IBRD Global Portfolio · Companion to Part 15 of the Zero Club Series · MDB Reform Platform
Two findings, both windows: of the IBRD projects approved since 2020, not one carries a calculated economic rate of return. A project rated Unsatisfactory disburses 89.2% of its commitment; a project rated Satisfactory, 99.3%. Outcome exerts only a limited influence on financial execution in the non-concessional window — the 10.1-percentage-point gap is wider than IDA Africa’s 1.8pp, but still leaves nearly nine-tenths of a failing loan deployed regardless of outcome. The Bank has not merely become disconnected from results. It has become disconnected from appraisal itself — and this is as true in its non-concessional window as in its concessional one.
The companion IDA Africa paper documented that $104 billion — 65.8% of all IDA Africa disbursements — went to operations that did not achieve a satisfactory outcome, and that ERR coverage had fallen from 71% to zero. A natural objection: IDA lends to fragile states and the world’s poorest countries; perhaps the pattern reflects the difficulty of those environments rather than any institutional failure.
IBRD is the test. The International Bank for Reconstruction and Development lends at near-market rates to middle-income and blend countries — governments with functioning finance ministries, professional project-management capacity, and access to capital markets. If the IDA Africa patterns were fragility artifacts or concessionality artifacts, they should not appear here.
They do. The appraisal collapse is Bank-wide. The disbursement decoupling is institution-wide. The sovereign guarantee operates the same way in both windows. The conclusion is structural.
Across 4,535 evaluated IBRD projects — $621 billion in original commitment, $554 billion disbursed, all regions, all income groups — $257 billion (46.4%) went to operations that did not achieve a satisfactory outcome. ERR coverage fell from 70% in the 1970s to zero since 2020. A project rated Unsatisfactory disburses 89.2% of its commitment, against 99.3% for one rated Satisfactory: a 10.1-percentage-point gap that leaves nearly nine in ten dollars of a failing loan deployed regardless of outcome.
The MENA region records a 33.6% S+ rate on $41 billion — worse than IDA Africa, on a portfolio of middle-income borrowers. The Water practice records 36.0% globally, the lowest of any major GP. MTI runs $98 billion at 40.7%. DPF accounts for 43% of all IBRD disbursements at the worst instrument S+ rate. The largest projects (over $500 million) record just 50.2% S+ on $189 billion.
The Bank has not merely become disconnected from results. It has become disconnected from appraisal itself. And this is as true in the non-concessional window as in the concessional one.
The appraisal collapse (ERR coverage by decade, IBRD vs IDA). The decoupling table (six outcome bands). The Eskom case file. Decoupling by region, GP, instrument, and size. The IDA–IBRD comparison table. Five reform proposals.
The Front End: Appraisal Has Lapsed in Both Windows
Of the IBRD projects approved since 2020, not one carries a calculated economic rate of return. Coverage was 70 percent in the 1970s. The collapse followed the same trajectory as IDA Africa, with a roughly decade-long lag — IBRD maintained higher coverage through the 1980s and 1990s because middle-income project teams had more capacity for economic analysis, and because the IBRD portfolio was historically heavier in infrastructure sectors where the methodology applied naturally.
Compare IDA Africa: 70% (1970s) → 36% (1980s) → 25% (1990s) → 15% (2000s) → 1% (2010s) → 0% (2015+). Both windows reach the same end state; IBRD arrived a decade later.
Where the calculation survives in the older IBRD portfolio, it shows the same optimism bias as IDA. Among the 1,355 projects carrying both an appraisal and a completion estimate, the median appraised return of 20.0 percent was realised at 17.7 percent on completion — a shortfall in 59 percent of cases. Of 1,993 IBRD projects with a closing ERR, 20.9 percent reported a rate at or below the Bank’s traditional 10–12 percent hurdle: zero-or-negative NPV on realised returns for about one in five projects that even reported a number. And of the 72 that reported negative ERRs at closing, 64 sat at exactly −5 percent — the same censoring pattern IEG documented across the broader portfolio.
How the collapse happened
The mechanism is documented Bank-wide in IEG’s 2010 cost-benefit evaluation. Of 51 task team leaders: only 5 gave cost-benefit analysis significant weight at identification; 82% said it had never been the key funding criterion; 80% said a “tick a box” sufficed; 92% said contributing did not improve their promotion chances. Management decisions were typically made before cost-benefit information was available. The OP 10.04 exemption — “benefits cannot be measured in monetary terms” — became the default route for avoiding the calculation.
The institutional response — and its consequence
In July 2010 the Board’s Committee on Development Effectiveness called the non-compliance “unacceptable,” questioned why management had not acted on a trend running since 1989, raised the Board’s fiduciary responsibility, and escalated to President Zoellick. What followed: OP 10.04 consolidated into OP 10.00 (2013); the Quality Assurance Group disbanded (2014); ERR coverage zero in both windows by the early 2020s. Three Presidents, fifteen years. The Board that called it unacceptable continued to approve every project without the measure it had said was essential.
The Central Finding: Disbursement Is Decoupled From Outcome
There is more of a gradient in IBRD than in IDA Africa — middle-income borrowers with access to alternative financing cancel more readily when projects fail badly — but the fundamental pattern holds: the bulk of the money deploys regardless of the result.
| IEG Outcome | Projects | Median Disbursed | Mean Cancellation |
|---|---|---|---|
| Highly Satisfactory | 170 | 99.8% | 6.0% |
| Satisfactory | 2,306 | 99.3% | 8.0% |
| Moderately Satisfactory | 1,020 | 97.5% | 11.1% |
| Moderately Unsatisfactory | 376 | 93.1% | 18.7% |
| Unsatisfactory | 588 | 89.2% | 23.8% |
| Highly Unsatisfactory | 55 | 25.6% | 62.0% |
The satisfactory half of the portfolio disburses a median 99.0 percent; the unsatisfactory half, 90.0 percent. The 10.1-percentage-point gap is wider than IDA Africa’s 1.8 percentage points — but Unsatisfactory projects still disburse 89 percent of their commitment. The mechanism is the sovereign guarantee. IBRD loans carry a sovereign obligation to repay at near-market terms whether or not the project delivers. No party in the chain absorbs a loss from continuing to fund a failing operation.
The mechanism in four projects across four regions
These cases span four regions, two lending instruments, three decades, and borrowers ranging from lower-middle-income to upper-middle-income. The common feature is not geography or income level, but the completion of disbursement despite poor outcome performance — and the sovereign guarantee that ensured all parties were repaid regardless of the development result. Russia’s SAL 2 (P050491, $800M, 100% disbursed, Unsatisfactory, FY1998) shows the same dynamic reaching back to 1998 — OED rated both outcome and Bank performance Unsatisfactory, finding insufficient economic and sector work before the loan was approved and fully disbursed.
Not a Regional Artifact: MENA as the Sharpest Case
| Region | n | S+ Rate | Disbursed | To Non-S+ |
|---|---|---|---|---|
| East Asia & Pacific | 1,010 | 63.9% | $129bn | $46bn |
| Europe & Central Asia | 914 | 53.7% | $117bn | $53bn |
| Latin America & Caribbean | 1,500 | 51.1% | $192bn | $92bn |
| South Asia | 223 | 48.9% | $45bn | $26bn |
| Middle East & North Africa | 301 | 33.6% | $41bn | $24bn |
| Africa (Eastern & Southern) | 172 | 57.6% | $10bn | $7bn |
| Africa (Western & Central) | 179 | 54.2% | $8bn | $5bn |
The MENA region is the striking outlier at 33.6 percent S+ — worse than IDA Africa’s 35.6 percent, on a portfolio of $41 billion lent to middle-income governments with functioning finance ministries. The income-level explanation for the outcome record has firm limits: the worst-performing IBRD region is not the poorest one. Latin America carries the largest volume at $192 billion and the largest absolute quantity of non-S+ disbursement at $92 billion.
The Practices and the Instrument
| Global Practice | n | S+ Rate | Disbursed |
|---|---|---|---|
| Water | 272 | 36.0% | $26bn |
| Health, Nutrition & Population | 167 | 38.3% | $17bn |
| Governance | 199 | 38.7% | $22bn |
| Macroeconomics, Trade & Investment | 307 | 40.7% | $98bn |
| Agriculture and Food | 151 | 41.1% | $13bn |
| Urban, Resilience and Land | 385 | 43.9% | $38bn |
| Education | 246 | 47.2% | $22bn |
| Transport | 376 | 51.1% | $61bn |
| Finance, Competitiveness & Innovation | 279 | 52.0% | $59bn |
| Energy & Extractives | 275 | 52.7% | $53bn |
| Social Protection & Jobs | 149 | 53.0% | $30bn |
Water is the worst-performing major practice globally at 36.0 percent — a finding with immediate operational significance given the Bank’s water security commitments under Mission 300 and related initiatives. MTI ranks fourth-worst at 40.7 percent but carries $98 billion, the largest disbursement volume among low-performing practices. DPF accounts for $237 billion — 43 percent of all IBRD disbursed — at a 50.8 percent S+ rate. PforR outperforms both at 67.6 percent, consistent with its structural requirement for independently verified disbursement-linked indicators.
Time in Portfolio: Duration, Timing, and What They Predict
The IBRD data carries board approval dates, effectiveness dates, and closing dates for virtually all 4,517 matched projects, permitting a timing analysis comparable to the IDA Africa paper — with two findings that diverge from the IDA pattern in informative ways.
Duration predicts outcome, but the signal is weaker than in IDA
Satisfactory projects close in a median 6.1 years from Board approval; unsatisfactory projects take 6.6 years. The 0.5-year gap is substantially smaller than IDA Africa’s 1.9-year gap. IDA fragile-state operations that go badly wrong tend to persist — extensions accumulate because borrowers lack institutional capacity to cancel and close. IBRD borrowers, with stronger institutions and access to alternative financing, close bad projects somewhat faster.
| IEG Outcome | Projects | Median Duration | Mean Duration |
|---|---|---|---|
| Highly Satisfactory | 170 | 5.6 yr | 5.2 yr |
| Satisfactory | 2,306 | 6.1 yr | 5.9 yr |
| Moderately Satisfactory | 1,020 | 6.7 yr | 6.1 yr |
| Moderately Unsatisfactory | 376 | 6.5 yr | 5.9 yr |
| Unsatisfactory | 588 | 6.7 yr | 6.5 yr |
| Highly Unsatisfactory | 55 | 5.8 yr | 5.8 yr |
The Highly Unsatisfactory band’s short median (5.8 years) is itself a finding: the most catastrophic IBRD failures are recognised and closed relatively quickly — the opposite of the IDA pattern, where HU projects are among the longest-running. IBRD borrowers exit faster when projects fail completely. Fast projects (≤ 6 years) succeed at 58.9%; slow ones (> 9 years) succeed at 42.7% — a 16-percentage-point gap.
Effectiveness lag is mildly predictive in IBRD — unlike in IDA
In the IDA Africa analysis, time from Board approval to first disbursement showed no relationship with outcomes. In IBRD there is a mild gradient: Satisfactory projects become effective in a median 142 days; Unsatisfactory projects take 174 days; Highly Unsatisfactory projects take 241 days. In the middle-income IBRD setting, where delays are less driven by systemic constraints, a slower start more often reflects genuine preparation problems that foreshadow implementation difficulty.
Disbursement runs to the end regardless of outcome
Across all six outcome bands, the median last-disbursement date falls within six months of the official closing date in 83 to 88 percent of projects. Whether a project is rated Highly Satisfactory or Highly Unsatisfactory, the money keeps moving until shortly before close. Disbursement is not a signal of project health; it is a function of project lifecycle.
Appraisal Quality Predicts Outcomes — The Same Gradient as IDA Africa
If the appraisal collapse has the consequences the paper argues, design quality at entry should predict outcomes in IBRD as it does in IDA Africa. The QAE ratings in the matched set are the direct test.
| IEG Quality at Entry | Projects | S+ Rate | Commitment |
|---|---|---|---|
| Highly Satisfactory | 172 | 95.3% | $30bn |
| Satisfactory | 1,544 | 70.9% | $266bn |
| Moderately Satisfactory | 653 | 26.6% | $130bn |
| Moderately Unsatisfactory | 349 | 4.9% | $51bn |
| Unsatisfactory | 482 | 11.4% | $57bn |
| Highly Unsatisfactory | 35 | 2.9% | $5bn |
When QAE is Satisfactory or better, 73.4 percent of IBRD projects achieve satisfactory outcomes. When QAE is below satisfactory, 8.4 percent do — on $114 billion in committed resources, of which $89 billion was ultimately disbursed. The gap is 65 percentage points. In IDA Africa, the equivalent gap is approximately 62 percentage points. The two windows produce almost identical QAE-to-outcome gradients despite serving borrowers at very different income levels and institutional capacities.
QAE also predicts disbursement. Projects rated Unsatisfactory on quality at entry disburse a median 82.6 percent of their commitment — the lowest in the matched set outside Highly Unsatisfactory. The connection between the two findings is direct: $89 billion was deployed to projects already rated below standard at the point of Board approval, at an 8.4 percent S+ rate.
The Comparison That Confirms the Structural Argument
| IDA Africa | IBRD Global | |
|---|---|---|
| Matched projects | 2,576 | 4,535 |
| Disbursed | $158bn | $554bn |
| S+ rate | 35.6% | 54.6% |
| Disbursed to non-S+ | $104bn (65.8%) | $257bn (46.4%) |
| Satisfactory half median disb | 99.3% | 99.0% |
| Unsatisfactory median disb | 97.5% | 89.2% |
| Gap | 1.8pp | 10.1pp |
| ERR coverage, 1970s | 70% | 70% |
| ERR coverage, 2020s | 0% | 0% |
| Appraisal → completion | 21% → 16% | 20% → 18% |
The S+ rates differ by 110.1 percentage points — largely compositional, reflecting borrower income and institutional quality. The appraisal optimism bias is virtually identical across both windows. Both reach zero ERR coverage. The decoupling holds in both because the sovereign guarantee operates the same way regardless of concessionality, and the incentive architecture rewards disbursement rather than results in both.
What We Are Proposing
For investment lending, where cost-benefit analysis plainly applies, no project should reach the Board without a credible estimate of whether its benefits exceed its costs. The methodology exists. The decline was institutional, not technical.
Report, for every Global Practice and region, the share of disbursement going to operations rated below the line in supervision. The MENA region’s 33.6% S+ rate on $41bn of IBRD lending deserves the same scrutiny as IDA Africa’s figures.
Consecutive Unsatisfactory ISRs should trigger mandatory Board review of whether to restructure, suspend, or close — in both windows. Not quiet continuation to maturity, as in Eskom.
The Water practice at 36.0% globally and MTI at 40.7% on $98bn warrant specific practice-level accountability in IBRD, just as MTI’s 21.1% in IDA Africa warranted it in the companion paper.
The reform is not that the Bank should lend less. It is that disbursement should answer to outcomes. That re-coupling begins by restoring the analytical function the institution let lapse — and insisting that the answer, once known, is allowed to move the money.
Companion Paper — Part 15
The IDA Africa paper: 2,576 projects, $158bn, $104bn (65.8%) to non-S+. ERR 71% → 0%. The front-end and back-end findings for the concessional window that this note replicates for IBRD.
The Structural Argument
Why the System Does Not Learn: A Game Theory Analysis (mdbreform.com) explains the equilibrium that produces these numbers in both windows. The sovereign guarantee is the parameter that makes the Nash equilibrium stable.
The DPF Record
Policy Without Performance: Isomorphic Mimicry and the DPO Incentive Trap analyses the instrument that now moves 43% of IBRD disbursements at the worst instrument S+ rate in both windows.
South Africa Eskom
Does the World Bank Learn from Project Failures? (Part 3, Country Case Studies) provides the full Eskom case file — the nine ISRs, the state-capture procurement record, and the disbursement timeline.
ERR coverage by decade (IBRD vs IDA). Full decoupling table. Eskom case file. Regional, GP, instrument, and size breakdowns. The IDA–IBRD comparison table. Five reform proposals. Methodological note.
$257 billion went to IBRD operations that did not achieve a satisfactory outcome. ERR coverage reached zero. An Unsatisfactory project still deployed 89% of its commitment. The QAE gradient — 73.4% S+ when design quality is adequate, 8.4% when it is not — is virtually identical to the IDA Africa pattern on a portfolio of middle-income borrowers.
The IBRD evidence matters precisely because it removes the usual explanations. The same patterns observed in fragile IDA countries appear in middle-income borrowers with stronger institutions, greater implementation capacity, and access to alternative financing. Fragility did not produce this. Concessionality did not produce this. Country capacity did not produce this. What remains is an institutional problem: a guarantee structure that insulates every party from the development outcome, an incentive architecture that rewards moving money rather than delivering results, and a thirty-year retreat from the analytical discipline that once asked whether the money was worth committing. Ajay Banga has the mandate and the moment to correct this in both windows simultaneously — and the IBRD data makes the case that this is a structural correction to the institution as a whole, not a concession to critics of IDA fragile-state lending.