and IMF 2003). Moreover, the demands of managers and oversight agencies evolve from changes at the top, and turf considerations may also lead to a desire for separate products.

At the procedural level, arrangements for peer reviews and other quality control mechanisms vary considerably among development agencies, which reflects their managerial styles and administrative cultures. As a result, there may be genuine professional concerns (as opposed to efforts made for self-protection) about the quality of assessments produced by other agencies. Finally, as in all bureaucracies, groups within agencies fight for their turf and for maintenance of the status quo. Thus, even when boards or top managers announce policies intended to reduce overlap and turf battles within and between agencies, line managers and staff may be able to subvert or ignore such instructions for a considerable period.

These constraints have three important implications. First, to the extent that integration occurs, it is somewhat more likely to happen within than between agencies. Second, when a recipient government has enough capacity, clarity, and political power, it can usually compel coordination among development agencies. Third, thinking in terms of fewer instruments misses the point. Although further proliferation should be halted, some instruments can rely largely on the analysis of others without forsaking the agencys independent judgment or desire to have its own separate product. Thus efforts to improve assessments should not focus on eliminating or merging instruments to reduce their number. Rather, they should focus on ensuring the substance and quality of analyses on which all concerned parties can rely, and on improving coordination and reducing costs for recipient governments. And as noted, there is scope for streamlining the scope and coverage of instruments to reduce unnecessary overlap and fill gaps- and possibly going further and developing a programmatic and modular approach, as discussed below.

Exchanging information and programming intentions

Despite the risks and costs of duplication, there is no systematic exchange of information among agencies on their plans for conducting assessments. The issue here is not the need to create a reliable database on fiscal and budget issues shared by development agencies and recipient governments. That vast task is already being addressed through other donor coordination mechanisms. For example, a lot of information is available from IMF databases and statistical publications on government finance. In addition, the

OECD is preparing (with World Bank support) a new public expenditure database containing budget information for a range of countries, including some developing and transition economies.

Of concern here is the narrower but still important need of agencies to exchange systematic, timely information on their plans for conducting assessments, and to share the reports. This task should be aided by the Country Analytic Website being developed by the World Bank and other donors.2 In addition, steps are being taken to give development partners online access to IMF and World Bank documents. Both efforts should make it easier to improve the sequencing of assessment instruments. But better programming also requires the cooperation of country project managers-and is achieved when a recipient government has sufficient capacity and political strength to take the lead in coordinating assessment work among agencies and instruments.

Coordination between agencies

The nature of budget support provides a powerful incentive for donors to collaborate. Indeed, it was one of the forces driving the creation of PEFA. There are many examples of successful collaboration between multilateral institutions: the Bank and the IMF on HIPC AAPs, multilateral development banks (and several bilateral agencies) on CPARs,3 and the Bank and the Inter-American Development Bank on CFAAs. Most donors must justify their provision of budget support using sound assessments of recipient governments financial management and accountability systems.

But while aid agencies are generally unwilling to risk relying entirely on assessments conducted by other agencies, most bilateral agencies do not have the resources and skills required to conduct robust assessments. The obvious response would be to conduct joint assessments, but in most cases these are impractical.4 Such assessments would result in multiagency teams of 20-30 people on three- or four-week missions-a managerial nightmare that would also force the recipient government to put aside its regular activities for a long period.

So what is the solution? This predicament creates an important avenue for coordination and raises an equally important corollary. Better coordination and lower transaction costs can be achieved if bilateral development agencies rely on the assessments of multilateral agencies equipped to perform them-normally the World Bank and, on fiscal transparency issues, the IMF. But these institutions must be far more open to cooper-

ation with each other and to suggestions, requests, and criticisms from their partners.

At a minimum, given that the quality of the analysis is paramount for improving public expenditure management and financial accountability, the IMF should strengthen its ex ante consultation with the Bank on the terms of reference for its Fiscal ROSCs, and the Bank should reciprocate for its PERs and CFAAs (once consistency between these two instruments has been assured, as proposed below).

The Bank and IMF already share some documents and information and request comments from one another. A broad division of labor has been established-beginning with 1981 guidelines for Bank-IMF collaboration-that has largely been respected, though not without friction. Peculiarities persist, especially in the separate public expenditure management activities of the Banks Public Sector Governance Board and the IMFs Fiscal Affairs Department, sometimes leading to divergent results. There are also strong differences of approach between the areas and regions of the Bank and the IMF. The ongoing review of Bank and IMF collaboration on public expenditure issues should be used to propose changes and streamline work on both public expenditure and revenue management.

Bank-IMF collaboration also raises the issue of overlap between CFAAs and Fiscal ROSCs. Both institutions seem to accept that such overlap is substantial, particularly in assessing transparency and fiduciary risk. All the elements described as being central to assessing risk to Bank funds in the recent revision of CFAA guidelines are explicitly defined elements of the IMFs fiscal transparency code. So, in principle, if a ROSC has been completed, a CFAA may only have to deepen or broaden certain aspects of its analysis to achieve its fiduciary risk objective. And if the sequence is reversed, the ROSC should be able to apply CFAA findings directly to its assessment of public expenditure management rules and procedures.

Thus it should not require much effort to combine the collection and analysis of information, through joint teams and missions, using a combined questionnaire and database, and produce a joint report that satisfies both institutions fiduciary and surveillance objectives. The two institutions could also use the information for other purposes-for example, the Bank could use it to feed into the dialogue with countries on reforms linked to Country Assistance Strategies and Poverty Reduction Strategy Papers.

The Bank and IMF have also suggested other changes to enhance cooperation: sharing more information on work plans and mission schedules, exchanging information between mission teams working in the same coun-