Developed countries reported providing US$115.9 billion of climate finance in 2022. Oxfam and CARE, working from the same underlying reporting, put the real value at US$28–35 billion. Neither figure is a mistake, and neither side is accused of inventing data. They are answers to two different questions, and the gap between them is almost entirely a question of instrument.

A loan is reported at face value. Ten million dollars lent counts identically to ten million dollars given, even though one is repaid with interest and one is not. Compute the grant equivalent instead, netting out repayment and interest, and the headline falls by roughly two thirds.

The division is deliberate. A practitioner who understands the Union Budget still cannot read a concession agreement, and the two skills are taught by different literatures.

Three audiences, with different reasons to be here. People working in organisations affected by a financed project, who need to know what documents exist and who is obliged to produce them. People designing or evaluating programmes funded through these channels, who need to read the conditions attached. And researchers and students who need the institutional vocabulary before the critical literature makes sense.

Two countries each receive a headline US$100 million. One receives a grant. The other receives a twenty-year loan at a market rate. The press releases look the same; the fiscal consequences do not.

The first question locates the guarantee, the offtake agreement or the revenue floor. The second separates a grant from a loan and a loan from a contingent liability. The third is the one most reporting omits: a financing package is negotiated, and the concessions given in exchange — a tariff formula, a procurement rule, a regulatory commitment — are part of its price.

Each shift redefined what the money was for, and each left institutional residue behind. The safeguard policies and complaint mechanisms of module nine are residue from the project-lending era. The conditionality literature is residue from adjustment. Both still operate inside an architecture now organised around the fourth row.

This is why a course written twenty years ago would be a poor guide today. The critical vocabulary of conditionality still applies, but it no longer describes where most of the decisions are made.

• A finance figure is a construct. Ask what it counts before comparing it to anything.
• Concessionality, not headline size, determines what a loan does to a country.
• A guarantee is real money with an invisible balance sheet entry.
• The system's centre of gravity has moved from lending to states towards making projects investible for private capital.
• The documents that answer the hard questions usually exist, and are more often public than assumed.

One caution before going further. None of the five points implies that development finance is a bad thing or that the institutions act in bad faith. The argument of this course is narrower and more useful than that: the form money takes has consequences that its headline size conceals, and those consequences fall on people who were not party to the negotiation. A practitioner who can read the form is in a position to say what those consequences are, specifically, with a document behind the claim. A practitioner who cannot is left arguing about intentions, which is an argument nobody wins.

The distinction that matters most for accountability is the fourth row. When an MDB lends through its private arm, the borrower is a company, the safeguard regime is different, and the complaint mechanism is a different body.

At the IBRD, voting power is tied to capital subscription. A member's share of the votes reflects its share of the capital, adjusted by a small allocation of basic votes given equally to all members. The distribution is published, and the Articles of Agreement set the framework.

Read this beside a safeguard policy requiring free, prior and informed consultation with indigenous peoples, or a governance loan conditioned on procurement reform, and the tension is immediate. A great deal of the institutional and legal literature on the MDBs is an attempt to reconcile the political-prohibition clause with lending practice that is unavoidably political.

Graduation from IDA to IBRD is therefore a large event in a country's public finances, and the eligibility thresholds are a live political question. India's own transition out of IDA eligibility is a useful worked case for anyone studying how the terms change when a country crosses a line.

The standard institutional account is that the conditionality of the structural adjustment era was substantially reformed, with fewer conditions and more attention to social protection. Kentikelenis, Stubbs and King coded the actual conditions attached to IMF programmes across three decades and found the reform is largely presentational.

• The count of conditions changed more than their substantive reach
• Conditions continued to extend into labour markets, public employment and social policy
• Stated attention to social protection appeared more in framing than in binding requirements
• The coding scheme is published, which is what makes the finding disputable rather than rhetorical

For a practitioner in an advocacy organisation this is the single most portable technique in the module: if you disagree with an institution's account of its own behaviour, code its documents.

Woods's argument in The Globalizers is that outcomes are produced by the interaction of three things, not by any one of them: the professional norms of staff, the voting arithmetic of the board, and the domestic coalitions inside the borrowing country. The same conditions imposed on two countries produce different results because the third factor differs.

The reverse error is just as common and less discussed. Where a domestic government genuinely wants a reform that its own coalition blocks, an external condition can supply political cover: the reform proceeds and the blame is exported to the lender. This is one reason conditionality persists despite poor evidence of its effectiveness on paper. It is doing work the published rationale does not describe, and a study that measures only the stated objective will find it failing while the actual function succeeds.

• Lenders differ by ownership, borrower and instrument, and the private arms differ most
• Voting is weighted by subscription, and the distribution is published
• The political-prohibition clause explains why safeguards read as technical standards
• IDA and IBRD terms differ enough that graduation is a fiscal event
• Conditionality reform is measurably more presentational than substantive
• Outcomes come from staff, board and domestic coalition together

The thread connecting these is that the institutional detail is not decoration around the politics; it is where the politics happens. Voting weights determine whose preferences survive a contested policy revision. The political-prohibition clause determines the register in which a social standard must be written to be adoptable at all. And the difference between IDA and IBRD terms determines whether a ministry of finance can afford a programme that a line ministry wants. Someone campaigning on any of these without knowing which body decides what will aim at the wrong target, and the institutions are under no obligation to correct them.

Most confusion in public debate comes from treating the first and the fifth rows as the same thing.

A long grace period makes the early years of a loan look painless, which is precisely the period a political cycle covers. The repayment burden arrives later, frequently after the government that signed has left office.

The test to apply: plot the debt service year by year over the full maturity and ask which administration faces the peak.

The same logic applies to concession length. A thirty-year concession signed today allocates revenue for a period longer than most of the officials signing it will remain in post, and longer than any demand forecast can responsibly project. Long tenors are frequently necessary, because the asset lasts that long and the capital has to be recovered over its life. The point is that length is not a neutral technical parameter: it determines how far into the future present decisions bind, and who will be around to be answerable for them.

The question to ask of any guarantee is not whether it will be called, but what has to happen for it to be called, and how correlated that event is with everything else going wrong at the same time.

Policy-based lending deserves particular care because its disbursement trigger is a policy action rather than a physical output. That makes verification political rather than technical: whether a tariff order was issued, whether a law was notified, whether an agency was established. Each can be satisfied formally while leaving the intended substance untouched, and the incentive to satisfy it formally is created by the disbursement itself. Read the condition to see whether it specifies an outcome or merely an instrument, because the two behave very differently once money depends on them.

The grant element of a loan is the difference between its face value and the present value of its future repayments, expressed as a percentage of face value. The discount rate chosen changes the answer, which is why the choice of discount rate is itself contested in the reporting standards.

• A grant has a grant element of 100% by definition
• A loan at the discount rate has a grant element of zero
• Longer maturity and longer grace raise the grant element
• A higher discount rate raises the measured concessionality of any given loan
• Reporting standards that permit a generous discount rate therefore inflate reported concessional finance

Debt sustainability analysis asks whether a country can service its obligations without an implausible fiscal adjustment. It is a projection exercise, and its conclusions are sensitive to assumptions about growth, exchange rates and interest rates over long horizons.

• The instrument determines who bears risk; the headline figure does not
• Six terms decide a loan's real burden, and currency is the quietest of them
• Guarantees are unpriced public risk with a delayed and correlated cost
• Results-based lending makes measurement quality a financing question
• Grant element depends on a chosen discount rate, so the standard is contestable
• Most of the documents you need exist; the concession agreement is the hard one

If you retain one operational habit from this module, make it the currency question. Interest rates are negotiated in public and scrutinised; currency denomination is treated as a technical detail and settled quietly. A project earning rupees and servicing dollars has an unhedged exposure that no clause in the loan describes as a risk, and a movement of twenty per cent in the exchange rate can convert a comfortable debt service into an impossible one without a single term of the agreement changing. Ask what currency the revenue is in, ask what currency the debt is in, and if they differ, ask who absorbs the gap.

Ahead of the 2015 Addis Ababa conference on financing for development, the multilateral development banks and the IMF published a joint paper arguing that the sums implied by the Sustainable Development Goals were far beyond what aid budgets could supply, and that the gap should be closed by using public money to mobilise private capital at a multiple.

The World Bank's Maximizing Finance for Development approach sets this as a decision sequence for project appraisal. Read as drafting, the significant feature is where the burden of proof sits: public financing is the residual, to be justified after the commercial and regulatory-reform options have been exhausted.

Whether that is prudent stewardship of scarce public money or a structural bias against public provision is the argument. The sequence itself is not hidden; it is published policy.

Daniela Gabor's argument is that Billions to Trillions, Maximizing Finance for Development and the G20's Infrastructure as an Asset Class agenda are one project: reorganising development around escorting institutional investors into a new asset class.

If public money is justified by the private capital it mobilises, then the ratio is the test. The OECD publishes an annual series on amounts mobilised from the private sector by official development finance interventions, broken down by instrument, sector and recipient income group.

• Guarantees and syndicated loans account for a large share of reported mobilisation
• Instrument mix matters: some instruments mobilise far more per public dollar than others
• The distribution across income groups is the politically significant cut
• The attribution method — who gets credit for a given private dollar — is itself a standard, not a fact

Blended finance mixes concessional public money with commercial capital in a single structure, typically with the public tranche taking first-loss or subordinate position. The OECD DAC principles set standards: anchor to a development rationale, design to attract commercial finance, tailor to local context, manage for results, and monitor transparently.

• The shift from lending to mobilising is documented in the institutions' own published policy
• The Cascade places public financing last in the appraisal sequence
• De-risking transfers specified risks to the public and constrains later policy by contract
• Mobilisation ratios are the testable core of the agenda, and OECD publishes the data
• Additionality is the hardest and most important question, and is rarely answered
• The mechanism performs worst where need is greatest, which is structural

The fair summary of this debate is that both sides are describing the same mechanism accurately and disagreeing about the counterfactual. Supporters are right that aid budgets cannot fund the estimated need and that private capital will not move without risk mitigation. Critics are right that the mitigation transfers real risk to the public and contracts away policy room. What would settle it is evidence on additionality, and that evidence is thin because the counterfactual is unobservable and nobody with a stake in the answer is well placed to produce it. Treat confident claims in either direction with the scepticism the evidence base deserves.

In corporate lending, a bank lends to a company and looks to that company's whole balance sheet for repayment. In project finance, a lender lends to a single project, usually housed in a special purpose vehicle, and looks primarily to that project's own cash flows.

The single most consequential row for public policy is the last but one. A change-in-law or stabilisation clause can require the state to compensate the concessionaire when it changes policy, which is the contractual form of the policy-space argument in module four.

Every project has a financial model: a spreadsheet projecting capital cost, revenue, operating cost, debt service and equity return over the concession period. The negotiated terms are whatever makes that model produce an acceptable return.

• The equity internal rate of return is the number the sponsor is solving for
• Tariff, concession length and guarantee structure are the levers that move it
• A small change in assumed demand can change the required tariff substantially
• The model's assumptions are therefore the substance of the public interest question

Optimistic traffic and demand forecasts are the most documented pathology in infrastructure appraisal internationally. Forecasts made by parties with an interest in the project proceeding tend to be higher than outturn.

A large share of long-concession infrastructure contracts are renegotiated. Concessions run for decades; no party can foresee three decades of demand, policy and cost. The Kelkar Committee's central insight was that renegotiation should be designed for rather than treated as a scandal.

• If renegotiation is unplanned, it happens under duress and asymmetric information
• The party threatening to walk away has leverage the public interest does not
• A pre-agreed renegotiation framework converts a crisis into a procedure
• Without one, the state's choice is between a bad deal and a stalled asset

The appraisal document tells you what the project is for and what risks were identified. The concession agreement tells you who carries them. The environmental and social assessment tells you who lives there. The procurement award tells you who is building it and for how much. The implementation report, published years later, tells you what actually happened.

• Project finance lends against a project's cash flows, so every risk is allocated by contract
• The legal counterparty may be a thin SPV rather than the recognised sponsor
• Availability payments and take-or-pay both move demand risk to the public side
• Change-in-law and stabilisation clauses are where policy space is contracted away
• The financial model's assumptions are the substance of the public interest question
• Renegotiation is normal, and is best handled by a framework agreed in advance

The practical lesson of this module is that a concession is not a procurement with a longer timescale. A procurement buys a thing; a concession allocates an uncertain future between parties with different appetites for risk and very different information about it. That is why the document runs to hundreds of pages and why the schedules matter more than the recitals. When you are handed a summary of a concession, what you have been handed is the part the drafter was comfortable summarising. The risk allocation table is the document.

NaBFID is the fifth All India Financial Institution, after EXIM Bank, NABARD, NHB and SIDBI.

Read the table as a sequence rather than a list. India dismantled its development financial institutions in the 1990s and 2000s on the reasoning that banks and capital markets would supply long-term finance more efficiently. Banks then lent heavily to infrastructure, funded by short-term deposits, and a large share of that lending became stressed. NaBFID in 2021 is a reversal of the 1990s judgement, made in the light of what followed it. Whether the second attempt works depends on whether the bond market that the first attempt lacked now exists.

India spent the 1990s and 2000s winding down the development financial institution model, on the view that term lending was better done by banks and markets. NaBFID reverses that judgement for infrastructure specifically.

Judge the institution on both mandates. Lending volume alone would miss half of what it was created to do.

The National Monetisation Pipeline, developed by NITI Aayog on a Union Budget 2021-22 mandate, identified an aggregate monetisation potential of Rs 6 lakh crore across roads, railways, power, gas pipelines, telecom and civil aviation, over FY2021-22 to FY2024-25.

• The framing is explicit: structured contractual partnership, not privatisation or slump sale
• Ownership is retained; the right to operate and collect revenue is transferred for a period
• The state receives capital up front, to recycle into new construction
• Government reporting put realised monetisation at about Rs 3.85 lakh crore over the first three years

Viability gap funding is a capital grant to a PPP project that is economically desirable but not commercially viable at a tariff users can bear. The state contributes a share of capital cost so that the project clears the investor's return threshold.

The Kelkar Committee reported to the Finance Minister on 19 November 2015; the report was released publicly on 28 December 2015. Its diagnosis was that PPP contracts had been drafted around fiscal transfer rather than service delivery, and that disputes had nowhere sensible to go.

• Recommended an Infrastructure PPP Project Review Committee
• Recommended an Infrastructure PPP Adjudication Tribunal, headed by a former Supreme Court or High Court judge
• Recommended model concession agreements be reviewed sector by sector
• Recommended a national PPP policy document from the Ministry of Finance
• Recommended an inbuilt renegotiation mechanism rather than ad hoc renegotiation

Comparing a Bill as introduced with the Act as assented is the fastest way to see which safeguards survived committee, and PRS makes that comparison straightforward.

Two of these sources are systematically underused. Parliamentary and assembly answers are given under an obligation of accuracy and are often the only place a specific figure appears in public, and they are searchable by subject. CAG reports are slow, arriving years after the events they examine, but they are produced by an authority with statutory access to records that no outside researcher can obtain, and their findings carry weight in forums where an advocacy report would not. Both reward the patience of searching them properly.

• NaBFID revives the DFI model for infrastructure, with Rs 1,00,000 crore authorised capital and a market-development mandate
• Its permitted shareholder list runs from the Central Government to pension funds and sovereign wealth funds
• The Monetisation Pipeline trades future revenue for present cash; the discount rate is the crux
• Viability gap funding is defensible when competitively bid and not stacked on other guarantees
• Kelkar's recommendations are a checklist to trace against actual implementation
• Indian project information is scattered but substantially public

Two threads run through this module and both are worth stating plainly. The first is that India has now tried, at different times, most of the available models: budget-funded public works, PPP concessions with viability gap support, a dedicated development financial institution, and the sale of future revenue from assets already built. Each was adopted partly because the previous one disappointed, which should make anyone cautious about the current one's permanence. The second is that the disappointments were rarely about the model in the abstract; they were about forecasting, risk allocation and renegotiation, which are execution questions that follow whichever model is chosen.

Every financed infrastructure project occupies land, and in India two statutes principally govern what it owes the people there: the Right to Fair Compensation and Transparency in Land Acquisition, Rehabilitation and Resettlement Act, 2013, and the Environment Impact Assessment Notification, 2006, issued under the Environment (Protection) Act, 1986.

The practical shift is procedural as much as substantive: the Act creates steps that must be documented, and documentation is what makes a process reviewable.

One structural feature deserves emphasis because it is frequently missed. The 2013 Act extends entitlements beyond people who hold title to the land. Families whose livelihood depends on the area, including agricultural labourers, artisans and those dependent on common resources, fall within its definition of affected families. In practice this is the provision most often under-implemented, because identifying the landless requires survey work that acquiring authorities have an incentive to do narrowly, and because the people affected are least likely to have documentation of their own.

• The assessment identifies affected families, not just landowners, which matters for the landless
• It is required to consider whether the acquisition is the minimum necessary
• The public hearing produces minutes, which are a record of objections
• Entitlements attach to categories defined in the Act's schedules

Category A projects are appraised centrally; category B at state level. The screening decision therefore determines the forum, and the forum affects both scrutiny and the route for challenge.

Two practical notes on obtaining these documents. Draft EIA reports are published before the public hearing precisely so that objections can be informed, which means the window to read and respond is narrow and known in advance rather than discretionary. And the minutes of the hearing are a formal record: objections made there are on file whether or not they are accepted, and a later challenge can point to the fact that a specific risk was raised at the time and how the appraisal dealt with it.

An environmental clearance is granted subject to conditions. Those conditions are not advisory. They are the terms on which the project was permitted, and compliance with them is reportable and challengeable.

The 2013 Act's consent requirements apply to defined categories of acquisition and are expressed as thresholds of affected families. Consent in this statutory sense is a procedural requirement with a numerical threshold. It is not the same as the free, prior and informed consent standard used in international instruments and in some lenders' safeguard policies.

Note also what consent attaches to. A consent threshold is generally measured against affected families at a defined point in time, which means the composition of the affected group is itself a contested question, decided by the survey that precedes the vote. Who is counted as affected determines both the denominator and who has a say, and disputes about a project's consent are very often disputes about that survey rather than about the vote it produced.

Both statutory regimes contain exemptions, exclusions and category boundaries, and in practice a large share of dispute concerns whether a project falls inside or outside them rather than whether its substantive obligations were met.

• Whether an acquisition falls under a state-specific or sector-specific regime
• Whether a project is category A or category B under the EIA Notification
• Whether an expansion requires fresh appraisal or rides on an earlier clearance
• Whether urgency provisions were properly invoked
• Whether a project was split so that each part falls below a threshold

• Two Indian statutes generate the public documents describing a project's social and environmental cost
• The 2013 Act's entitlements are a statutory floor a contract cannot go below
• The EIA process produces minutes, objections and enforceable clearance conditions
• Compliance with a promoter's own clearance conditions is usually the most tractable line of challenge
• Statutory consent and international free, prior and informed consent are different standards
• Category and threshold disputes, including project splitting, are where the action is

It is worth being precise about why a finance course spends a module on two statutes. It is not because a finance practitioner needs to litigate. It is because these statutes are the only mechanism that compels a project to produce a public description of its own social and environmental cost, in a form that can be checked against what is happening on the ground. Without the social impact assessment and the environmental clearance, the public record of a large project consists of a press release and a contractor's board. With them there is a document trail that names affected people, sets entitlements, and records commitments the project made in order to be permitted at all.

Module one opened with the gap between US$115.9 billion reported and US$28–35 billion assessed. This module explains how a gap that large is produced without anyone falsifying a number.

Every row is a defensible convention that someone argues for. Each one moves the total in the same direction, which is the pattern worth noticing.

There is a third question the two above tend to crowd out: predictability. A commitment that arrives in unpredictable amounts at unpredictable times cannot be planned against, and a ministry cannot build a multi-year adaptation programme on a flow it cannot forecast. Predictability rarely features in headline reporting because it is harder to express as a single figure, but for the officials who have to spend the money it frequently matters more than the total does.

Mitigation projects frequently generate a revenue stream: a solar plant sells power. Adaptation frequently does not: a sea wall, an early warning system, a drought-resistant cropping programme produce avoided losses rather than income.

• Private capital requires a return, and avoided loss is not a return to the investor
• So mitigation attracts blended and mobilised finance more readily
• Adaptation therefore depends disproportionately on grants and public budgets
• Which are precisely the sources under pressure
• The UNEP Adaptation Gap Report series quantifies the resulting shortfall

For a South Asian practitioner the category matters concretely rather than semantically. Slow-onset losses, such as land becoming saline or a glacier-fed river changing regime, sit awkwardly in every existing funding window: they are not a discrete disaster that triggers humanitarian finance, and they are not something adaptation spending can prevent. How eligibility is drawn will determine whether these losses are fundable at all, which is why the drafting of eligibility criteria deserves more attention than the headline pledge figures usually get.

Reading the OECD series and the shadow report side by side, on the same year, is the single best exercise in this module. The disagreement is transparent and the methods are published.

Read these in a deliberate order. Start with the treaty body's assessment, because it sets out the definitional problems most candidly and will make the other sources legible. Then take the OECD series for the official totals and their disaggregation. Then the shadow report, which recalculates the same underlying data and shows you where the two methods part company. Reading them in the reverse order tends to leave a reader with a conclusion and no sense of how it was reached.

Used loosely the word means little. Used precisely it names specific, checkable practices.

• Relabelling: existing development finance reported as climate finance without change in activity
• Over-attribution: counting a project's full value when only a component is climate-related
• Double counting: the same flow reported by more than one party
• Instrument inflation: reporting loans at face value as though they were grants
• Definitional drift: widening what qualifies, so totals rise without flows changing

India's updated Nationally Determined Contribution and its Long-Term Low-Carbon Development Strategy, both submitted in 2022, set out the country's targets and its assessment of what the transition requires. For this course the financing paragraphs matter more than the targets.

• Year: reporting lags two to three years, so a current-sounding figure is usually old
• Instrument: face value or grant equivalent changes the answer by a factor of three
• Source: public, mobilised private, or both added together
• Method: which attribution and qualification rules were applied

Apply the checklist to a friendly source as well as a hostile one. The discipline is worth little if it is used only to attack figures you already doubt, and a claim from an organisation whose conclusions you share is exactly the one you are least likely to interrogate. If your own side's number fails the same four questions, you need to know that before an opponent tells you, because the cost of discovering it in public is much higher than the cost of checking.

• A large reported-versus-assessed gap is produced by conventions, not by falsification
• Every convention in common use moves the total in the same direction
• Adequacy and direction are separate questions, and direction carries the distributional consequences
• Adaptation is structurally underfunded because it generates no revenue stream
• Loss and damage is contested because it implies responsibility rather than investment
• Greenwashing is checkable once you name which specific practice is alleged

A closing caution against the easy conclusion. The accounting problems documented here are real and they are not, by themselves, evidence that climate finance is a fiction or that the institutions reporting it are dishonest. Reporting conventions are genuinely difficult: reasonable people disagree about whether a loan at below-market rates delivers value equal to its face value, its grant equivalent, or something between. What is fair to say, and what the evidence supports, is that every convention in wide use happens to raise the reported total, that the bodies choosing the conventions are the bodies being assessed by them, and that the resulting numbers should therefore be read as negotiated figures rather than as measurements.

The World Bank's Inspection Panel was created by the Board in September 1993 and began operating on 1 August 1994. It was the first independent accountability mechanism at any international financial institution, and its innovation has since been copied at more than twenty other development banks and bilateral institutions.

• It provides a route for people who believe a Bank-financed operation may harm them
• It reports to the Board, not to management, which is the source of its independence
• It assesses compliance with the Bank's own policies and procedures
• It cannot award compensation; it can find non-compliance and trigger a response

The first task in any complaint is identifying which institution financed the specific component causing harm, because that determines which mechanism has jurisdiction and which safeguard policy sets the standard.

A complication worth anticipating: a large project often has several financiers, and they need not share a standard. A transmission line might take senior debt from one development bank, a partial risk guarantee from another, equity from a bilateral fund and commercial debt from domestic banks. Each financier owes its own policies, each mechanism has its own jurisdiction, and a harm caused by the project as a whole may not map neatly onto any one of them. Establish early which institution financed the component that produced the harm, because that is the first question the mechanism will ask.

• Financier: from the appraisal document, procurement notices, or the project's own signage
• Policy: the safeguard framework in force when the project was approved, not today's
• Harm and link: the requirement is harm plausibly connected to the breach
• File: mechanisms have eligibility criteria and time limits; read them first
• Follow-up: the action plan and its monitoring is where outcomes are won or lost

The gap between a mechanism as designed and a mechanism as reached is the single most useful thing to understand about this architecture, and it is not primarily a legal gap.

Every mechanism publishes its cases. Reading two or three end to end teaches more than any summary of the procedure, because a case file shows what the mechanism actually treats as sufficient.

• What the requesters alleged, in their own framing
• Which policies the mechanism considered in scope
• What management said in response
• What the investigation found, and on what evidence
• What the action plan committed to, and what monitoring followed

Read at least one case that the mechanism declined to register, alongside the ones it investigated. The eligibility decisions are where the boundaries of the system are actually drawn, and they are more informative than the substantive findings about what the mechanism will and will not treat as its business. A refusal on time limits, or on the ground that the harm was not connected to a policy the institution owed, tells you more about how to draft a future complaint than a successful investigation does.

• The Inspection Panel, from 1993, is the model for more than twenty later mechanisms
• They assess the lender's compliance with its own policies, not the state's conduct
• They cannot compensate or halt a project; they produce findings and action plans
• Jurisdiction follows the financier, so identifying it is the first task
• The binding safeguard version is generally the one in force at approval
• Most of the barrier is practical, not legal, and intermediaries are what close it
• In India the domestic forums carry the remedies; the mechanism carries the record

The honest assessment of this architecture is mixed and should be taught as such. These mechanisms were a real innovation: before 1994 there was no route at all by which a person harmed by an internationally financed project could be heard by the financier. They have produced findings that changed project designs and, in some cases, stopped them. They are also slow, procedurally demanding, unable to compensate, and reachable in practice mainly by people who have an organised intermediary. Both halves of that are true at once, and a practitioner deciding whether to invest months in a filing needs both halves rather than either one.

“Who is financing this project?” is not one question. It is four, and they have different sources and different difficulty.

It helps to know in advance which answers you are entitled to and which you are merely hoping for. Appraisal documents and environmental clearances are published as a matter of policy by most lenders and regulators, so failure to produce them is itself irregular. Concession agreements sit in a different category: commercial confidentiality is routinely claimed over them, sometimes legitimately and often more broadly than the exemption supports. Knowing which of the two you are asking for changes how the request should be framed and what refusal means.

• Name it precisely: official project name and any phase or package number
• Approving authority: which ministry, state department or board signed off
• Appraisal: MDB project pages, ministry sites, clearance portals
• Agreements: concession agreement, loan agreement, power purchase agreement
• Parties: sponsor, SPV, lenders, contractor, offtaker, and their corporate parents
• Compliance: clearance conditions, safeguard reports, audit findings

The order is not arbitrary. Each step supplies the identifiers the next one needs: the official project name unlocks the clearance portal, the approving authority tells you whose records to request, and the appraisal document names the parties whose filings you then search. Starting in the middle, which is the natural instinct when a project is already controversial, tends to produce a folder of documents about a project you cannot conclusively identify, which is the state in which most half-finished investigations are abandoned.

Work top to bottom. An RTI request drafted after the first six sources are exhausted is far more precise, and far harder to deflect, than one drafted first.

Pick one infrastructure project near where you work. Something with a name, a site and a visible contractor board. Then answer, in writing, with a source for each:

• What is the project's official name and sanctioned cost?
• Which authority approved it, and on what date?
• Is it budget-funded, a PPP, or a monetised asset?
• Who are the sponsor, the SPV and the lenders?
• Does it hold an environmental clearance, and what conditions attach?
• Was land acquired under the 2013 Act, and what R&R was due?
• If demand disappoints, who absorbs the loss?

• Split the question: who lends, on what terms, who builds, who carries the downside
• Six steps, from naming the project precisely to checking compliance
• Work the sources in order of yield; RTI last and therefore sharper
• Date every retrieval and keep the document rather than the summary
• Inability to answer the risk-allocation question from public sources is itself a finding

A final point about what this method is for. Tracing a project is not an end in itself, and a document collection is not an argument. The purpose is to be able to state, precisely and with sources, what a specific project commits the public to and what it owes the people affected by it. That statement is what a journalist can publish, a parliamentarian can ask about, a regulator has to answer, and a court can consider. Everything in this module is in service of producing one paragraph that can survive being checked.

• Year: reporting lags, and the lag is often longer than people assume
• Publisher: an institution reporting on itself is a source, not an independent one
• Method: face value or grant equivalent; public only or mobilised included
• Comparison: a total means nothing without the denominator or the counterfactual
• Currency: has the underlying series been revised since?

These five questions are most of what separates a citable claim from a repeated one.

A sixth question, for figures about India specifically: is the number for the Union or for the general government? Union figures exclude state spending, and in sectors where states do most of the spending, a Union figure quoted as a national one can understate the real total severalfold. The same trap appears in reverse with schemes that are centrally sponsored but state-implemented, where the Union release and the actual expenditure are different numbers reported by different bodies.

The Deep Dive is the natural next step: this deck teaches the architecture, and the reading list gives you the primary documents and the arguments about them.

• A finance figure is a construct; establish what it counts before comparing it to anything
• Concessionality and instrument, not headline size, determine what money does to a country
• The system has moved from lending to states towards making projects investible, and that shift is documented in the institutions' own policy
• In a financed project, risk allocation lives in the contract, and the contract is the document to obtain
• Two Indian statutes generate the public record of a project's social and environmental cost
• Accountability mechanisms judge the lender against its own rules; domestic forums carry the remedies

Export credit agencies are state-backed bodies that insure or finance their own country's exporters. They rarely appear in development finance discussions and they move very large sums into infrastructure in developing countries, because a power plant or a metro system is, from the exporting country's point of view, an export order.

A substantial share of infrastructure lending to developing countries now comes from bilateral lenders outside the traditional OECD donor group, with China the largest single source over the past two decades. The terms and the disclosure practices differ from MDB lending in ways that matter for anyone trying to trace a project.

Debt distress is not a binary. A country moves through rising debt service as a share of revenue, then difficulty rolling over maturing debt, then arrears, then restructuring. Each stage narrows the fiscal space available for everything else, which is why this is a development question and not only a finance one.

The sequencing matters for a practitioner because capital budgets are cut before salaries and transfers, so the first visible effect of a debt problem is often a stalled construction site rather than an announced austerity programme.

Coordination failure is the binding problem: each creditor class has reason to wait for the others to take losses first. The consequence falls on the debtor, which stays in limbo while the negotiation runs, and limbo is itself expensive.

Most Indian urban infrastructure is delivered by bodies with weak own-revenue and limited borrowing capacity. That constraint shapes what gets built more than any national policy does.

The fourth row is the one to look for and the easiest to miss. Where an MDB issues a partial risk guarantee, the borrowing state commonly counter-guarantees it, which means the risk the guarantee appeared to move off the public balance sheet has travelled back onto it by a second document.

Development impact bonds and outcome funds pay on verified results rather than on activity. An investor funds delivery up front and an outcome payer repays with a return if agreed results are achieved and verified.

Development finance discussion focuses on official flows, which are the ones with policy attached. For many countries they are not the largest external flow, and keeping the relative magnitudes in view prevents a common category error.

The rows behave differently in a crisis, which is the practically important point. Remittances have historically held up or risen when a receiving economy weakens; portfolio flows reverse fastest. A resilience plan built on the wrong row will fail at the moment it is needed.

A financing decision allocates money; a procurement decision determines who receives it. For a project's local economic effect, and for most corruption risk, the second matters more.

The deck closes here rather than with a call to action, deliberately. Development finance attracts strong positions, and most of them are held by people who have not read a concession agreement. The purpose of a foundational course is to put you in a position to form a view that survives contact with the documents, and then to change it when the documents say something different. If you finish this deck more uncertain than you started, but better able to say precisely what you are uncertain about and which document would resolve it, the deck has done its work.