Governance Analysis · MDB Reform Advisory · July 2026
The Deliberate Fog: The World Bank’s Information Architecture of Institutional Opacity
The World Bank has built a document architecture rather than a data architecture. Individual reports are technically available. But the information needed to evaluate institutional performance — project preparation costs, supervision costs, portfolio-wide ISR data, Board compensation, ASA budgets — is either not published at all, or scattered across disconnected systems that make independent portfolio-level analysis practically impossible. When a former Country Manager with twenty years of Bank service formally requested this data to underpin an independent accountability platform, the Bank refused — citing, for consolidated ISR access, the need to “create, develop, or collate information or data that does not already exist.” That information exists on every task team leader’s dashboard. The issue is not whether documents exist. It is whether the information needed to evaluate institutional performance is publicly available. The documentary evidence assembled here suggests it is not.
The Insider–Outsider Contrast
MDB Reform Advisory was established in 2023, following the author’s completion of twenty years of service at the World Bank as a Country Manager and Lead Governance Specialist, including a final posting as Lead Governance Specialist for West Africa. The platform was set up specifically to provide independent empirical analysis of the World Bank Group — drawing on the same project-level data the author had used throughout his career.
The experience that followed was revealing. As an insider, the data needed to analyse portfolio performance — project costs, supervision records, outcome ratings, country pipelines — was available on screen, consolidated, filterable, and updated in real time. The moment that insider status ended, the information wall went up.
- Country portfolio — 1 click
- Active projects — 1 click
- Closed projects — 1 click
- ISRs (portfolio view) — 1 click
- Supervision history — 1 click
- Financial data — 1 click
- ASA portfolio — 1 click
- Pipeline — 1 click
- Dashboards — 1 click
- Country portfolio — Not available
- Active projects — Project by project
- Closed projects — Project by project
- ISRs (portfolio view) — Not available
- Supervision history — Project by project
- Financial data — Multiple disconnected systems
- ASA portfolio — No public dataset
- Pipeline — Registered projects only; pre-registration preparation not visible
- Dashboards — Not available
What the Public Information Architecture Does Not Provide
Before examining specific gaps, it is worth naming the structural issue. The World Bank has built a document architecture rather than a data architecture. A document architecture makes individual reports, appraisals, and evaluations technically available. A data architecture would make the underlying information machine-readable, consistently organised, and linked across lending, supervision, preparation, and evaluation so that independent researchers could analyse institutional performance without reconstructing the database from thousands of individual files. The IEG ratings dataset is a partial exception — it is a genuine data product. What surrounds it is not.
Project Volume in five broad bands: less than $10M, $10–25M, $25–50M, $50–100M, and “$100M and above.”
Actual commitment amounts, actual disbursements, total project cost to borrower, and supervision costs. A $105M project and a $3B project are indistinguishable in the dataset. Cost per satisfactory project cannot be calculated.
Nothing. No public dataset of preparation costs, preparation staff weeks, or time from concept to Board approval.
The full cost of originating lending operations — including projects that were prepared but never approved. Without this, the true cost of the Bank’s lending pipeline cannot be assessed externally.
Individual ASA reports are often publicly available. No consolidated dataset of all ASAs, their costs, or their measurable outcomes.
Cost per ASA, usage analytics, evidence of influence or policy impact. The Bank spends approximately $1 billion annually on advisory services and analytics with no public evaluative framework. See companion paper: The Knowledge Bank.
Individual ISRs accessible on each project’s page — one at a time, with no portfolio-level view, no bulk download, and no country-level consolidation.
A consolidated, filterable portfolio view equivalent to what Bank staff access on the intranet. A country government wishing to see the Bank’s full active portfolio in its own country faces the same obstacle as an independent researcher: there is no public equivalent of what every task team leader sees on their dashboard.
Executive Director salaries — set annually by Board of Governors resolution — are technically findable as PDFs in the documents portal if one knows to search for them. Current salary: $296,390 for EDs, $256,400 for Alternates (effective July 2024).
No consolidated governance disclosure page. No public disclosure of individual pension arrangements (EDs may opt in or out of the Staff Retirement Plan). No consolidated statement of total Board compensation costs in any published financial report. A formal A2I request on Board compensation (Case AI9272, July 2024) was answered with two web links. The case was closed.
Phase 1 of the Doing Business data manipulation investigation — covering irregularities in the 2018 and 2020 reports — was published in 2021.
Phase 2, understood to cover earlier years and potentially additional officials, was never completed or released. The Bank discontinued the Doing Business report entirely rather than publish Phase 2’s findings. Had Phase 2 been completed, it could have clarified whether the identified problems were isolated incidents or reflected broader institutional weaknesses in data governance. A formal A2I request on this matter (Case AI9271, July 2024) was answered with four links to Phase 1 documents. The case was closed.
The Documentary Record
These gaps are not inferred. They are documented in a formal correspondence record spanning June–August 2024.
In June 2024, a request was submitted to the Bank’s documents team for bulk PDF access to Project Appraisal Documents and Implementation Completion Reports for the Bank’s full evaluated portfolio — a defined list attached to the request. No substantive response was received despite a follow-up ten days later noting that a promised data scientist had never made contact and that bulk downloads were essential for the research to proceed.
In early July 2024, those requests were filed formally under the Bank’s A2I process, covering project documents, ISR access, Board compensation, and the Doing Business Phase 2 investigation. The responses are instructive.
The Bank directed the requester to its Advanced Search portal and advised downloading an Excel spreadsheet of search results — not the documents themselves — and clicking through individual project pages. No bulk download facility was acknowledged to exist.
This position is difficult to reconcile with the Bank’s operational systems, where ISRs are routinely used to monitor active portfolios by every task team leader, Regional VP, and management dashboard in the institution. The information is not missing — it is not assembled for external users.
Two links provided: the Board of Directors webpage listing current Executive Directors, and the policies page for regulations relating to EDs and Alternates. No salary figures, pension terms, or total compensation costs disclosed. Case closed.
Four links to Phase 1 documents already in the public domain. A suggestion to email doingbusiness@worldbank.org. No information on Phase 2 provided. Case closed.
The Architecture of Opacity
None of the individual omissions described above is accidental. Each represents a decision point at which an institution that genuinely prioritised external accountability would have chosen differently. Taken together, they constitute an information architecture whose cumulative effect is to make independent cost-effectiveness analysis extraordinarily difficult.
A useful test of functional transparency is this: if Bank staff can answer a portfolio question in one minute using internal systems, while an external researcher requires days of manual reconstruction to answer the same question, the institution has not achieved meaningful transparency. It has achieved legal compliance. The two are not the same. Transparency should be judged by the ease with which an independent researcher can answer legitimate questions — not simply by whether individual documents can eventually be found somewhere on a website.
The IEG ratings dataset is real and useful. It is also carefully bounded. It tells a researcher whether a closed project was rated Satisfactory. It does not tell them what the project cost to prepare, what it cost to supervise, what the Bank’s internal systems say about projects still disbursing, what the institution spent on knowledge work that produced no evaluable output, what the governing Board costs its member countries in total compensation, or what an investigation into data manipulation actually found.
The gap between what is disclosed and what would be needed for genuine accountability is not a data management problem. It is a governance problem.
The World Bank is the largest single source of development finance for low-income countries. Its accountability architecture — inspection panels, IEG, the Access to Information Policy — is structured to give the appearance of transparency while preserving management’s ability to control what the public can see and analyse.
The issue is therefore not whether the information exists. It plainly does. The issue is whether the institution has designed its public information architecture to permit meaningful external scrutiny. The documentary evidence assembled here suggests that it has not.