A visual essay on measurement
The Road That Doesn't Count
Two expressways. The same foreign financier. Only one of them counts as foreign-funded, and the difference decides which governments get called opaque.
Road one · Uganda · public record
Follow the loan, find the project
The Entebbe-Kampala Expressway. For decades, this is what foreign-funded infrastructure looked like, and every definition still knows how to see it.
USD 350 million lent by the Export-Import Bank of China to the Government of Uganda, at 2 percent, repayable over about forty years. Uganda added USD 126 million of its own. Total: about USD 476 million, built by a Chinese state contractor, opened in 2018.
The money enters the treasury, joins the budget, and leaves a paper trail the whole way: a signed loan, a public debt record, budget lines, audit reports. Follow the loan and you find the project.
Ask any monitoring framework whether this is foreign-funded infrastructure and the answer is instant.
counts · foreign-fundedRoad two · unnamed, deliberately
The money that never touches the treasury
Now a second expressway, financed from the same country. Two billion dollars. I will not name it, because it stands for a pattern, not a scandal.
A foreign company finances the road, builds it, and will operate it for decades, earning its money back from tolls before handing it over. The state granted the land, approved the route, and gave the concession.
No loan to government. No entry in the public debt book. No budget line. The money reaches the same kind of public asset, directly.
Concessions, equity stakes, joint ventures, supplier's credit, guarantees: a growing share of foreign infrastructure finance now arrives this way. The clearest description of the shift came from a research team studying it in their own country: lending has moved from direct concessional loans to build-operate-transfer frameworks.
Your call
Is road two foreign-funded infrastructure?
Public asset. Foreign money. Two billion dollars. What would you say?
Most working definitions say no.
They look for a loan to a government, or aid administered through a budget. This project has neither. So a two billion dollar piece of public infrastructure, financed entirely from abroad, quietly drops out of the study before scoring begins.
Counted as foreign-funded: 1 of 2. Same financier country. Same public purpose. Different shape.
The mechanism
Definitions written for instruments go blind by design
An instrument-keyed definition does not fail randomly. It keeps seeing the finance that still flows the old way, and it cannot see the finance that moved.
The definition sees
Invisible to it
Both of our roads can come from the same country's finance. One counts, one vanishes. The problem is the shape of the money, not the flag on it. And a study that only sees one shape is not comparing countries. It is comparing instruments.
Distortion one · flatters the old shape
Grading the bank's website
Multilateral and bilateral lenders publish rich project pages: appraisals, costs, procurement notices, completion reports. Useful documents. But watch what happens when a scoring exercise credits them as government transparency.
The funder's project page
The implementing government's page
If any published information earns the credit, regardless of who published it, then the score measures the funder's publishing culture and calls it government openness. The government's own silence disappears inside its banker's diligence.
Distortion two · condemns the new shape
The same zero tells three different stories
Road two never held a tender: the financier and the builder are the same company. A scoresheet that expects a tender record writes a zero. But what does that zero mean?
The record exists, or should exist, and was not published. A signed contract on no portal. This is the thing transparency work is meant to catch.
No tender was ever held, so there is no tender record for anyone to publish. Scoring this zero as concealment measures the deal's structure and calls it secrecy.
A completion report for a project that finishes in three years. Nobody on earth has it. A fair method waits; an unfair one calls it opacity today.
One caution cuts the other way: a structure that skips the tender does not erase the public's questions. The expectation shifts to the concession award, the direct-award justification, the negotiated terms. The expected record moves. It does not vanish.
The repair
Three boundary rules fix the comparison
None of them changes a scoring scale. They change what enters it, what a missing record means, and who gets credit. Settle them before anyone scores, because they decide more than the scale ever will.
Define by purpose, stake, and control
Is it a public asset or a public service under a government-granted concession? Does any party hold a recorded financial interest, of any kind? Is that interest ultimately controlled from outside the country, once you follow the ownership chain past the local company? Yes to all three: it counts, whatever the instrument is called.
Ask why a record is missing before scoring it
Hidden scores zero. Never-existed is excluded, with the expectation shifted to the records the structure did generate. Not-yet-due waits. The denominator adjusts, so honest reporting stops looking worse than silence.
Credit the publisher, and give figures a passport
Classify information by who controls its publication, not where you found it. A funder's page proves the information exists; only the government's own publication is government disclosure. And every figure travels with its source, date, currency, and exchange rate, or it is not evidence.
The same two roads, fairly
Both roads in. Both asked the same questions.
Counts, as it always did
Now scored on what the government itself published: the loan, the budget lines, the audits. The funder's documents stay as evidence, not as a substitute.
in scopeCounts, at last
Public asset, foreign stake, foreign control: in. Scored on the records its structure actually generates: the concession award, the terms, the tolls, the handover.
in scopeDrivers pay tolls on both roads. On road one, the tolls help service a forty-year public loan. On road two, they repay a private financier directly. Different paper, same public. The questions a citizen is entitled to ask do not change with the financing structure. The definitions just have to catch up.
The point
Transparency rankings often measure the shape of the money and the publishing culture of financiers. Before you compare governments, make sure you are not comparing their bankers.
Provenance and sources
This essay draws general lessons from recent cross-regional work mapping how foreign-funded infrastructure is defined and measured. That work is not mine to publish, so no programmes, partners, or study countries are named here, and road two is an unnamed archetype of a build-operate-transfer concession. Road one is a matter of public record:
- Entebbe-Kampala Expressway financing: USD 350 million preferential buyer's credit from the Export-Import Bank of China at 2 percent, repayable over about 40 years, with about USD 126 million from the Government of Uganda; contractor China Communications Construction Company; commissioned June 2018. Sources: AidData project 14235, Uganda Ministry of Works.
- The shift of Chinese infrastructure finance from concessional lending toward concession and equity structures is widely documented in public research on Belt and Road era financing.
Figures are as reported by the sources above. Nothing in this essay scores, ranks, or characterises any country.