Insight, Industrial Policy
A new World Bank report finds that around 60 per cent of Africa's trade costs sit inside countries rather than at their borders. West Africa's development bank is putting US$2 billion into industrial-grade power. Both point at the same variable, and it is not price.
The World Bank has launched Integrating Africa: From Threads to Hubs, arguing that the next stage of the African Continental Free Trade Area is interoperability: customs, transport, payments, standards, energy, finance and digital systems that work across borders rather than only tariff schedules that permit trade across them. The report identifies four priorities, including strengthening regional value chains and reducing behind-the-border trade frictions.
One reported finding deserves to be pulled out and left on its own. Around 60 per cent of Africa's trade costs come from behind-the-border barriers rather than from border controls, with customs clearance inefficiencies, regulatory misalignment, weak logistics services, fragmented transit systems and poor infrastructure named as the sources.
Sit with what that means for a free trade area. Reducing tariffs is what a free trade area does. Tariff reduction acts on the smaller share of the cost, and the tariff schedules themselves are only part of even that share.
The larger part is inside the country, and it has a common character across every domain the report names.
Consider a component maker in one country supplying an assembler in another.
The border crossing takes two days on average. It also ranges from six hours to nine days depending on the shift, the paperwork, the inspector and the season. The assembler does not plan around two days. The assembler holds nine days of buffer inventory, because a line stoppage costs more than warehousing does.
The working capital tied up in that buffer, across every input, is the actual cost of the border, and it does not appear in the average. Averages are what get published. Variance is what determines whether anybody specialises.
The same structure governs electricity, and this is where the second announcement fits.
The ECOWAS Bank for Investment and Development has said it intends to invest a further US$2 billion in West African energy infrastructure by 2030, on top of roughly US$1 billion invested across generation, transmission and distribution by June 2026, framed explicitly around reliable power for manufacturing, agro-processing, mining, digital industries and urban development.
That framing is the important part. Household electrification and industrial-grade supply are different products, and the statistics that track the first say almost nothing about the second. Industry needs three-phase supply at stable voltage, no unplanned interruption during a process run, a tariff it can forecast across a debt schedule, and a connection date it can plan a factory around.
A firm can price expensive electricity. It cannot price unpredictable electricity, because unpredictability forces it to size and finance backup generation for the worst case and then run two systems. That doubling is the real cost, and it is invisible in a tariff comparison.
Two distributions of delivery time or hours of supply interruption can share exactly the same mean and impose very different costs. The firm sizes its buffer inventory and its backup generation against the tail, not the mean.
Only the average is usually published. The tail is what actually decides whether a firm depends on somebody else's supply.
The case for regional value chains is that a battery, a fertiliser plant or a food processing industry does not need every input made in one country. It needs specialised capabilities connected across borders.
But specialisation means depending on somebody else's reliability, and every link added multiplies exposure to variance. A firm that makes everything itself is inefficient and predictable. Under high variance that is the rational choice, and it is exactly the choice that prevents a hub from forming.
This is why interoperability is not a softer version of trade liberalisation. It acts on the variable that actually decides whether firms are willing to depend on each other, and it is slower, less announceable and harder to attribute than a tariff cut.
It also explains something otherwise puzzling about the continent. Firms integrate vertically, hold large inventories, generate their own power and handle their own logistics, and these look like symptoms of poor management. They are rational responses to distributions with long tails, and they will persist until the tails shorten, regardless of what the tariff schedule says.
Set alongside this, Indonesia has launched fourteen solar projects totalling around 5.3 gigawatts, within a programme targeting roughly 100 gigawatts by 2029 and potentially around US$62 billion of investment, explicitly linked to displacing around 13 gigawatts of diesel generation while building domestic solar-cell, battery and transmission industries.
That is an attempt to do inside one jurisdiction what the AfCFTA proposes across many. No customs posts, one currency, one standards regime, one regulator.
Which makes it unusually informative. If localisation succeeds there, the border was a real constraint elsewhere. If it does not, and the obstacle turns out to be the same set of behind-the-border problems in domestic form — weak supplier reliability, unpredictable permitting, slow payments and unstable grid quality — then the border was never the main thing. Anyone building a case for regional integration should be watching Indonesia closely for that reason, and the capability question will decide it.
There is an intervention here that costs almost nothing and is almost never done.
Publish the variance.
Not average border clearance time but the ninetieth percentile. Not electrification rates but outage minutes and voltage events per customer. Not an average connection timeline but the distribution of days from application to energisation for an industrial connection. Not average payment settlement but the tail.
These figures largely exist already, in customs systems, utility outage management systems and payment platforms. They are not published, and the consequence is that a firm deciding whether to depend on a supplier across a border, or to size a factory without a generator, has to assume the worst. A firm that can see the tail can price it and insure against it. A firm that cannot must plan as though the worst case is normal, and that assumption is itself a large part of the cost.
Transparency about your own bad numbers is an uncomfortable form of industrial policy. It is also, on this evidence, cheaper than most of the alternatives and faster than all of them.
The Lab works on this in local manufacturing and energy access, across the markets where industrial-grade reliability is the binding question.
If you are financing industrial infrastructure and want to know what firms actually experience rather than what the average reports, tell us what you need to know.
This is an independent insight piece by Transitions Lab. For the Lab's applied work, see Local Manufacturing & Supply Chains and Energy Access & Off-Grid Systems. See also The Customers Who Can Leave on the same infrastructure being funded from the public purse while the most creditworthy users step off it, and The Village Twenty Kilometres Off the Road on why variance rather than distance is what stops firms trading across a corridor. To discuss a study, see Contact.