What We Look For

01 Underwriting, in the open

Most investors keep their underwriting private, treating the criteria as an edge. The coalition publishes its guidelines for two reasons that outweigh any edge. First, efficiency: a sponsor who can read the bar self-selects, and the pipeline that arrives is already half-screened. Second, integrity: standards that live in a drawer bend under pressure; standards on a public page bend in public, which is to say rarely.

What follows is the actual sequence a proposal travels — not a sanitised summary. It is the same document, in substance, that the Investment Committee scores against and that Risk & Compliance audits. Where numbers are ranges, the ranges are real; where a rule says “no exceptions,” the waiver register — also public — shows none.

One caution for readers building a submission: the guidelines describe how proposals are judged, not how to phrase them. A proposal engineered to recite the four tests back at the committee scores worse than one that simply answers them, because the committee has read every version of the recitation before.

02 Screen one — eligibility

Before any analysis, a proposal must clear four threshold questions. The screen is deliberately cheap to run — a day, not a month — because its purpose is to protect diligence capacity for proposals that can actually proceed.

Two of these deserve elaboration. Nexus is not a residency check; it asks whether the member’s own authorities have named this priority, because the coalition does not import agendas into member economies. And the scale floor is not elitism — below roughly ten million dollars, the fixed cost of proper diligence and a real capacity component consumes the development value of the deal. Smaller needs are served through intermediated structures: a guarantee pool or fund commitment that reaches small tickets through a local institution built for them.

• Nexus — does it serve a member economy, originated with or endorsed by its authorities?
• Sector — does it sit inside one of the ten verticals, or genuinely across their seams?
• Scale — is it large enough to carry a capacity component and diligence cost? (Indicative floor: $10M)
• Counterparty — is the sponsor identifiable, solvent, and willing to accept coalition standards?

03 The four tests

Everything that passes eligibility is underwritten against the same four tests, each scored one to five and minuted. A proposal needs no perfect scores, but a one on any test is a decline, and the scores travel with the proposal to committee — origination cannot soften them en route.

Additionality is the test most often failed and most often misunderstood. It does not ask whether the project is good; it asks whether the coalition is necessary. A sound toll road that three commercial banks would finance tomorrow fails additionality however excellent its economics — funding it would simply displace private capital while wearing a development label. The test forces the coalition to live at the frontier of what markets will not yet do, which is the only place concessional capital earns its concession.

Durability is the counterweight that keeps additionality honest. It is easy to be additional by funding things nobody sensible would fund; durability asks what still stands in ten years. The two tests together define the target zone: deals markets refuse today and will embrace tomorrow, because the coalition’s participation builds the record, the institution, or the framework that changes the price.

• Additionality — would this happen anyway without the coalition? If yes, decline
• Mobilisation — how much private and institutional capital does the structure crowd in?
• Durability — what institution, skill, or asset survives after the financing exits?
• Integrity — does the counterparty and structure clear anti-corruption, sanctions, and safeguard screens?

04 Credit and risk criteria

The financial underwriting is deliberately conventional — the innovation belongs in the structuring, never in wishful analysis. Debt commitments are sized against demonstrated or contracted cash flows, stressed for currency, commodity, and demand shocks at severities drawn from each economy’s own history rather than a global average. Equity is sized against governance rights, not just valuation: the coalition does not take minority positions without information rights and reserved matters, because capital without visibility is hope, not investment.

Guarantees — the instrument most tempting to misprice — are priced from expected loss with a capital charge, never issued as free comfort. Every guarantee carries a published trigger, a defined tenor, and a reserve held against the modelled book. Currency risk follows one principle across all instruments: it must land on the balance sheet best able to carry it, which is usually not the member government’s. Where no natural holder exists, the coalition’s local-currency solutions or pooled cover absorb it explicitly and price it visibly.

Counterparty analysis goes one layer deeper than convention: past the sponsor to the operator, and past the operator to the institution that will own the asset at exit. A strong sponsor with a weak eventual owner scores poorly on durability no matter how clean the construction-phase credit looks — this is where the capacity plan enters the credit file, as underwriting rather than decoration.

05 Safeguard screens

Environmental and social review scales with risk category, assigned at eligibility and re-confirmed at diligence. Category A — high-risk — commitments require an independent assessment before committee, disclosed in affected communities’ languages, with a functioning grievance channel as a condition of first disbursement.

Three findings are deal-breakers rather than negotiating points: involuntary resettlement without a compliant framework, conversion of critical habitat, and child or forced labour anywhere in the direct supply chain. Safeguard conditions enter the legal documentation with the same force as financial covenants and are monitored through delivery — a commitment can be suspended for safeguard breach exactly as it can for payment default, and the register of suspensions is disclosed.

06 Portfolio limits

No single commitment, however good, may bend the book. Concentration limits cap exposure by economy, sector, instrument, and counterparty group; they are set by the Investment Committee, monitored continuously by Risk & Compliance, and published in the annual risk disclosures.

When a proposal would breach a limit, the answer is a smaller ticket or a syndication — never an exception, because the first exception is a precedent and the third is a policy. The limits exist for a reason the portfolio has already validated once: in the coalition’s second year, discipline on a single-economy cap kept a regional payment crisis from touching more than four percent of the book.

07 Pricing principles

Concessionality is a tool, not a giveaway. Pricing sits at the minimum concession needed to make the transaction happen and crowd others in — priced too soft, the coalition substitutes for markets and fails additionality retroactively; priced too hard, nothing moves and the mandate goes unserved.

The discipline is traceability: every basis point of concession must map to the development return that justifies it, in the pricing memo the committee sees. Where a commitment can carry market terms, it does; where it cannot, the memo says exactly why, and the gap between coalition terms and market terms is recorded as the measured subsidy — a number the results framework then holds against the outcomes.

08 The deal-breakers

Some findings end the conversation regardless of returns, and publishing the list is a courtesy to counterparties: do not spend a quarter preparing a proposal that dies on page one.

The integrity screen deserves one elaboration: beneficial-ownership opacity is itself the failure. The coalition does not need to prove wrongdoing behind an opaque structure; the opacity is disqualifying, because capital that cannot see its counterparty cannot answer for it. Zero waivers of the integrity screen have been granted, and the number is published each year precisely so that it stays zero.

• Integrity — credible corruption findings, sanctions exposure, or beneficial-ownership opacity
• No additionality — a market or another institution would fund it on similar terms
• Capacity vacuum — no credible institution to own the asset after exit, and no plan to build one
• Safeguard red lines — the exclusions above, without exception
• Standards refusal — a counterparty unwilling to accept disclosure and independent evaluation

09 The decision path

Origination with the member authority; independent review by Risk & Compliance, staffed separately from the team that sourced the deal; Investment Committee recommendation against the four tests and portfolio limits; Board approval, minuted and published; then independent evaluation of outcomes, reported above management. Five gates, same order, no shortcuts — whatever the size, whoever the sponsor.

The funnel data below shows what the gates do in practice: of every hundred structured inquiries, roughly forty-four clear eligibility, twenty-seven enter full diligence, sixteen reach committee, and eleven are committed. The attrition is the system working. A funnel that passed everything would mean the screens were decorative; one that passed nothing would mean the mandate was.

10 What a complete submission contains

Sponsors ask what the diligence file should hold; the honest answer fits in a paragraph. Corporate: legal identity, beneficial ownership to natural persons, three years of audited statements. Transaction: the ask, the use of proceeds, the cash-flow model with assumptions exposed, the security package. Delivery: who operates it, their track record, the capacity gap analysis if there is one. Alignment: the member-authority endorsement, and the development case in the sponsor’s own numbers. Nothing exotic — the file is judged on candour, not thickness.

The most useful advice is the least intuitive: state the weaknesses yourself. Diligence will find them regardless, and a submission that names its own risks with mitigants signals a counterparty worth partnering with. The files that stall are rarely the ones with problems; they are the ones where problems were dressed.

11 Timelines, and what to expect

From complete submission: eligibility answer within fifteen working days; diligence, where opened, typically eight to sixteen weeks depending on category and geography; committee scheduling monthly; Board approval on the same cycle. Financial close then depends mostly on the counterparty’s side of the documentation. The published medians — six months from committee to close under common terms — are floors the operations report audits annually.

Sponsors can shorten their own path materially: beneficial ownership documented to natural persons before submission, the member-authority endorsement secured early rather than retrofitted, and the capacity plan drafted with the operating institution rather than about it. The pipeline’s fastest transactions share those three features almost without exception.

12 After a decline

A decline arrives with its reason — the failed test or screen, in writing. Roughly a third of declined sponsors return within two years with the defect cured, and cured resubmissions are welcome: the guidelines are a bar, not a blacklist. Additionality failures are the exception that proves the design — a sponsor declined because the market would fund them has been given good news, and the decline letter says so.

What a decline never includes is an invitation to renegotiate the standards. The tests, screens, and limits move through the annual framework review, on portfolio evidence — never inside a live transaction, however large. Sponsors who have watched other institutions bend under deal pressure eventually recognise this as the feature it is: the same immovability protecting the portfolio from someone else’s bad deal is protecting theirs.

13 The annual framework review

The guidelines on this page are versioned, and the version changes exactly once a year. Every screen, test, threshold, and deal-breaker enters the annual framework review, argued from portfolio evidence: decline statistics, delivery outcomes, claims experience, and the waiver register. Proposals to amend can come from any region, committee, or the evaluation panel — but never from inside a live transaction, which is the firewall that keeps the standards from being negotiated by whoever has the largest deal pending.

The review publishes its decisions with reasoning, including the rejections. Recent cycles have tightened beneficial-ownership documentation, added the named-role continuity requirement to capacity plans after churn data demanded it, and — the review’s most cited decision — declined to raise the eligibility floor despite diligence-cost pressure, on the evidence that intermediated structures were serving small tickets better than direct lending ever had. The guidelines you have just read are, in other words, the current output of a running argument with the portfolio’s own data — which is the only authority underwriting standards should answer to.