The question came from a training course.
A development agency was advertising a free course on designing public-private partnerships that are sustainable, responsible and aligned to the Sustainable Development Goals. It is a good course, run by serious people, and the material is better than most.
And reading the syllabus I found myself stuck on a question I could not answer: why this modality and not the programme?
Why should a road built under a thirty-year concession be held to sustainability, gender and environmental standards that the identical road, built by the same ministry with the same design, is not held to when it is procured conventionally?
Two answers are possible. Either partnerships are more likely than direct public provision to be socially irresponsible, environmentally damaging or blind to gender — and therefore need correcting. Or direct provision already has those attributes, and what is delegated to a private party does not.
Neither is ever stated. Neither, as far as I can find, has been examined.
First, the size of the thing
Before asking whether the asymmetry is justified, it is worth knowing how much of the problem it covers.
Using the IMF’s Investment and Capital Stock Dataset, cross-checked against the World Bank’s database of private participation in infrastructure, the modality carrying this entire architecture of sustainability obligations amounts to about 3% of public capital spending in the countries that report it.
That figure is a floor, not a point estimate, and the paper documents exactly how it is built and where it could be wrong. But no reasonable construction of it makes the share large.
Whatever the architecture achieves, it achieves on a small and shrinking slice.
Then, what the rules actually say
I took the rulebook that governs the appraisal of every project seeking public funds in seven jurisdictions — Chile, Peru, Colombia, Mexico, the United Kingdom, France and the European Union — and counted the same terms in each, against two chapters of the 2026 PPP Certification Guide covering the same ground.
Gender appears ninety times in the two PPP chapters. It appears zero times in the Chilean, Peruvian, Mexican, British, French and EU rulebooks, and three times in Colombia’s — where, on inspection, all three are a demographic data field. Colombia asks you to count beneficiaries by sex. No system asks you to assess a project’s differential impact on women.
At which point a good critic says: counting is superficial. One mention can be enough if it binds, and ninety can be rhetoric.
That objection is correct, and it is the most useful thing anyone said to me while writing this.
So the counts are not the argument
They are an index of where to read. Every passage in which any document touches gender or climate was then read and coded for normative force — is there an actor, a deontic verb, a procedure, and does non-compliance block anything — on a scale from absent through purposive and discretionary to required with a method.
And the PPP curriculum was put through the same test. Applying a strict standard to one side and none to the other would have been worthless.
Coded, two things changed, and both corrected me.
Nobody has a method for gender — including the PPP curriculum. Its 102 passages contain no procedure, no parameter, no threshold. The strongest is a duty for the legal framework to align with national gender policy. That is real content and more than anyone else offers, but it is not a method. So the gap is between abundant aspirational language and silence, not between rigour and silence — and the recommendation cannot be to transplant, because there is nothing there to transplant.
On climate, the ranking inverts. The three strongest treatments in the study are not in the PPP curriculum. They are public investment manuals: the UK Green Book, which carries an arithmetic procedure and an instruction to assess whenever in doubt; the French guide, which publishes shadow prices in euros per tonne of CO₂; and the EU Vademecum. An earlier draft of mine implied the opposite. The coding does not support it.
The test that changed the paper
The comfortable explanation is that national systems lag and international guidance is pulling them along. The way to test that is to hold the publisher fixed.
The World Bank publishes the PPP Reference Guide. It also publishes the Public Investment Management Reference Guide. The IMF publishes the PIMA Handbook. Same genre, same finance ministries, same years, same two institutions that jointly built the PPP certification. Between them, 526 pages on how a country should appraise and select all of its public investment.
Gender appears once in each, and both times it is rhetoric. The World Bank’s instance is a clause listing what a social impact assessment looks at. The IMF’s is in the opening case for why infrastructure matters. Neither asks anyone to do anything.
Climate, in the World Bank’s guide, appears four times and not once about climate: the fiscal climate, the business climate, the name of a Canadian agency, and a footnote in a table of other countries’ discount rates.
The IMF’s Handbook explains itself. One of its four uses of “climate change” is a footnote recording that the Fund is piloting a separate PIMA climate module. Climate did not enter the fifteen institutions of the assessment; it was built alongside them. Gender never got a module at all.
So the lag explanation does not survive. The same institutions, in the same years, for the same governments, wrote one body of guidance that names gender ninety times in two chapters and another that names it twice in five hundred pages.
This is not a gap between rich and poor practice. It is a gap between two product lines.
And it relocates the responsibility. A ministry that adopts both manuals as written will hold its concessionaires to a standard it never applies to its own works — not through hypocrisy, but because that is what the two books on its shelf say.
What good looks like
The United Kingdom is the case that breaks the pattern, and the most useful thing in the study.
A word count reports the Green Book as having zero mentions of gender, and is badly wrong about what it requires. The obligation is statutory rather than lexical: the Public Sector Equality Duty binds decision makers to the Equality Act, and the Green Book instructs practitioners to use distributional analysis to show effects on those with protected characteristics.
And beside it sits a method — the welfare weight of a group as a ratio of median incomes raised to a published power, with equivalisation scores per household member, and the hard requirement that weighted estimates be presented alongside unweighted ones so decision makers can see what the weighting did.
Zero mentions of gender, and more enforceable content than ninety.
The gap in the other systems is therefore not inevitable, and not a property of cost-benefit analysis. It is a choice about what the appraisal rules are asked to contain.
The same shape, three levels down
I set out to test one asymmetry and found it three times.
In the rulebooks, where obligations attach wherever there is a counterparty to bind. In the guidance that writes the rulebooks, confirmed independently by international lawyers reading loan agreements as treaties, who describe the Bank operating through financial leverage rather than formal hierarchy. And in the survey instruments that would be needed to measure any of it: care-related travel is not a category in conventional transport surveys, short walking trips are dropped, and the reason given is that they are not considered relevant to infrastructure policy.
What is not valued is not counted, and what is not counted cannot later be valued.
The recursion is what makes the problem stubborn, and it is also where the leverage is. A requirement written into an appraisal manual with a method attached breaks the loop at the first level. A survey category for care travel breaks it at the third, and costs almost nothing.
Neither requires anyone to agree that partnerships are held to too high a standard — which is the argument this began with, and the least important thing in it.
The full paper — the seven-jurisdiction table, the coding scale, the symmetric test in full, and a worked gender weight built on time-use data — is at austral-intelligence.com/research/ppp-vs-programme/.
Disclosure: I co-developed PFRAM, published jointly by the IMF and the World Bank, and led the World Bank’s work on PPP fiscal risk assessment. I am therefore a contributor to the body of PPP-specific guidance whose asymmetry this paper documents. I also founded Austral, which builds appraisal software, and the paper’s final section describes a module of it as a worked proof that the gap is closable — with the method set out in full so it can be reproduced without the software, and including a parameter error of mine that reversed an earlier conclusion. Readers should weigh those interests.
The Price of a PPP is published monthly, on the third Thursday — on where public infrastructure obligations get counted, where they don’t, and what that costs.