The World Bank policy scorecard: The new conditionality?

22 November 2004 | At Issue

Available in fully-formatted PDF. (PDF does not include references.)

While accepted rhetoric says that donors respond to nationally-owned development plans, the reality is that these plans have little impact on either policy outcomes or the volume of loans a country receives(1). Opaque assessments conducted by the World Bank – known as Country Policy and Institutional Assessments – do. Critics argue that the scorecard is a way to coerce borrowers into adopting the Bank’s preferred model of economic development(2).

Policy conditions attached to loans simply do not work. Often the prescriptions are wrong for the patient. This can be due to analytical failings or ideological bias in their design. Regardless whether the prescription is right or wrong, a government that does not want to follow the conditions may be able to find a way around them. The debate over the content of the conditions continues, but there is growing consensus that the mechanism is broken. Even those conditions which are ‘successfully’ imposed distort government accountability, undermining the legitimacy of both the policies and the institutions which implement them.

Slower than most, the World Bank and the IMF are realising this. Pitched street battles, lengthy multi-stakeholder reviews and extensive advocacy efforts forced the institutions to change their discourse. Structural adjustment loans became poverty support credits. Ownership became the new touchstone. Civil society was invited to participate in the design of national development strategies. Four years after the introduction of the new nomenclature, civil society organisations said that the fundamental picture had not changed.

In the past year however, a perceptible shift has taken place in the IFI’s stance on conditionality. First, an admission on the part of the Fund that it’s conditions had extended beyond its mandate and competency. Now grudging agreement from the Bank to review its use of conditionality. Some critics argue that the changes are cosmetic. But if the change does result in the reduced use of conditionality, is it because the Bank and Fund recognise the shortcomings of conditionality? Or is it because conditions make loans unattractive, leading to shrinking portfolios and institutional prestige? Or have the major shareholders simply figured out a better way of getting the policies they want implemented without having to use such ham-fisted methods?

Banking the World Bank way

John needs £50,000 to open a new restaurant. John arrives at his bank, and the loan officer tells him – before looking at his business plan – that his credit limit is only £10,000. But that won’t be enough to open the restaurant, he protests. But she won’t hear it. The bank has done an assessment of all its customers – not just their financial history, but their education, their employment record, their lifestyle choices. And that is all that John can get. John asks to see the assessment – maybe a mistake has been made. But that, she says, is impossible – strictly against bank policy. More money might be available, she helpfully suggests, if John were to open a hair salon.

Country Policy and Institutional Assessments

Since the late 70s, at the behest of the deputies of the International Development Association (the Bank’s concessional lending window), the Bank has been making assessments of the policies of governments of the poorest countries to guide the allocation of cheap loans. In 1997, came the inception of a new, more formalised system. Since that time, the countries have annually received a country performance rating, known as the ‘IDA Country Performance’, or ICP.

The ICP is obtained by calculating a weighted average of the policy scorecard (80 per cent) and the Bank’s rating of the performance of outstanding loans to the country (20 per cent). This weighted average is then multiplied by a ‘governance factor’. The governance factor itself is drawn from the governance-related criteria in the scorecard (see box 2). The ICP, in conjunction with an assessment of need (based on gross national income per capita) determines the allocation of available funds. What this convoluted calculation means is that the subjective judgement of World Bank economists over the quality of a country’s public sector management plays an enormous role in the decision over how much money is available to that country.(3)

The policy scorecard, known as the Country Policy Institutional Assessment (CPIA), is made up of 16 indicators (see box 1). These ratings are prepared annually in all countries by Bank country teams and then subjected to a process of internal review. The exercise takes six months and is estimated to cost $1.5 million. Each criterion is given a score on a scale from one to six.

Box 1: CPIA rating criteria

A. Economic management
1. Monetary and exchange rate policy
2. Fiscal policy
3. Debt policy

B. Structural policies
4. Trade
5. Financial sector
6. Business environment

C. Policies for social inclusion
7. Gender
8. Equity of public resource use
9. Building human resources
10. Social protection and labour
11. Policies and institutions for environmental sustainability

D. Public sector management and institutions
12. Property rights and rule-based governance
13. Quality of budgetary and financial management
14. Efficiency and equity of revenue mobilisation
15. Quality of public administration
16. Transparency, accountability and corruption in the public sector

Lifting the veil

In 2000, the Bank began disclosing the CPIA ratings. But only in an aggregated format that revealed little about the differences between individual countries, why those differences existed, or how the ratings were calculated.

The Bank’s board members recognised that this secrecy would leave the Bank open to charges that lending decisions were driven by poorly substantiated and subjective decisions, if not outright realpolitik. In 2002, the governors of IDA urged full disclosure of the rating system to “allow it to benefit from open scrutiny”(4). They asked management to report to the board on the readiness of the system for public disclosure at a meeting scheduled for October 2003.

But management dithered on full disclosure. Instead, it proposed that ratings be disclosed in “half-point ranges”(5). Leaked documents from the October 2003 board meeting reveal a split amongst the Bank’s directors. While some directors “were concerned that disclosure of IDA’s ratings could have a negative effect on foreign investment”(6), others chastised management for taking a “step in the wrong direction” and leaving the “impression that progress was being made when this was not the case.” Why the delay?

Revealing in this respect was the discussion at the board around the rating methodology. A “large number of speakers” argued that disclosure should be delayed until the ratings had been improved. The current methodology “was based largely on staff judgement rather than clear, objective measurable indicators”. It was decided to put the ratings to an external review process to “assure that the Bank was on solid theoretical footing”. This frank admission raises serious questions about the legitimacy of the Bank’s past allocation decisions.

An external panel met for two days in Washington in February 2004 to review CPIA ratings and methodology. The panel was made up of nine experts: including 6 from US and European academic or research institutions, the Indonesian deputy minister for international cooperation, and a member of the NEPAD peer review secretariat. The panel made a number of recommendations(7):

  • Simplification of CPIA criteria from 20 to 16;
  • Undertake analytic work to better inform the weighting of the various criteria;
  • Reconsider the weight given to the ‘governance factor’, calling this calculation “highly non-transparent” and “excessive in light of the available empirical literature”;
  • Provide country authorities with an opportunity for comment on the assessments;
  • Establish an independent committee to review the CPIA methodology every three years; and
  • “Strongly in favour” of full disclosure of the numerical ratings of the 2005 CPIA exercise for IDA borrowers.

The Bank has accepted most of the recommendations of the panel. Importantly, full disclosure of the ratings will start for low-income countries with the 2005 ratings(8). No progress was made in making the governance factor “simpler and less volatile”.

Box 2: The ‘governance factor’

The ‘governance factor’ contains the following five criteria drawn from the CPIA(9):

  1. Property rights and rules-based governance: a good score requires, inter alia, that property rights be protected in “practice as well as theory”; laws and regulations affecting businesses are “transparent and uniformly applied”; obtaining licences is a small share of the cost of doing business; police force functions well and is accountable.
  2. Quality of budgetary and financial management: assesses extent to which budget is linked to policy priorities in national strategies; effective financial management; timely and accurate fiscal reporting; and clear and balanced assignment of expenditures and revenues to each level of government.
  3. Efficiency of revenue mobilisation: a good score requires that “bulk of revenues” be generated from “low-distortion” taxes such as sales/VAT, property, etc.; low import tariffs; tax base is free from arbitrary exemptions.
  4. Quality of public administration: assesses “policy coordination and responsiveness, service delivery and operational efficiency, merit and ethics, and pay adequacy and management of the wage bill”.
  5. Transparency, accountability and corruption: a good score requires accountability reinforced by audits, inspections and adverse publicity for performance failures; an independent, impartial judiciary; conflict of interest and ethics rules for public servants.

Undermining accountability

There are three question marks against the CPIA. The first is over the objectivity of the ratings; the second over their reliability; and the third is whether scorecards should be used at all in aid allocation decisions.

As for objectivity, several of the indicators explicitly reflect an economic bias: the indicator for trade policy, for example, rewards low tariffs, the absence of state marketing boards and the removal of controls on capital inflows. While most of the governance criteria reward behaviour which is lauded across the political spectrum, there is a bias towards such factors as stringent private property rights and low regulation of business.

As regards reliability, a number of indicators, such as those on gender, labour and environmental sustainability, are in areas where the Bank’s mandate and expertise, relative to other agencies, is in question. For all the indicators, the creators warn of “substantial margins of error” which mean that cross-country comparisons “should be made with due caution”(10). Small policy changes can have a significant impact on the ‘governance factor’. A move to harmonise the scorecards of the different multilateral development banks would further amplify this effect. What if the new harmonised scorecard is wrong?

Do ‘good policies’ as indicated by the CPIA foster economic growth and poverty reduction? Here there is great controversy and a growing literature. On one side, an “extraordinarily influential” study by Bank economist David Dollar says they do(11). However, a team of independent economists, lead by a former senior economist at the Bank, William Easterly, was given access to the CPIA database, and concluded that “foreign aid does not raise growth in a good policy environment”(12).

The third question is whether there is a role for any kind of scorecard in aid allocation decisions. Opponents of scorecards will rightly argue that they circumscribe policy space and promote external accountability, undermining that of governments to their citizens. Proponents counter that, unlike normal lending, there is a finite limit to the quantity of cheap loans and grants available; in the absence of scorecards, lending will continue to be made on even more opaque criteria. Even if it were possible, giving governments what they say they need risks punishing the citizens which eventually have to pay for overzealous borrowing. It would also fail to provide the accountability which citizens of donor countries demand of their governments.


The opening up of the CPIA process will provide an opportunity to revisit both its methodological and instrumental validity. Both critics and proponents believe that the ‘sunshine effect’ of increased transparency will lead to vigorous debate over the accuracy and role of the scorecards.

Before any further steps are taken towards harmonisation, an independent review should be commissioned to examine the relationship between rating criteria and poverty reduction. This review should inform an open debate where borrower country governments and civil society are able to express their opinions on the case for/against the use of scorecards and what their content should be.

If the use of scorecards survives such a debate, this review suggests:

  • Outcomes-based criteria, based on a government’s ability to improve the lives of its citizens, should take precedence over policy criteria which act as a form of ex-ante conditionality. Immediately, it will be pointed out that outcomes-based criteria have been used in the past to justify lending to brutal and/or corrupt regimes. This is why outcomes-based criteria must be accompanied by fiduciary criteria and some measure of respect for internationally agreed human rights.
  • A greater role for specialised agencies in the calculation of those ratings which are chosen would lend legitimacy to the exercise.
  • Needs assessment done in conjunction with any scorecard might be made more effective if human development measures were combined with the purely income-based indicator used currently. Research into this issue could examine what impact this would have.
  • The appropriate weighting of chosen criteria requires further independent, empirical investigation.

Without action on these points, the failed ‘one-size fits all’ policy prescriptions of the bad old days of structural adjustment lending may simply be reborn in policy scorecards. If this is the case, Bank and Fund moves towards ‘reduced conditionality’ may prove a boon to lending volumes but a pyrrhic victory for the poor.

Available in fully-formatted PDF. (PDF does not include references.)

Thanks to those who provided comments on a draft version of this briefing. Any errors are the fault of the author. Comment on this briefing.

(1) Evaluation of the IMF’s role in PRSP and the PRGF, IEO, August 2004. The PRS initiative: An independent evaluation of the World Bank’s support through 2003, OED, July 2004.

(2) See, for example, Judge and Jury: The World Bank’s Scorecard for Borrowing Governments, Citizen’s Network for Essential Services, April 2004.

(3) Governance accounts for an estimated 66 per cent of the final rating. IDA rating disclosure and fine-tuning the governance factor, September 2004. p. 3.

(4) Additions to IDA resources: Thirteenth replenishment (IDA/SecM2002-488), 17 September 2002.

(5) Disclosing IDA country performance ratings (IDA/R2003-0187), 9 October 2003.

(6) Disclosing IDA country performance ratings (SD2003-59), 28 October 2003.

(7) CPIA: An external panel review (SecM2004-0304), 15 June 2004.

(8) However, the double standard will remain whereby low-income countries scores are made public while middle-income countries scores are not. CPIA ratings are done for all countries, while ICP ratings are only calculated for IDA recipients.

(9)CPIA Assessment 2003: Assessment questionnaire.

(10) Governance matters III: new indicators for 1996-2002 and addressing methodological challenges, Kaufmann and Kraay. p. 1.

(11) ‘Aid, Policies, and Growth’, Craig Burnside and David Dollar. American Economic Review, September 2000, 90(4), pp. 847-68.

(12) Quoted in: Judge and Jury: The World Bank’s scorecard for borrowing governments, Nancy Alexander. CNES, April 2004. Confirmed with William Easterly.