural geography that is put into secondary use, they are forgiven Y % of their universal service contributions. There are many possible variations on this theme that might compensate incumbent network operators for spectrum they were not planning to use anyway.

The concept that spectrum licenses shouldn’t confer a right to exclusive use of a given frequency but rather, more modestly, a right to protection from interference was first proposed (that I am aware of) in 2012 by the US President’s Council of Advisors on Science and Technology entitled Realizing The Full Potential Of Government-held Spectrum To Spur Economic Growth. It was this report that led to the development of the CBRS band in the US.

Mexico is one of the few countries that has formally recognised the challenge of mobile coverage in underserved rural areas by setting aside GSM spectrum specifically to promote network development in areas where it is not economically viable for the big operators.

What is needed now is to put secondary-use and rural strategies together into. Secondary access to unused LTE spectrum in rural areas could dramatically change the face of affordable access and open up innovation in service delivery.


Banner image – CC0 Public Domain

I am grateful to Carlos Rey-Moreno and Peter Bloom for feedback that made this a better article. Any errors are all mine.

This article was made possible through my ongoing association with APC’s LOCNET Project, Mozilla, and NSRC.

Posted by Steve Song

Local hands building local networks. Policy advisor @mozilla Consultant https://apc.org Collaborator @nsrcworld Founder @villagetelco