Disbursement Disconnect · Part 15 of the Zero Club Series · MDB Reform Platform
Two findings. The front end: of the IDA Africa projects approved since 2015, not one carries a calculated economic rate of return — the figure that once told the Bank whether a project’s benefits exceeded its costs. Coverage was 71% in the 1970s. It is zero today. The back end: a project rated Unsatisfactory disburses 97.5% of its commitment; a project rated Satisfactory disburses 99.3%. The difference is 1.8 percentage points. The Bank stopped measuring whether projects were worth doing — and stopped letting the answer move the money.
When the World Bank approves a project, it does not hand over the money at once. It makes a commitment — a credit line of a fixed size — and then releases funds against that commitment over the life of the project as expenses are incurred and verified. The share of the original commitment that is actually paid out is the disbursement ratio. A project that draws down its full credit has a disbursement ratio near 100 percent; a project that draws down half and lets the rest lapse has a ratio near 50 percent. Disbursement is the clearest financial signal of whether the institution kept funding an operation or pulled back from it.
The only mechanism that reduces disbursement is cancellation — the formal decision, by the Bank or the borrower, to give up part of an undisbursed commitment before the project closes. Cancellation is the financial system’s one way of saying “this money will not be spent.” It can happen because a component is dropped, because conditions are never met, or because the project is failing badly enough that continuing to fund it is indefensible. The cancellation rate — the share of commitment given up — is therefore the one execution-side number that can, in principle, register that a project has gone wrong.
Both numbers sit downstream of a structural fact: the sovereign guarantee. IDA credits are sovereign obligations. The borrowing government is legally bound to service and repay the credit on its concessional terms — typically a multi-decade maturity at a fraction of a percent — regardless of whether the project the money financed achieved anything. The loan, in other words, performs as a financial instrument whether or not the project performs as a development intervention. No party in the disbursement chain — not the task team, not management, not the Board, not the Treasury — absorbs a balance-sheet loss when a funded project fails. This is the background condition against which every figure in this paper should be read.
The critical implication: disbursement and cancellation are decisions the institution controls in real time, while it can already see how a project is performing. The Bank rates each active project’s development objectives at every supervision mission, on a six-point scale, in an Implementation Status and Results Report (ISR). So at any moment the Bank knows whether a project is on track — and separately decides whether to keep disbursing. This paper places those two records side by side. It asks a single question: when the Bank’s own supervision says a project is failing, does the money stop? The answer, across 2,576 evaluated projects, is that it does not.
Of the Africa projects the World Bank has approved since 2015, not one carries a calculated economic rate of return — the figure that tells the institution whether a project’s benefits are expected to exceed its costs. Coverage was 71 percent in the 1970s. It is zero today. IEG’s own 2010 evaluation explains how the collapse happened: 82 percent of task team leaders said cost-benefit analysis had never been the key criterion in deciding to fund a project; 80 percent said it was sufficient to “tick a box”; management decisions to proceed were typically made before cost-benefit information was available. The Board’s own committee called the non-compliance “unacceptable” in 2010 and escalated to the President. The standalone policy was then consolidated, the Quality Assurance Group was disbanded, and coverage reached zero. The Bank has stopped asking the question at the front end.
At the back end, the money does not respond to the answer. Across 2,576 evaluated IDA Africa projects, roughly $158 billion was disbursed. Of that, $104 billion — 65.8 percent — went to operations that did not achieve a satisfactory outcome. A project rated Unsatisfactory disburses a median 97.5 percent of its commitment. A project rated Satisfactory disburses 99.3 percent. The gap is 1.8 percentage points. Only at the very floor of the scale does disbursement materially fall. The single execution-side variable that tracks failure is cancellation, and it is partial and late: $10.7 billion cancelled against $158 billion disbursed.
The Bank has not merely become disconnected from results. It has become disconnected from appraisal itself.
The decoupling exhibit (disbursement and cancellation by outcome band, all six tiers). The ERR-coverage collapse by approval decade. The sovereign-guarantee mechanism. Fragility test (FCS vs non-FCS). MTI and the DPF instrument. Time-in-portfolio and project-size gradients. Bank-versus-borrower performance and the lockstep finding. Three project case files: DRC Multi-modal Transport (disbursement after the flag), Nigeria Federal Roads and Ethiopia Road Sector Support II (appraisal that did not hold). Full methodological note.
The Front End: Appraisal Has Lapsed
Begin with a fact that ought to be a scandal and is instead unremarked: of the Africa projects the World Bank has approved since 2015, not one carries a calculated economic rate of return — the figure that once told the institution, and the public, whether a project’s benefits were expected to exceed its costs. Coverage falls steadily across every decade before reaching zero.
Part of the decline is defensible: a growing share of operations does not lend itself to a clean rate-of-return calculation, and for those a forced ERR would be spurious. But coverage fell even within investment lending, where cost-benefit analysis plainly applies. Where the calculation does survive, in the older portfolio, it is unflattering: among the 584 projects carrying both an appraisal and a completion estimate, the median appraised return of 21 percent was realised at 16 percent, and the completion figure fell short of appraisal in 64 percent of cases.
How the collapse happened: IEG’s own evidence
IEG’s 2010 evaluation — Cost-Benefit Analysis in World Bank Projects — explains the mechanism. From interviews with 51 task team leaders: only 5 said cost-benefit analysis was given significant weight at the identification stage; 82 percent said it had never been the key criterion in deciding to fund a project; 80 percent said it was sufficient to “tick a box”; 92 percent said it did not improve their promotion chances. The analysis was typically produced after the decision to proceed — a test applied after the decision cannot reject the decision. IEG documented a conflict of interest: the same staff responsible for the cost-benefit assessment were responsible for guiding the project through Board approval.
The exemption was the main channel. OP 10.04 allowed projects to skip the ERR if “benefits cannot be measured in monetary terms.” Of 93 investment projects closing in FY2008 without any cost-benefit information, 60 gave no explanation at all, and 24 invoked cost-effectiveness analysis but none actually applied it. Failing projects were selectively excluded from ex-post recalculation — the economic-accounting twin of the disbursement finding at the centre of this note: the money keeps flowing after the flag, and the ex-post reckoning disappears for the same cohort.
The institutional response — and its consequences
The Board’s own Committee on Development Effectiveness called the non-compliance “unacceptable” in July 2010, questioned why management had not acted on a trend running since 1989, raised the Board’s fiduciary responsibility, and escalated to President Zoellick. Management promised a policy consolidation with a timeline. What followed: the standalone cost-benefit policy (OP 10.04) was absorbed into OP 10.00 in 2013; the Quality Assurance Group — the body management had cited as evidence of improving quality — was disbanded in 2014; and ERR coverage reached zero by 2015. Three Presidents, fifteen years: a problem diagnosed, escalated, promised a remedy, and allowed to reach completion.
The Back End: Disbursement Is Decoupled From Outcome
If financial execution were disciplined by development results, disbursement rates would fall as outcomes worsened. The institution would slow or stop the flow of money to projects that were not working, and the money would concentrate in the projects that were. That is not what the record shows. Across the evaluated IDA Africa portfolio, disbursement is almost flat across the entire outcome scale, and only collapses at the extreme floor. The table below reports, for each IEG outcome band, the median share of original principal actually disbursed and the mean share cancelled.
Median gross disbursement against original principal. Source: IEG ICRR/PPAR database (March 2026) matched to IDA Statement of Credits (31 May 2026). Africa only.
The bars are almost the same length. A project that the Bank’s independent evaluators ultimately judge Unsatisfactory draws down 97.3 percent of its commitment — a figure statistically indistinguishable from the 99.4 percent drawn down by a project rated Satisfactory. Collapsing the scale into the two halves the Bank itself uses, satisfactory projects disburse a median 99.3 percent of commitment and unsatisfactory ones 97.5 percent. The difference is 1.8 percentage points. The resources move at essentially the same pace toward success and toward failure. Only at the extreme floor — the 38 Highly Unsatisfactory projects — does disbursement fall away, to 80.7 percent, and even there four-fifths of the money is spent.
This is the empirical content of a familiar observation about development finance: a lender protected by a sovereign guarantee is repaid whether or not the operation it financed delivers. That guarantee is the mechanism behind the flat line in the table. Because the credit is serviced by the sovereign regardless of whether the project works, no party in the disbursement chain absorbs a loss from continuing to fund an operation that is already visibly failing — so nothing in the financial machinery has cause to arrest the flow. What the data adds to the familiar observation is that the indifference runs upstream of repayment, into disbursement itself. The money is not merely recovered regardless of outcome. It is deployed regardless of outcome.
The Mechanism in a Single Project: DRC Multi-modal Transport
The portfolio-level pattern is visible in the supervision record of almost any large failing operation. The DRC Multi-modal Transport Project (P092537) — an IDA investment operation approved in 2010 to restore the national railway and strengthen transport state-owned enterprises — was rated Unsatisfactory at completion, with Risk to Development Outcome High. Its Implementation Status and Results Reports record when the money moved relative to when the Bank knew the project was failing.
| Implementation report | Date | Development-objective rating | Cumulative disbursed |
|---|---|---|---|
| ISR 7 | Sep 2013 | Satisfactory | $175.6m |
| ISR 9 | Aug 2014 | Unsatisfactory | $252.8m |
| ISR 11 | Dec 2015 | Unsatisfactory | $324.7m |
| ISR 14 | Oct 2017 | Moderately Unsatisfactory | $366.9m |
| Closing | Jun 2018 | Final outcome: Unsatisfactory | ~$385.6m |
The Bank’s own implementation reports rated the development objective Moderately Satisfactory or better for four years, peaking at Satisfactory in September 2013 with $175.6 million disbursed. The rating was then cut to Unsatisfactory in August 2014, with about $252.8 million already out, and it never recovered — staying at Unsatisfactory or Moderately Unsatisfactory through every subsequent report to closing. Yet disbursement continued, to roughly $385.6 million. About a third of the total — some $133 million — went out the door after the Bank’s own supervision had flagged the project as Unsatisfactory.
The decision to keep funding was explicit rather than inadvertent. The completion report records that suspension of the railway component was proposed by the project team in 2014 and again in 2015 but did not occur, because it “would have had negative impact on the Bank–Government relationship.” A $180 million additional financing had been approved in June 2013 — two months before the mid-term review that might have prompted an exit. And the appraisal that launched the operation was thin against the stake: the Bank spent about $1.1 million preparing the project and roughly $3.2 million supervising it — three times as much watching the operation as deciding whether to start it. An economic rate of return was calculated for the largest component only, resting on the assumption that the railway would collapse and force freight onto the road, a shift that had already occurred by appraisal. Efficiency was ultimately rated Negligible. The completion report’s own verdict on quality at entry was an “unrealistic” central objective, set without reliable financial data, with institutional diagnostics deferred to the implementation phase.
Two Appraisals That Did Not Hold
The DRC railway is the decoupling case — money disbursed after the rating turned. Two further project files show the appraisal failure directly, and both were rated Moderately Unsatisfactory on quality at entry by the Bank’s own reviewers.
The Nigeria Federal Roads Development Project ($333 million in IDA financing, approved 2008) was designed around an untested output- and performance-based contracting model and the creation of three new road institutions — a design built, the completion report concedes, on “fast-track pre-feasibility studies which were technically deficient,” with no market sounding for the contracting approach and no analysis of the fiscal-federalism risks it depended on. Contractors declined to bid; the entire approach was abandoned in a 2011 restructuring that dropped both the contracts and the institutions. The Bank spent about $0.9 million preparing the operation and roughly $1.4 million supervising it — the same lopsided ratio as the DRC railway.
The Ethiopia Road Sector Development Support Program II ($348 million in IDA financing, approved 2004) did carry an economic rate of return — and the record shows both its optimism and its fragility. Insufficient engineering design at preparation left cost estimates that bids exceeded by as much as 200 percent; the final cost reached 236 percent of the appraisal estimate, absorbed through two successive additional financings of $87 million and $100 million. The appraised economic return of about 27 percent was re-estimated at 19 percent on completion, and the independent evaluators downgraded the project to Moderately Unsatisfactory.
The Appraisal Mechanism in Detail: IEG’s 2010 Evaluation
The coverage figures document the disappearance. They do not, by themselves, explain it. A separate IEG evaluation — Cost-Benefit Analysis in World Bank Projects (Independent Evaluation Group, 2010) — does, and the findings bear directly on the pattern this note documents because they describe the institutional machinery that allowed appraisal to lapse.
The decision precedes the analysis
IEG interviewed 51 task team leaders chosen randomly from projects closing in FY2006–09. Only 5 of the 51 reported that cost-benefit analysis was given significant weight at the project identification stage. When asked whether a cost-benefit analysis had ever been the key criterion in deciding to fund a project, 82 percent said it had not. Management decisions to proceed were typically made before cost-benefit information was available — so the positive-NPV test required by OP 10.04 was not, in practice, screening projects at the point of approval.
Tick a box
Eighty percent of task team leaders agreed that for a project proposal it was sufficient to have a cost-benefit analysis present — what several described as needing to “tick a box.” Forty-one of the 51 reported that an outside consultant was hired to perform the analysis; only 10 did it themselves. Time allocated varied starkly: Public Sector Governance task leaders allotted just two days — the lowest in the sample — against about 20 days in Transport. Ninety-two percent said contributing to the analysis did not improve their chances of promotion. IEG’s framing: there is a conflict of interest when the Bank places responsibility for the cost-benefit assessment in the hands of the same staff responsible for guiding the project through Board approval.
Quality declined
A review by Belli and Guerrero found economic analysis in appraisal documents to be acceptable or good in 54 percent of cases, against approximately 70 percent in comparable reviews in the 1990s. The sectoral variation was wide: Water scored 81 percent, Agriculture 67 percent, while Urban Development, Education, and Health/Nutrition/Population were rated acceptable or good in only 20 to 40 percent of cases — the same sectors where the “benefits cannot be measured in monetary terms” exemption from OP 10.04 was most routinely invoked.
The exemption became the default
OP 10.04 allowed an exemption from the ERR requirement “if the project is expected to generate benefits that cannot be measured in monetary terms.” IEG found that of 93 investment projects closing in FY2008 without any cost-benefit information, 60 provided no explanation or simply asserted non-applicability. Twenty-four invoked cost-effectiveness as the standard — but none actually applied cost-effectiveness analysis. The exemption designed for genuinely non-monetizable benefits had become the default route for avoiding the calculation altogether.
Failing projects disappear from the ex-post record
The complement to the front-end disappearance is a back-end one: projects with lower outcome ratings were less likely to have their ERRs recalculated at closing. The difference in recalculation probabilities between low- and high-rated projects averaged 0.24 after 1987, against 0.18 before — suggesting the bias grew over time. This is the economic-accounting twin of the disbursement finding: the money keeps flowing after supervision flags a project as failing, and the ex-post economic reckoning quietly disappears for the same cohort.
Among the projects that did report an ERR at closing, the bottom of the distribution was censored. Of 1,299 projects closing since 1990 with a reported ERR, only 24 reported negative returns — and 16 of those sat at exactly −5 percent. The overall correlation between reported ERRs and IEG outcome ratings was only about 40 percent. And in the cross-tabulation of 1,535 rated projects with a reported ERR (1987–2008), roughly 285 — about 19 percent — reported an ERR of 10 percent or lower, at or below the Bank’s traditional hurdle rate: a zero-or-negative net present value on realised returns for about one project in five that even reported a number.
The Institutional Response: From “Unacceptable” to Zero
The IEG evaluation did not disappear into a filing cabinet. It was discussed by the Committee on Development Effectiveness (CODE) on 21 July 2010, with management present, and the institutional record of what followed is printed inside the report itself. What the record shows is an institution that recognised the problem, pledged to fix it, and then presided over the completion of the collapse it had promised to arrest.
Management disputed the severity
Management ran its own review of 795 investment operations approved between July 2007 and December 2009 and concluded that “at least 72 percent of all operations meet the strict requirement of OP 10.04,” with a further 12 percent “substantively acceptable.” But the measure was different from IEG’s. IEG had counted whether a project carried an economic rate of return. Management counted anything that could be argued to satisfy the policy — including the “benefits cannot be measured in monetary terms” exemption route that IEG had identified as the principal channel through which projects avoided the calculation altogether. The 72 percent compliance figure was built, in significant part, on the exemption IEG had just documented as the mechanism of the decline.
Management also cited the Quality Assurance Group (QAG), which had conducted eight reviews between 1997 and FY2008 and found “a major improvement in the quality of economic analysis,” with 96 percent of projects rated marginally satisfactory or better. IEG’s own assessment, in the same report: QAG’s quality-at-entry ratings “are performed after projects have been approved by the Board and therefore have no effect on approval decisions.”
The Board was sharper
CODE members expressed “serious concerns regarding the non-compliance of OP 10.04 and accountability issues in this regard.” They questioned why management had not identified and acted on the trend earlier — noting the decline had begun in 1989–90, two full decades before the evaluation. Several members “considered it unacceptable that Bank operational policies are simply disregarded and also raised the issue of the Board’s fiduciary responsibility.” The Committee Chair escalated the findings “to the attention of other committee chairpersons and the President.”
What followed
Management committed to consolidate the policy framework for investment lending by fall 2010. IEG had already noted that the investment lending reform concept note of January 2009 — the reform management was working on — “does not mention cost-benefit analysis.”
In 2013, OP 10.04 — the standalone economic-evaluation policy — was consolidated into OP 10.00, the new umbrella policy for Investment Project Financing. The standalone ERR requirement was absorbed into a broader framework. In 2014, the Quality Assurance Group — the body management had cited to the Board as evidence of quality — was disbanded. Nothing replaced it. And ERR coverage, which had already been falling for three decades, completed its descent to zero by 2015.
Cancellation: The Only Signal, and It Is Blunt
The one place where execution and outcome do move together is cancellation. Mean cancellation rises monotonically across the rating scale, from roughly 2–4 percent in the satisfactory bands to 16 percent at Unsatisfactory and 34 percent at Highly Unsatisfactory. Failing projects, in other words, do not draw down less of what they retain — they retain less, and disburse the remainder in full.
| IEG outcome | Projects | Median disbursed | Mean cancellation |
|---|---|---|---|
| Highly Satisfactory | 53 | 98.8% | 2.1% |
| Satisfactory | 864 | 99.4% | 3.9% |
| Moderately Satisfactory | 813 | 97.8% | 4.8% |
| Moderately Unsatisfactory | 383 | 97.8% | 8.8% |
| Unsatisfactory | 414 | 97.3% | 16.0% |
| Highly Unsatisfactory | 38 | 80.7% | 34.3% |
Cancellation is therefore the portfolio’s de facto, and only, financial accountability mechanism. But it is a blunt and trailing one. Across the matched set, total cancellations come to about $10.7 billion against $158 billion disbursed; the signal arrives, where it arrives at all, after commitment and often after substantial drawdown. It registers trouble; it does not prevent it. The DRC railway again illustrates the timing: the $23 million eventually cancelled was a sliver of a $385 million drawdown, processed at closing — years after the supervision record had turned, and chiefly to preserve the country’s IDA allocation rather than to discipline the operation.
The Pattern Is Structural, Not a Fragility Artifact
The most obvious objection to the decoupling finding: the Bank disburses to failing projects because it operates in failing states, where outcomes are bound to disappoint and stopping disbursement is not a realistic option. The data does not support this reading. Projects in countries classified as fragile or conflict-affected (FCS) rate slightly higher on the S+ standard than those in non-FCS countries — 38.2 versus 35.2 percent — cutting against the assumption that fragility mechanically depresses outcomes. More importantly, the disbursement decoupling is present in both groups in the same form.
| Environment | S+ outcome rate | Disbursement — satisfactory half | Disbursement — unsatisfactory half |
|---|---|---|---|
| FCS (fragile / conflict-affected) | 38.2% | near-full | near-full |
| Non-FCS | 35.2% | near-full | near-full |
In fragile and non-fragile settings alike, satisfactory and unsatisfactory projects disburse at near-full and near-equal rates, with cancellation the only differentiator. The decoupling is thus a property of the portfolio’s operating model — the sovereign-guarantee incentive structure that runs through every environment — not of the difficulty of the environments in which the Bank works. The point matters because “these are hard places” is the standard institutional answer to the outcome record, and on the disbursement question it is not the explanation.
The Macroeconomic Practice and the Instrument of Least Accountability
Disaggregating by Global Practice surfaces a second, narrower pattern that this paper records descriptively and takes up at length in two companion analyses. The Macroeconomics, Trade and Investment practice (MTI) is the largest single GP in the evaluated Africa portfolio and has the lowest S+ rate of any major practice, against a GP-wide average near 29 percent. It is also the practice that runs policy lending: roughly 71 percent of Africa’s Development Policy Financing operations sit under MTI, and DPFs as a class underperform investment lending.
| Cut | S+ rate | Note |
|---|---|---|
| MTI (all instruments) | 21.1% | Lowest of any major Global Practice; ~29% portfolio average. |
| Development Policy Financing | 28.9% | Against 38.1% for investment lending. Carries no ERR by design. |
| Investment lending (IPF) | 38.1% | Cost-benefit analysis applies; ERR coverage has still lapsed. |
| MTI on its investment-lending portfolio | 10.0% | 30 projects — a thin sample, but the practice rates lower still. |
The instrument and the practice are difficult to separate, because they substantially coincide. But one observation cuts against the easy explanation that policy lending is simply harder to rate well: on its small investment-lending portfolio, MTI rates lower still — 10 percent S+, on 30 projects. The practice underperforms whatever instrument it touches. Whether this reflects mandate, capability, or the design role macroeconomic staff have come to play in operations is a question of institutional political economy taken up in the companion notes Policy Without Performance: Isomorphic Mimicry and the DPO Incentive Trap and Institutional Power Architecture and Portfolio Distortion at the World Bank. What the data establishes here is the descriptive fact: the worst-performing major practice also owns the worst-performing instrument — the instrument for which the appraisal economist’s central question is never posed.
Time and Size: The Operations That Disburse Slowest Fail Most
Two further structural regularities round out the picture. The first concerns time in the portfolio. Implementation duration is strongly and monotonically associated with outcome: among investment projects, the S+ rate falls from 48.6 percent for operations completed in under three years to 30.4 percent for those running beyond seven. The relationship is partly endogenous — troubled projects are restructured and extended, so duration is in part a symptom of difficulty rather than a cause of it — but the direction is unambiguous. Prolonged time in the portfolio is the company of failure. Read alongside the disbursement finding, it completes the picture: money leaves regardless of outcome, and the operations that take longest to push it out are the ones most likely to fail.
| Original principal (IPF) | Projects | S+ rate | Median implementation |
|---|---|---|---|
| Under $25M | 952 | 44.6% | 5.8 yr |
| $25–75M | 675 | 31.7% | 6.3 yr |
| $75–150M | 227 | 30.0% | 6.8 yr |
| $150–300M | 132 | 28.0% | 7.8 yr |
| Over $300M | 56 | 33.9% | 8.0 yr |
Outcomes are strongly graded by size, and not in the direction a “larger means better-resourced” intuition would predict. Operations under $25 million reach a satisfactory outcome about 45 percent of the time; everything larger clusters between 28 and 34 percent. The small, narrowly-scoped operation outperforms the rest of the portfolio by a wide margin, while the mid-to-large investment project — complex enough to be demanding but not large enough to command flagship attention — is the structural weak spot. Size also entangles with the duration finding: implementation length rises steeply with size, from 5.8 years for the smallest operations to 8.0 years for those above $300 million, and large projects almost never finish quickly. The best-performing profile in the data is the small, quickly-implemented project. Size and prolonged implementation travel together, and length is itself the company of failure.
Performance Ratings That Move in Lockstep
The evaluation record attaches two performance ratings alongside the outcome — one for the Bank’s own conduct, one for the borrower’s — and these are, in principle, where responsibility for a result is located. In practice, for the Africa portfolio, they do little independent work. On the 1,770 Africa projects rated on both, in the pre-2018 window when both were still recorded, Bank performance and borrower performance are judged almost identically, and almost identically poorly: 38.4 and 35.8 percent satisfactory-or-better respectively. The two ratings are equal in 70 percent of projects.
What the ratings do not do is discriminate responsibility when a project fails. Among the operations whose outcome fell below the line, IEG marks down both the Bank and the borrower together in 88 percent of cases; it faults the borrower alone in 7 percent and the Bank alone in just 4. The Bank’s own performance — its design and its supervision — is rated sub-satisfactory in 92 percent of failed projects. The performance assessments are co-determined with the outcome rather than independent diagnoses of it: when the result is bad, both parties’ performance is marked bad in tandem, with a slight institutional inclination to spare the Bank.
This closes a circuit with the central finding. The evaluation record acknowledges, for the large majority of failing projects, that the Bank supervised them poorly — and yet, as the disbursement evidence shows, those same projects drew down some 97 percent of their funds. The supervision rating registers the failure; the disbursement system proceeds as though it had not. And the retirement of the borrower-performance rating after FY2017 is of a piece with the disappearance of the economic rate of return: within a few years the institution let lapse two of the measures by which a project’s performance, and the responsibility for it, might have been independently judged.
The Board: 53 Years of Reports, Zero Corrective Action
Every project in this record was approved by the Board of Executive Directors, and every disbursement was made against a commitment the Board authorised. The Board owns the outcome risk and holds the sovereign guarantee. IEG has been transmitting independent evaluations of those outcomes to the Board for more than half a century. Across that period, the Board has taken no formal corrective action in response to the persistent disbursement-outcome gap those evaluations document. The reports are received, noted, and shelved. The money continues to flow at near-full rates to projects the same reports rate as failures.
This makes the evaluation function, in its present institutional position, largely cosmetic on the financial question — not because IEG fails its mandate, but because its findings reach a Board that authorised every commitment in the record and therefore cannot discipline the disbursement pattern without implicating its own decisions. Management receives IEG’s findings first and frames the Board’s encounter with them. The evaluation exists; the consequence does not. The situation the Bank is in is, ultimately, the Board’s responsibility.
What We Are Proposing
The reform is not more analysis. The Bank’s own data already identifies the problem with precision. What is required is a change to the machinery that commits money at the front end and keeps disbursing it at the back end, irrespective of results:
For investment lending, where cost-benefit analysis applies, no project should reach the Board without a credible estimate of whether its benefits exceed its costs — and a completion estimate against which the appraisal can be judged. The diagnostic was not made obsolete; it was simply dropped.
Report, for every Global Practice and country department, the share of disbursement going to operations rated below the line in supervision. What is measured and surfaced to the Board can be governed; at present the gap is visible only in retrospect, project by project.
A project that records consecutive Unsatisfactory ISRs should not continue to disburse on autopilot. Sustained sub-satisfactory supervision should trigger a mandatory Board-level review of whether to restructure, suspend, or close — not a quiet continuation to closing, as in the DRC railway.
Internal performance is measured substantially by commitment and disbursement against targets. Link the standing of practices and managers to the IEG outcome record of what they financed, so that delivering results — not moving money — is what advances a career and a unit.
The reform implication is not that the Bank should disburse less. It is that disbursement and outcome should answer to one another. That re-coupling begins where the decoupling did: by restoring the analytical function the institution let lapse, and insisting that the answer, once known, is allowed to move the money.
Series Context
Quality at Entry in IDA (Part 14)
The companion front-end paper. When design quality and M&E are both strong, 73% of IDA projects succeed; when both are weak — 52% of the portfolio — 5.4% do. QAE is the design failure; this paper is the financial consequence that follows it.
Policy Without Performance — the DPO Incentive Trap
The instrument analysis. Why Development Policy Operations persist despite the outcome record: conditionality met on paper, isomorphic mimicry, and the incentive structure that rewards the prior action rather than the reform.
Institutional Power Architecture and Portfolio Distortion
The practice analysis. Economist dominance in country management and DPF instrument bias across eight Sub-Saharan African CMUs — the political economy behind MTI’s position as the worst-performing major practice.
Game Theory: Why the System Does Not Learn
The mechanism. The decoupling is a Nash equilibrium: under a sovereign guarantee, no actor’s payoff improves by arresting disbursement to a failing project, so the flow persists regardless of the documented outcome record.
| # | Paper | Scope | Headline | Status |
|---|---|---|---|---|
| 11 | Zero Club — Transport | Africa | 0% Satisfactory | Published |
| 12 | Zero Club — Education | Africa | 0% Satisfactory | Published |
| 13 | Zero Club — Cross-Sector Synthesis | Africa | 0% Satisfactory | Published |
| 14 | Quality at Entry in IDA | 2,141 projects | 5.4% when design & M&E weak | Published |
| 15 | Disbursement Disconnect (this paper) | 2,576 projects, $158bn | 65.8% disbursed to non-S+ | This paper |
The decoupling exhibit (six outcome bands). ERR-coverage collapse by decade. The sovereign-guarantee mechanism. DRC, Nigeria and Ethiopia project case files. Fragility test, MTI/DPF, time and size gradients, Bank-versus-borrower lockstep, and the full methodological note.
IDA controls two things absolutely: whether it measures the value of a project before committing the money, and whether it keeps disbursing once its own supervision says the project is failing. On the first, it has stopped measuring — ERR coverage has fallen from 71% of Africa projects to zero. On the second, it does not stop — a project rated Unsatisfactory disburses 97.5% of its commitment, against 99.3% for one rated Satisfactory, and $104 billion of the $158 billion disbursed went to operations that did not achieve a satisfactory outcome.
The cause is structural, not a matter of effort or environment. Under a sovereign guarantee the credit is serviced whether or not the project works, so no party in the chain absorbs a loss from funding a failing operation, and the internal incentive rewards moving money over delivering results. The reform is to re-couple the financial system to the results system: restore the appraisal diagnostic at the front end, make the disbursement-outcome gap a governed Board metric, and stop disbursing on autopilot through consecutive Unsatisfactory supervision. The Bank has not merely become disconnected from results. It has become disconnected from appraisal itself. The evidence is in. The choice is institutional.