Yesterday, Brookings launched a book, aptly named Breakthrough, coauthored by 15 people active in the worlds of technology, innovation, investment and development.

The book is a truly majestic sweep of what technology innovation can do to get us over the SDGs bar by 2030, or simply put: make this planet a beautiful home for everyone. And, no, I am not being so kind just because I contributed a chapter (here); there are some truly compelling vistas of the near future in the volume.


From creating a multilateral bank dedicated to non-human species to applying synthetic biology to rural agronomic transformation, the authors’ visions are surreally beautiful in their combination of the practical with the truly lofty.

Contributing to Breakthrough was a very nice opportunity to sit back and think about the three passions of my life: technology, activism and social innovation/entrepreneurship. I am at a point where, after nearly two decades in the trenches, I am revising most of what I know. A group of connected agro-tech projects in Malawi, with which I am intimately familiar, thus served as the perfect backdrop for me to use my chapter in the book to tease out my new horizons and perspectives.

The words “transmediary” and “transmediation” are the brushes and ink-pots with which I try to bring out the features in the landscape I am seeing in my world right now, a world in which standalone social enterprises are sadly incapable of making enough of a dent to change the world. A world in which technology is changing much too fast for most people to properly trace its real contours at the intersection of Political Economy and the Knowledge Economy. To get to my point in the chapter, I had to first, in my own head, get to basics.

What is technology?

A technology is a system that produces quick and precise responses to a defined problem.

Therefore, we have to start with “problem”. In the context of the SDGs and changing the world to benefit all its inhabitants, we are clearly talking about social problems.

Given the times, let me give you a familiar example: the last major episode of the bubonic plague in London, in 1665. City planners began to believe that stray animals were involved in spreading the disease. The King therefore ordered the wholesale massacre of dogs and cats. By some estimates some 240,000 of them.

Well, they were right in part. Animal vectors were a major spreader of the disease. But they got the wrong animal.

Turns out rodents were most to blame. So, by killing cats and dogs, the dangerous furry balls now had free rein to run amok, exacerbating the plague.

In our far more complex world, problems are obviously harder to define and diagnose properly compared to Stuart England, so we can expect more unintended consequences when we go chasing silver bullets. But, even worse, problems nowadays also tend to have multitudes of secondary and tertiary effects and adaptive features that confound precise gauges of their causes and impacts.

An easy way to put it is that all problems are connected, and so all solutions must be connected.

Because our fast-changing world, however, also requires very quick and precise solutions, it means that all solutions must be technological at their roots, and all the technologies must be connected.

It is amazing how poorly appreciated this fact is.  

Take the automobile for instance. And compare it with the modern notion of a self-driving car.

The classic automobile was described in terms of how many component manufacturers – about 200 in fact – had to collaborate in its making, and how many inputs – about 30,000 – were needed to get it on the road. Many treatises have been written about the interconnectedness of the modern supply chain and how only connected solutions can address concerns like the need to shift to greener inputs and outputs whilst addressing labour rights.

But in the 20th Century world, once the car was made, it was safe to switch our analysis to treating the car as a discrete object with a discrete impact on society. We can look at emissions, vehicular accidents, the blight of traffic on suburbia and many other such technology-society interactions by treating the technology and its producers as discrete elements in a classical mechanical plane.

A self-driving car, on the other hand, is going to be very different. Its ontology is more in the quantum mechanical plane, to use a bad metaphor. To work effectively, it must continuously be wired to other solutions and solution-providers. Municipal risk management systems. Novel satellite complexes. Networked semantic IoT. New safety algorithms developed by collaboratives. Connected car insurance platforms. A whole host of as yet poorly understood meshes of datacenters and super-algorithms developed and managed by a wide range of actors will form the quantum mesh to make the self-driving car a truly viable institution in a world grappling with automation.  

What that means is that the actual technology will be EMERGENT through networking across solutions. It is not like Tesla will screen its suppliers, make the car, and release a compact object whose interactions with society can then be mapped in terms of which problems it solves and which new ones it creates. Rather the problems and the solutions will emerge in tandem and evolve alongside the ongoing interventions of a whole host of other actors whose evolving contributions will shape the car’s interaction with society.

Because the problems and solutions will be emergent through continuous relations among actors, which as we all know is a function of power but also of the passion of many who will create these sub-solutions, the worlds of technology, regulation, activism, design and social impact will fuse, and grate far more intensely than we can imagine today.

Consumer safety activists will demand “standardised APIs” between pedestrian safety algorithm stores and municipal road safety engines, and reinsurers will redesign contracts to accommodate on-demand predictive analytics on crashes and the likes. Advocacy won’t be limited to people on the “outside” wanting to check the technology. Activists will be able to “embed” in the technology-complex itself.

Transmediation is a powerful concept in aligning critical tradeoffs on the path to the SDGs

Most of the problems that confront us will be solved by technologies that have emerged through continuous interactions of power and passion, of activist technologists suddenly relevant through other instruments besides the market. The dissolving boundaries between government and business domains (consider the number of public cyber-security systems managed by private contractors) will become even more complex as civic actors outside the two spheres also start to assert themselves.

As unintended consequences, feedback loops, and complex causal rings make it clearer to most people that every problem is in fact a problem-solution-problem-solution chain, the infrastructure of our lives will become woven into quantum entanglements connecting everything.

Fintech solutions will not be one thing and government anti-money laundering systems another thing. Crypto will not be one thing, and regulations to deal with its effects another thing. Different actors with different angles to these issues will come at these problem-solution-problem chains with multiple chimeric and intermediate responses, which working together integrate problems and solutions into manageable situations. This is the only way for society to reconcile its need for both speed and precision in addressing rapidly mutating risks.

 Those who stand above it all and succeed in creating higher-order systems for integrating the longest problem-solution-problem chains into effective and stable ecosystems where risk is effectively managed, even as social and economic value continues to multiply, are those I have called, “transmediaries”.

They are transmediaries because by taking a systems entrepreneurship view, they can build transmediary platforms that remove inefficient intermediaries. How? Providers of many chimeric and intermediate solutions rely on certain nodes in a network not being able to connect well with other nodes.

Banks are bad at talking to SMEs and millennials, hence the fintech boom. But fintechs are not necessarily any better at delivering financial literacy, suppressing excessive consumer debt, or sieving out triggers of dangerous boom-bust cycles in the macroeconomy. Only system actors working across large networks to build solutions, transmediary platforms, capable of reconfiguring nodes across the financial system can have a real impact on such problems.

So, whether the problem is how to ensure that credentials from MOOCs help address employment – skills mismatches or how to improve prescription surveillance to address antimicrobial resistance, neither social entrepreneurs nor commercial geniuses building discrete business models can fix it. Only transmediation and transmediaries can.

So, what is the call to action here then? I think the starting point is recognition. Narratives can direct energy and resources. The days of the heroic entrepreneur, mega-platform or statesman changing the world have been over for quite a while now. Yet, we refuse to truly embrace the power of networks. My duty is to sound the clarion call yet again.

This is a short postscript to my earlier note on Ghana’s proposed e-Levy.

In view of certain reactions to that piece, I felt that a quick span of the experiences of other countries that have gone this route, which I had suggested offhandedly in my note, would actually illuminate some of the points rather well. So here goes.

There are, near as I can tell, seven countries in Africa that have imposed or set in motion plans to impose mobile money taxes:

  1. Kenya imposes a 12% excise tax on transaction fees a while back. Whilst such excise taxes end up being passed on to consumers, and therefore can be retrogressive, the net absolute effect might still be smaller as when competition succeeds in bringing down fees the tax effects will reduce automatically. Moreover, after all, they apply to income earned and not just movement of funds.
  2. Uganda attempted to impose a 2% fee on transaction value but retreated after a massive public backlash to 0.5%.
  3. Malawi imposed a 1% fee on transaction value and has since retreated.
  4. Tanzania wanted to introduce a graduated levy of between 0.43 cents and 4.35 USD depending on the amount being sent. The President has asked that this be reviewed downwards. Reduction in transaction volumes and value have already greeted the policy announcement.
  5. Zimbabwe has a 2% fee on transaction value.
  6. The remaining countries – Cote D’Ivoire, Congo Brazzaville and DRC – apply the tax on the mobile operator’s turnover.

None of these countries make significant revenue from e-Money related taxes, yet there is considerable evidence of significant distortions of the digital economy since the imposition of these taxes.

  • Kenya makes less than $100m a year.
  • Uganda is around $27m a year (total transaction volume has fallen by about 27% in recent times).
  • In Zimbabwe, the 2% intermediated mobile tax is on course to yield about $50 million by the end of this year.

Ghana’s projected $1.15 billion return is, as should be clear from the foregoing, completely out of line with experiences elsewhere and a sign of weak policy formulation and prior research.

If actual outturn follows the experience of the country’s Communications Service Tax (CST, aka “Talk Tax”), where wildly optimistic expectations failed to pan out, we may end up with something in the region of say, $150m, and yet would be paying $40m to prevent vaguely defined “evasion”. The $40 million set aside for collection of the e-Levy is aggravating general discomfort about the whole enterprise. Given that the budget was expected to pass around next week and implementation of the tax meant to start on 1st January 2022, it is hard to believe the Government when it says that an open, competitive and transparent procurement process had been planned. How long does it take to even write up the Terms of Reference for a serious Expression of Interest (EOI) process?

All this makes one worry that the planned $40 million e-Levy administration service could end up as badly for the country in ratio terms as the CST revenue monitoring situation involving Subah, Kelni GVG and Afriwave. At one point, the country was spending $32 million in various electronic revenue protection measures, none of which made technical sense, yet total CST take was still hovering in the region of $54 million (2017). That is to say, Ghana was spending more than half of the money it was collecting on policing the same money.

Historical experience, like cross-country comparative experience, counsel serious caution and widespread consultation about how the country goes about designing and implementing any digital tax measures.

If these measures are motivated purely by problem-solving, rather than by procurement opportunities, then why is extensive stakeholder consultation always so hard?

The Leader of the Opposition (“Minority Leader”) in Ghana’s “hung” Parliament, who has been a perfect picture of parliamentary due diligence so far in the matter of the stranded 2022 government budget, startled the public by announcing prematurely that he will accept a 1% e-levy rate.

The government has proposed the imposition of the tax on all electronic financial transactions, a mainstay of its plan to increase revenue by 43% in 2022. A 43% revenue increase in one fell swoop would be mindboggling in normal times, but things are far from normal in Ghana’s fiscal circumstances now that servicing public debt is threatening to consume 60% of all government revenues (in the first 9 months of 2021, government revenue was $7.86 billion and it spent $4.23 billion, or 54%, of it servicing loans).

What analysis was the Minority Leader’s capitulation based on? Why 1% and not, say, 1.25% or 0.75%? He doesn’t say.

It is Not just about the 1.75% rate or 100 GHS floor

At any rate, it is not the absolute rate of the e-levy that matters most but rather its entire design as a digital taxation measure that will impact the growth of virtually all digital services well outside the financial sector.

Because of this factor, leaders of policy-oriented Civil Society Organisations (CSOs) are very close to asking the Government to remove the e-levy measure from the budget altogether in order not to derail the entire appropriations process. With e-levy out and a more comprehensive consultation process underway to flesh out a more detailed digital taxation strategy, the less contentious remainder of the budget review by Parliament can continue relatively smoothly at the committee level. Of course, there are equally or even more problematic provisions in the budget, such as a promise, or more appropriately, threat, to return the unconscionable Agyapa policy to Parliament for approval. But that is another matter.

Policy oriented CSOs are yet to come to any broad consensus on the e-Levy, but my survey of the mood of leaders in the community shows growing discomfort about the poor attention paid by the Government to the fact that there are many different design options available in crafting the e-levy, each with different implications. 

The absolute rate of 1.75% and the minimum daily cumulative transactional threshold of 100 GHS before the tax kicks in, as indicated in the budget, are by themselves meaningless concepts. The revenue estimate of $1.15 billion is also almost certainly wrong. These three data points are, to repeat, pointless without a ton of caveats about how precisely the e-levy will be designed

The Budget Process in Ghana is not designed for making fresh & complex policies

Yet, the budget is not the place to roll out complete policy. Policy is first debated by key stakeholders, such as the official opposition and CSOs, finalised, and then the revenue and expenditure implications are captured in the budget. The approach taken by the Government so far puts the cart before the horse. 

The minority leader and other important commentators like him who have jumped the gun to reduce this complex debate to a rushed haggling over absolute rates (1.75%) and minimum thresholds of taxable transaction value (100 GHS) are risking a Missouri Compromise which will come back to bite all of us later.

 In this brief analysis I hope to show, in broad outline, why the Minority Leader needs to align with the cautious camp and ask for the e-levy to be subject to a wider and deeper policy evaluation process instead of the perfunctory quibbling over narrow figures characteristic of the Ghanaian appropriations process.

Our call is even more poignant now that we know that the Government has set aside a full $40 million (35% more than the total allocation to the National Health Fund) to pay vendors for very suspicious services related to rollout and collections of the e-Levy.

If it turns out, as I will argue shortly, that the $1.15 billion expected to be raised from this tax could turn out to be far less, and yet the effects of the tax may well be what economists call “distortionary”, then its risks would far outweigh its benefits, and it will be sensible to return it to the drawing board.

$1.15 billion revenue estimate in the 2022 budget is naive

Even without a detailed policy document, which the government characteristically refuses to provide, most analysts have pointed out that the $1.15 billion seems to have been derived from a naïve application of the 1.75% tax rate to the portion of the total transactional value of Ghanaian electronic financial services – roughly $130 billion – made up of transactions above $17 in value.

Estimation of the e-Money Base in Ghana circa 2021 October (computation of this figure is constrained by many serious challenges of double counting. Systems like GhIPPS Instant Pay – classified above as an “interbank payments” channel – are used to connect multiple systems like ATMs, internet banking, USSD banking etc, meaning that volumes and values reported by the Authorities under that classification could also be reflected in other categories like card payments).

Everyone of course knows that this massive value of total transactions, nearly double Ghana’s GDP, does not represent actual economic value addition or even value generation. It is more akin to how global forex trading has a turnover of more than $200 trillion every month, yet global GDP is only $85 trillion.

To understand the real underlying economic value in electronic financial services, consider that despite the over $100 billion in transactional value, the nominal balance on mobile money float (how much the telecom networks actually retain with the banks to cover the liability claims of e-money) is about $1.3 billion. MTN, the overwhelmingly dominant market leader with over 92% market share, maintains a $1.15 billion float and annually makes about $216 million in profits from the $90 billion or more that circulates across its networks (0.0024% “margin”).

Not all electronic payment channels and e-Money transactions are the same

Interestingly, however, not all electronic financial channels behave similarly, and users of those channels both imprint their behavioural patterns and are in turn imprinted back.  That fact is seriously crucial in determining both what the real take from the e-Levy would be as a direct consequence of HOW it is designed and how the tax could impact various parts of the economy.

For example, whilst mobile money transactions now exceed 3 billion per annum in volume, the quantity of debit card transactions totalling $4.3 billion in 2018 was only 61 million, yielding an average transaction volume of $71 per transaction or nearly 400 GHS per purchase, with medians and means well above the proposed transaction threshold of 100 GHS, unlike the situation with mobile money. The overwhelming majority of these card transactions were on international rails like Visa and Mastercard, which have other significant charges, because customers continue to shun the supposedly cheaper Gh-Link product being imposed by GhiPPS (transactions on Gh-Link have dropped to barely $55 million from a high of $140 million in 2017).

As an aside, the trend of failing government-imposed digital financial products reflects in the abysmal float on E-Zwich, which is a measly estimated amount of ~$25 million (described as “value on card”), up marginally from $22 million in 2018 (total value of E-Zwich transactions exceeds a billion dollars per year, by the way). The reason is partly due to high leakage out of E-Zwich, mainly into cash-out, but also the result of government ignoring consumer behavior in its design of policy.

E-Money is overwhelmingly used for offline payments not digital services

Indeed, the vast majority of electronic payments relate to paying offline merchants or peers, or taking cash out. These expenditure patterns have stayed relatively stable. ATM transactional value, a major proxy for cashout preference, for instance, has grown from ~$4 billion in 2017 to ~$5.2 billion today. Whilst POS terminal spending, a good proxy for offline merchant engagement, reached ~$1.6 billion in 2018. Together they account for most card transactions volume, with online merchant spending trailing far behind.

Internet banking transaction value, which better reflects the growth of online digital services, on the other hand, has actually dropped from a high of more than $2.1 billion in 2017 to an estimated $1.5 billion today, and registered users are down by about 5% from highs of nearly one million in 2017.

Meanwhile, the large ticket size (~$260,000) of interbank settlements (RTGS/GIS) also reflects the weak presence of small and medium enterprises and a corresponding absence of boutique banks, particularly after the so-called “clean up” exercises in the banking sector.

All the above goes to show that whilst payment channels that enhance offline cash-substitute behavior are thriving, those that will foster the growth of spending online on actual digital services (such as e-commerce, edtech, e-health, e-insurance, agtech etc) are trailing far behind and in some cases declining. Most Ghanaians still see e-money as purely an offline cash substitute and not as a means to participate more deeply in the digital economy.

E-Money is not stimulating GDP growth and transformation well enough

An intriguing validator of this view is the concept of an “e-money base” as a loose mirror of the more conventional notion of “narrow monetary base”, i.e. liquid cash and on-demand banking. In Ghana, this monetary account line item (sometimes called M0) is about $6 billion. But if one strips the figure of its on-demand banking elements, and focuses purely on physical cash (or currency) in public circulation, one gets about $3.8 billion.

When looking for an electronic parallel to this physical cash amount for analysis, a host of complicated treatments is required, which I won’t go into here. Suffice it to say that the crude figure of $10 billion (aggregate e-money float in the economy stripped of multiplier effects) in the table above should be fine for the simple arguments in this note.

The elementary equation linking GDP and inflation to money supply will yield, in this crude analysis, a physical cash velocity of 19.5 and a e-money velocity of just 7.5. That result is fully consistent with the position canvassed above that, presently, electronic financial services and e-money in Ghana are not sufficiently stimulating the creation of new services. Nor are they proving an effective boost for digital production to stimulate GDP growth. E-money is still fundamentally a cash-substitute for offline transactions. For the size of Ghana’s digital economy to reach the level of, say, Indonesia (at ~2.5% of GDP), it would need to nearly quadruple.

E-Levy will elicit behaviors that can harm digitalisation

If the analysis above holds, then we must consider, as the late Martin Feldstein used to preach, not just the revenue estimate that will result in a steady-state GDP model but also the revenue implications of consumer behavioral responses to the tax and their impact on digital GDP. Here, an elementary economic analysis tool comes to our aid: cross-elasticities.

Given the fact that the most digitized payer in Ghana today is the government, with 85% or more of government payments now digitized in the wake of successive reforms culminating in GIFMIS, compared to less than 40% of private sector payments, the political economy reality is that government will exempt government payments from the tax. It will do this not only to avoid useless double-counting from government taxing its own revenue, but in other interesting ways.

Quasi-government payments such as cocoa cash transfers by licensed buyers, SSNIT related payments and a host of others will over time also qualify for exemption due to lobbying. Because government’s contribution to the e-Money base is significant, this political economy based behavioral response must be compensated for by a shift of the burden to the private sector by government trying to game private sector exemptions.

Meanwhile, the fact that different payment channels, as discussed above, are optimized for different average transaction values, cumulative frequency of payments, and exemption benefits, shall elicit even more complicated gaming responses from private payers.

For example, lobbying is likely to lead to Government setting an upper bound for the e-levy’s application to transactions. It is unthinkable to think that the Government will be able to resist lobbying from investors repatriating their cash abroad for instance to pay the full rate on the full amount. Whatever ceiling the government sets beyond which the total tax payment stays fixed, payers will have an incentive to aggregate and compress their payments across time to benefit from the cap.

Meanwhile, away from the wholesale level and in the more higher-churn retail-point businesses (like table-top hawkers), for which the 100 GHS floor is no comfort, the ability to pool deposits is more constrained, with tragic equity effects. Some smaller payers nonetheless would be able to splice and distribute their payments across time to try and avoid crossing the 100 GHS threshold or whichever floor is set.

Then there is of course the fact that if, as the analysis above suggests, e-money is primarily a cash substitute at offline merchants, then the tendency to carry cash in order to avoid the tax will increase. Cash use in the economy has proven highly resilient. Supply has risen in aggregate terms by about 70% since 2017, and approximately 65% of all transactions still use cash. Compare this to Sweden, for instance, where only about 5% of the population still shop with cash.

The likelihood that a poorly designed e-levy will lead to Ghanaians continuing to stay away from digital services and entrenching in their use of cash, whilst restricting e-Money to cash-substitute at offline merchants only when transactions fall below the threshold or are considerably above the cap will be high. Where a tax induces such complex behavioral responses, it can have wholesale distortionary effects on a major part of the economy, in this case on digital GDP.

In sum, the wide and complex range of exemptions, not least because of government’s strong role as a payer; the interplay between caps and ceilings; and the cross-transaction elasticities (or inducements to switch from payment types) all suggest that different e-levy designs could have dramatically different impacts on the economy. We know for sure that the $1.15 billion revenue estimate in the budget is seriously flawed, as the total eligible value to be taxed is less than the naïve base of total electronic transaction value the government is using. But we can’t be too certain about all the effects until we consider multiple design options. Just knowing the daily cumulative transaction taxability threshold (100 GHS or any other number) and the absolute rate (1.75% or any other number) tells you very little about the likely impact of e-Levy.

E-Levy is NOT just a quick tax measure; it can have lasting impact & need more analysis

In fact, even from a pure monetary policy impact point of view, different e-money transactions behave differently, both in terms of their impact on monetary aggregates and the money multiplier. Empirical evidence from places like Bangladesh show that should the e-Levy encourage cashouts from wallets and discourage B2P, B2C and C2C/P2P transactions, inflationary effects can ensue. Current Bank of Ghana exogenous shock models rely primarily on vector autoregression formulas that are seriously unequipped to deal with the complex behavioral dynamics of gaming by large numbers of agents.

Lastly, before the country rushes into setting absolute rates and floors, as the likes of the Minority Leader would have it, a deeper policy evaluation exercise would also definitely benefit from cross-country comparative analysis. Renowned Kenyan Economist Njuguna Ndung’u has shared empirical evidence from Kenya that shows emphatically that even plain vanilla excise taxes on financial services can dampen growth. What Ghana is proposing to do, depending on exact design, may well be unprecedented worldwide and will therefore require careful modelling by looking at its divergences from global experience.

So, once again, the Minority Leader and all the major commentators pushing for a quick and ready compromise based purely on Parliamentary assent to the absolute rate and floor, but without further legislative guidelines on other features such as caps, exemptions and variations and slopes, are seriously jumping the gun.

This e-Levy thing needs its own separate policy treatment, outside of the tight budget process and timeline. Mr. Government, please take it out, work on it carefully, solicit stakeholder input, and return it to Parliament in the New Year.