Lessons Learned from Modelling Global Affordable Broadband Strategies

An estimated investment of ~$2 trillion is required under business-as-usual conditions to achieve universal broadband via terrestrial 4G infrastructure (delivering at least 10 Mbps) in developing countries. However, cost reducing measures could bring down this figure by up to half!

In this article I summarize the findings from spending two years modelling the global capacity, coverage and cost of 4G and 5G universal broadband strategies. This work has been part of the World Bank’s 5G Flagship and the meticulous analysis involved has led to a new paper entitled “Policy choices can help keep 4G and 5G universal broadband affordable” (see link below).

Motivation - Delivering Universal Broadband

This challenge is already enshrined in the United Nation’s Sustainable Development Goals (Target 9c) to provide universal and affordable access to the Internet to all by 2030. The key point here, in my opinion, is the word affordable. There is no shortage of assessments addressing this topic. Yet, to their detriment, few focus on affordability. Or indeed, consider the viability of deployment strategies. That’s a key aspect we wanted to address in our paper.

The other important issue is the desired level of broadband capacity targeted universally. The UN Broadband Commission has been discussing ~10 Mbps per user, which we therefore assess in this work (although there are many limitations to this target, I won’t go into these here).

All technologies will play a role in SDG Target 9c, but this work focuses on cellular because it provides cost effective wide-area broadband.

What Options Do We Have to Keep 4G and 5G Universal Broadband Affordable?

There are three key areas identified which affect the affordability of universal broadband strategies:

  1. The choice of technology. This covers both network access (e.g., 4G vs 5G) and then how you connect new towers back into the internet (known as ‘backhaul’) (e.g., fibre vs wireless microwave). Note, 5G comes in multiple flavours, from 5G Non-Standalone (NSA) to Standalone (SA), with most 5G currently being deployed being NSA.
  2. The business model. Specifically the degree of infrastructure sharing that is allowed. Traditionally operators built their own networks. Increasingly, with weaker economic deployment conditions, operators have been exploring passive (non-electronics) or active (electronics) infrastructure sharing.
  3. The fiscal and regulatory landscape. This includes spectrum license costs, taxation and any other regulatory requirements placed on mobile operators.

All Models Are Wrong, But Some Models Are Useful

Despite quantitative models often being poor representations of reality, they underpin how we learn about the world, and are essential learning tools. Recognizing these limitations, and being fully transparent, we thus made the code open for everyone to access.

The broadband assessment model developed allows us to undertake comparative evaluation of different broadband infrastructure strategies. It’s a ‘bottom-up’ type of model, which just means we try to statistically estimate how many users there are in local statistical areas (1 km^2) and how much the Average Revenue Per User (ARPU) is likely to be. Least-cost network designs are then used to try connecting users. The comparative analytics produced enable us to explore the cost of each design.

The model is based on user-cross subsidization. So, after a mobile operator has taken a fair profit, excess profits (from mainly urban areas) are reallocated to less viable (usually rural) areas (which is essentially akin to a coverage obligation). When no more profits can be reallocated, public subsidies are (unfortunately) required.

We apply a cost-minimization approach to assessing universal broadband strategies. While this has limitations (like any method), the results make interesting reading.

The Findings

Our results suggest that if you really want to deliver at least ~10 Mbps per user universally via terrestrial infrastructure, you would probably need to invest approximately $2 trillion over the next decade. This is a much larger number than other estimates. Firstly, that’s because we don’t make monumental assumptions that the hardest 20% of users will be served by satellite! Our model uses 100% terrestrial 4G infrastructure. Not least because I’m not sure it’s possible to serve 20% of the global population by satellite, given physics-based constraints and spectrum limitations.

Secondly, this assessment also includes a statistical method to estimate quality of service. Previous work has focused more on a general peak speed, but doesn’t say how this would play out in the busiest hour of the day when everyone wants to connect. Importantly, if we played with all the cost reducing measures in the model, we found we could reduce this investment cost by up to half! While this finding – like every other universal broadband assessment – is riddled with assumptions and caveats. The main point is this. Governments have a whole range of ways to reduce the cost of infrastructure deployment.

If universal broadband is really the aim, then governments should consider:

  • Picking coverage over capacity. Ensure everyone has at least a basic broadband service. This means 4G using a wireless microwave backhaul may be the preferred option to provide a basic service. If you only have limited resources, you’re best off spreading a small amount of investment equitably. Rather than pouring money into future proofing a few areas and then running out of cash to connect other users.
  • Exploring infrastructure sharing options. Especially for the very hardest-to-reach places. We know infrastructure competition produces dynamic societal outcomes. But if you can’t viably support building even a single infrastructure under business-as-usual conditions, then we need to look at new sharing strategies in these areas of market failure.
  • Focusing on developing a favourable regulatory environment. Move towards reducing licensing and infrastructure deployment costs, and incentivizing infrastructure deployment in the hardest-to-reach areas.


Many low-income countries managed to leapfrog to 2G cellular, without having any previous fixed telecom network. Recently there has been quite a bit of hype around the idea of ‘leapfrogging’ straight to 5G. Thus, skipping deployment of 2G/3G/4G. We spent a lot of time working through this idea.

The key issue concluded was that from the disadvantaged users you want to connect, we might end up leaving some behind via this strategy. Leapfrogging a cellular generation (or two) would ultimately be a supply led initiative, therefore without commensurate support on the demand-side to ensure all users had 5G-enabled devices, many people would be unable to access the new network.

For example, rural users in low-income countries may still be on 2G or 3G devices, let alone 4G. Granted, device costs are decreasing rapidly, but even a $30 device cost needs to be evaluated against the annual income a potential user has. Indeed, this is a substantial amount for many on the poverty line (living on less than $2 per day) that we would like to connect.

Further materials and acknowledgements

The full citation for the open access paper published in Technological Forecasting and Social Change is as follows:

Oughton, E.J., Comini, N., Foster, V., Hall, J.W., 2022. Policy choices can help keep 4G and 5G universal broadband affordable. Technological Forecasting and Social Change 176, 121409. https://doi.org/10.1016/j.techfore.2021.121409

The Python Telecommunications Assessment Library (Pytal) code can be found at the pytal GitHub repository.

Thanks are due to many people and institutions including:

  • The governments of Kenya, Senegal, and Albania for providing cellular site data.
  • Telecommunications regulators in Malawi, Uganda, Kenya, Senegal, Pakistan, Albania, Peru and Mexico for research review meetings.
  • Mobile Network Operators including Entel (Peru), AT&T (Mexico), Telenor (Pakistan), Sonatel (Senegal) and TNM (Malawi) for research review meetings.
  • Valuable feedback from the World Economic Forum and the GSMA.
  • Specific expertise provided by Peter Curnow-Ford, Adnan Shahid, Andy Sutton, and Tom Russell, and Andoria Indah Purwaningtyas for operational support.
  • Funding from the World Bank 5G Flagship, UKRI grant EP/N017064/1 and a UK EPSRC Impact Accelerator Award.
Edward J. Oughton
Edward J. Oughton
George Mason University
Jim Hall
Jim Hall
Professor of Climate and Environmental Risk

Prof. Jim Hall FREng is Professor of Climate and Environmental Risks in the University of Oxford and Director of Research in the School of Geography and the Environment.