The budget and medium-term expenditure framework in Uganda

\ Africa Region Working Paper Series Number 24 0 0 The bu1gett 23439 * Medium-Term Expenditure 0 Framework in Uganda 0 0 0 ' i DvicI L. Bevan 0 X DPecemb~er 2001 0 0 0 0 -0 0 0 0 v 0 0 4 ,P:  The Budget and Medium-Term Expenditure Framework in Uganda December 2001 Africa Region Working Paper Series No. 24 Abstract The Medium Term Expenditure Framework (MTEF) in Uganda has been successful in shifting expenditure composition, most notably in favor of education, as well as protecting priority sectors against cuts. It has been less successful at ensuring that budget allocations translate reliably into actual expenditures. In the wider context of the policies that have successfully delivered macroeconomic stability, these achievements have been impressive and encouraging, but more remains to be done to improve public spending outcomes. Authors' Affiliation and Sponsorship David L. Bevan Department of Economics, University of Oxford E-mail: david.bevan@economics.ox.ac.uk THE WORKING PAPER SERIES The Africa Region Working Paper Series expedites dissemination of applied research and policy studies with potential for improving economic performance and social conditions in Sub-Saharan Africa. The Series publishes papers at preliminary stages to stimulate timely discussion within the Region and among client countries, donors, and the policy research community. The editorial board for the Series consists of representatives from professional Families appointed by the Region's Sector Directors. Editor in charge of the series: Antoine Waldburger, AFTM3, Email: awaldburger@worldbank.org, who may be contacted for hard copies. For additional information visit the Web site http://www.worldbank.org/afr/wps/index.htm, where copies are available in pdf format. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s). They do not necessarily represent the views of the World Bank Group, its Executive Directors, or the countries that they represent and should not be attributed to them.  The Budget and Medium-Term Expenditure Framework in Uganda November 2001 David L. Bevan Department of Economics University of Oxford E-mail: david.bevan@,)economics.ox.ac.uk Fhis paper was prepared as a background paper for the World Bank's first Poverty Reduction Support Credit (PRSC-1) for Uganda. The Credit was managed by Ritva Reinikka. We would like to thank the large number of people who were generous with their time, information and insights. Special thanks go to staff in the Ministry of Finance, Planning, and Economic Development and at the World Bank Resident Mission in Kampala who lirovided feedback on an earlier draft. Geremia Palomba also contributed to an earlier draft of this paper. We are grateful for the Department for International Development (UK) for its generous financial support. The findings. interpretations, and conclusions expressed in this paper are entirely those of the author(s) and do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. Working papers describe research in progress by the author(s) and are published to elicit comments and to further debate.   Contents 1. Introduction 2. Domestic Revenue The Current and Prospective Position The Revenue Profile in the Medium Term 3. Expenditures 3.1 Operation of the Medium-Term Expenditure Framework Aggregate Spending Ex Ante Aggregate Spending Ex Post Budgeted Composition of Spending Achieved Composition of Spending 3.2 The Poverty Action Fund The Share of the PAF in Government Expenditure Future Operation of the PAF 3.3 Decentralization Scale and Type of Fiscal Transfers Organizational Issues 3.4 Arrears 4. The Development Budget and Donor Financing The Financing of the Development Budget Program Support Implementation Rates 5. The Budget in its Macroeconomic Context Different Deficit Concepts Projected Budget Balance during the MTEF The Past Fiscal Year 6. Some General Procedural Issues Institutional Arrangements with Donors Mechanisms within the MTEF Costing and Monitoring Programs Long-Run Issues 7. Conclusions Tables Table 1. Sectoral Composition of Government Expenditure, 1994/95-2002/03 Table 2. Budget Performance, 1997/98-1999/00 Table 3. The Share of PAF Programs in Expenditure, 1998/99-2002/03 Table 4. Fiscal Transfers to Districts, 1994/95-2002/03 Table 5. Conditional and Unconditional Transfers to Districts, 1994/95-2002/03 Table 6. Domestic Stock of Arrears, 1996/97-1999/00 Table 7. Budgeted Financing of Sectoral Development Expenditures, 1999/2000-2002/03 Table 8. Budget/Import Support, 1992/93- 2002/2003. ii   1. Introduction The Government of Uganda's (GOU) approach to budgetary management has evolved steadily during the past decade. The cumulative changes in the institutional framework and in process have improved the macroeconomics of budgetary management and shifted the composition of spending toward social sectors and economic infrastructure. It is now widely appreciated that the GOU has been something of a trailblazer in developing and implementing budgetary institutions. It is important to see that the medium-term expenditure framework (MTEF) represents part of this cumulative development, and does not break with it. Similarly, there is both the scope and the need for considerable future improvement of the process within the MTEF itself. The institutional mechanisms now in place increasingly ensure that resources flow toward areas of policy priority and that aggregate spending is consistent with prudent macroeconomic policy. These mechanisms include the MTEF itself, the device of cash budgeting (the "cash flow"), and the poverty eradication action program (PEAP). The MTEF is intended to provide an ex ante framework to align aggregate resources with program priorities. Cash budgeting provides an ex post mechanism to ensure fiscal prudence in the event of an unanticipated revenue shortfall or a loss of expenditure control in some sector. The PEAP provides a coherent framework for balancing the claims of various sectors against each other and for setting goals. In principle, it also provides for estimating the likely costs of implementation, and for adopting and monitoring indicators of successful implementation. Taken together, this set of instruments offers the appealing prospect of a transparent process for agreeing on policy priorities, aligning budgetary finance in the light of these priorities, and tracking expenditures and their consequences to ensure that expenditures take place as budgeted and yield the intended outcomes. In reality, however, foreign aid-which accounts for the bulk of total resources-complicates the present arrangements because of the way that donors make aid available. These arrangements have implications for the budgeting apparatus of GOU and vice versa. When using the budget as the focal point for deciding priorities and consequent spending allocations, it becomes all the more important to consider all aid flows during the budgetary process, even project aid organized on an extrabudgetary basis. At a minimum, donors should make their intentions clear in advance, ideally on a three-year horizon, as part of the MTEF process, as well as reporting actual spending on a timely basis. There is also a great deal to be said for going beyond this and replacing specific aid with general budget support. One argument in favor of specific funding is that it helps reduce fungibility (when additional funds placed at the disposal of government for one purpose are matched by a transfer in its use of own resources to another purpose); another is that it eases monitoring (making sure funds are spent as intended). The second argument will lose force to the extent I  that the government's own procedures for financial control and auditing are perceived to be adequate. As to the first, fungibility will cease to be a problem if donors feel able to endorse the prioritization embedded in the MTEF, and if the MTEF is perceived as delivering on the planned allocations. All this innovative activity and actual change of substance are reflected in the PEAP and the Budget Framework Paper (BFP) , which together provide a detailed commentary on the strengths and weaknesses of the MTEF at its current stage of development.2 No attempt is made here to replicate or even summarize this discussion, which is both detailed and remarkably even-handed. Rather this paper highlights selective aspects of the present situation, how it has evolved in the past, and how it might evolve in future. Although the developments sketched in the opening paragraph have come to be regarded as exemplary for countries in similar circumstances, a number of problems and issues need to be addressed. Section 2 discusses the level of domestic resource mobilization, both in the short and long run, and the implications for spending. The problem lies not so much in the current relatively low level of domestic revenues, but in the perceived difficulty of raising these levels in future. Section 3 considers the recent history of government spending (recurrent and domestically financed development) and the current plans embodied in the MTEF. More specifically this section looks at how successful the government has been in executing the planned composition, particularly when this involved substantial changes in the shape of the budget (the question of arrears being relevant), and how successful the government was in protecting priority sectors. The past and prospective role of the Poverty Action Fund (PAF) is also considered here, as well as the implications of ongoing decentralization, and the effort to integrate the districts into the budget framework cycle. Section 4 considers the wider development budget and program support and section 5 looks briefly at the budget in its macroeconomic context, focussing on the budget deficit, and especially the domestic component of this deficit. Section 6 outlines a number of procedural issues and section 7 concludes. 2. Domestic Revenue The Current and Prospective Position Domestic revenue as a share of GDP is currently around 11.5 percent.3 This is low when compared with Uganda's own earlier (1960s/1970s) history, and when compared to ' The BFP and the MTEF overlap so closely that they are treated here as two names for the same instrument. 2 An illuminating discussion can be found in E. Tumusiime-Muteble's Uganda's Experience with the Medium-Term Expenditure Framework (1999). 3 This year it has in fact fallen as low as 11.2 percent of GDP, but the government argues plausibly that this was due to special factors. 2  many other Sub-Saharan countries. There has been a substantial recovery from the very low levels associated with earlier dysfunctional governments and civil war, but this has stalled since 1996. In the early 1 990s, following the establishment of the Uganda Revenue Authority (URA) in 1991, the share of revenue in GDP grew by about one percentage point per annum. In the later 1990s, the share has been stationary, but budgets continued to assume growth at one-half of a percentage point per annum. Finally, this year, the government accepted that the budget share is nearly flat.4 In a sense, this represents a shift from a normative approach to the budget (what we would like to see happen) to a positive approach (what we actually think will happen). In the budgeting context, the positive approach is more appropriate. Otherwise, implausible revenue forecasts expand the perceived resource envelope beyond what is likely to be financeable in practice. Unless there are offsetting errors, this leads to a built-in bias in which expenditures have to be curtailed below the budgeted amounts. Since spending cannot be evenly compressed across different components, this is not only inefficient, but is also likely to lead to unplanned shifts in composition. While the normative consideration should not be allowed to disrupt the budgetary process, it remains a serious issue. It is often claimed that governments in poor countries can usefully deploy resources on the order of 20 percent of GDP. This implicit rule of thumb seems now to be accepted as an appropriate target for Uganda. With a revenue share of only 11.5 percent, this leaves a large gap to be financed. Uganda's current approval rating with donors is high, and they seem happy to provide the requisite flow. However, doing this against a background where the gap is being narrowed albeit slowly by continued relative growth in domestic revenue is very different from filing a gap that appears to be stationary and hence indefinite. Consequently, it seems important for the GOU to take a serious look at stagnant revenue5: whether it is desirable to raise it, and if so, how that might be accomplished; if not, how that can be squared with the longer-run issues of size of government operations and aid dependency. The BFP discusses the stagnant revenue issue at length. It argues that the structure of taxes and the associated tax rates are broadly in line with neighboring countries and with best 4 In fact MTEF assumes that it will grow at about 0.2 percent per annum. However, this is based on a more disaggregated view of the likely movement, relative to GDP, of the main tax bases, and not on some unspecified efficiency gain." 5 Since 1996, a number of special factors reduced revenue growth: the disappearance of revenues from the coffee export tax following the end of the coffee boom, the cuts in fuel taxes from 1997, and the cuts in tariffs as part of the trade liberalisation program, also since 1997. Factors working the other way include the extension of VAT to imports in 1996 and the new Income Tax Act in 1998/99. But overall, the trend in growth of revenue from around 1995 shows a severe slow-down. 3  practice. Furthermore, this alignment has been achieved by rate reductions, so it would be undesirable for business confidence to reverse these. 6 At first glance, the argument that the problem does not rest in the design or rate structure of the system seems persuasive. But note that the African neighbors in question obtain very different revenue performances from similar systems: Tanzania is very similar to Uganda, at around 11-12 percent of GDP, whereas Kenya's performance is much stronger, with the tax share averaging roughly twice as much. It is tempting to attribute Kenya's stronger performance to its higher per capita income and more developed manufacturing sector, but recent cross-country analyses of tax revenue performance suggest that Uganda should be capable of generating substantially higher revenues, even when controlling for low per capita income and a disadvantageous economic structure. For example, when we feed Uganda's (1996) characteristics into the tax equations estimated for 39 Sub-Saharan African countries by Ghura, his base regression, which includes only per capita income and economic structure, suggests Uganda should be generating a revenue-to-GDP ratio of over 17 percent.7 Using Ghura's extended regression, which also includes macroeconomic and structural policies (where Uganda scores highly) and corruption (where it does not), it should apparently be capable of generating over 21 percent. These calculations-which should be taken with a grain of salt-suggest that Uganda's relatively poor performance cannot simply be attributed to unfavorable initial conditions. Taking the longer view, the PEAP assumes that the share of URA revenue in GDP will rise from 11.7 percent in 1999/00 to 18 percent in 2019/20 assuming per capita income grows from US$ 254 to US$ 550 (low case) or US$ 700 (high case).8 However, it is worth noting that the growth in the revenue share attributable to this amount of per capita income growth alone would be only 3 percentage points (in Ghura's base regression) and about half that (in Ghura's extended regression). Once again these estimates-treated with considerable caution-indicate that growth alone is unlikely to resolve the revenue problem. There is also the reverse connection to consider, running from taxation to growth. Uganda grew rapidly over the last decade while exhibiting a low share of private investment in GDP. Part of this capacity for high growth with low investment presumably reflects the extent of the recovery from the temporary loss of output associated with the civil war. It would be unwise to assume that this combination can continue over this 20-year horizon. The continued high growth of GDP, on which the GOU is relying, is likely to require a substantial 6 A World Bank study shows that prior to the recent rate reductions (for example in petroleum taxes), the marginal effective tax rate was higher for Ugandan firms than for those in neighboring countries. See D. Chen and R. Reinikka, Business Taxation in a Low-Revenue Economy: A Study on Uganda in Comparison with Neighboring Countries, Africa Region Working Paper Series, No.3, June 1999. 7D. Ghura, Tax Revenue in Sub-Saharan Africa: Effects of Economic Policies and Corruption, IMF Working Paper, WP/98/135. 8 Given an assumed population growth rate of 2.9 percent per annum, these figures imply economic growth at 5.5-6.5 percent and 7-8 percent, respectively. 4  increase in private investment. International evidence suggests that a high tax share is not inimical to rapid growth, but this evidence relates to countries where broadly based taxes at moderate rates raise large revenues. It does not suggest raising revenue share by setting high rates on narrow bases. The alternative avenues, as the BFP acknowledges, are either to reduce long-run spending plans or radically improve tax administration (e.g., raising taxpayer compliance and- reducing corruption). This is likely to prove a long road, but there seems no practical alternative. It will require a very specific program with detailed targets; in effect, a more seriously designed and implemented sequence of "business plans" for the URA. For example, the BFP notes that the problem with VAT collections is neither that the threshold is too high (there are 4,000 eligible firms above the current threshold), nor that the rate (17 percent) is too low. The serious deficiencies are in filing, compliance and collections for the vast majority of firms. Rectifying this is likely to require both resources and a sequential action plan. There may also be a case for more fundamental institutional reform, such as the scheme to contract out the customs administration advocated by the Uganda Manufacturers' Association. Within the URA it will also be important to develop detailed forecasts for collections in the different tax categories (conditional on intended administrative improvements as well as on sectoral forecasts), both as an input into the general budgeting process and as monitorable performance indicators. There may also be a need to make these operations more transparent. Reconstructing a culture of taxpayer compliance will require more than improved administration within URA, however. It will also require a change in the current perception of taxes as onerous, uneven, and arbitrary. To the extent that this perception is objective, further reform in tax design and in tax administration would change it; but to the extent that it is subjective, it will be necessary to improve public information and understanding to help correct it. In this context, the ongoing process of decentralization may offer considerable scope for legitimizing the tax process. There is also scope, as the BFP stresses, for a substantial rise in non-tax revenues from their exceptionally low level (presently 0.2 percent of GDP, compared to Kenya's 3.5 percent). Once again, this will require detailed scrutiny of government activities. Revenue gains will be individually small at first, cumulating to a significant total contribution over time. The long-run spending implications of these unresolved domestic revenue issues can hardly be overstressed. The GOU's present standing as a good macroeconomic manager, reformer of proven durability, and committed developmental state, coupled with its difficult history, means that donors are prepared to underwrite something close to a "normal" size spending program, despite Uganda's relatively modest revenue performance. It is of course possible that they will be prepared to do so indefinitely. However, it would not be wise to plan on this eventuality. More immediate is the risk that a failure to address the revenue issue will 5  begin to tarnish Uganda's manifold other achievements and conceivably lead to an earlier decline in net aid flows than would otherwise occur. This would involve something rather like a "matching grant" phenomenon; an extra shilling of domestic revenue would be accompanied by an enhanced aid flow, and vice versa. This argument is both speculative and (probably) long run, but the risks are nonetheless real. The Revenue Profile in the Medium Term This section discusses considerations relevant to the choice (and attempted implementation) of the revenue profile in the medium term. It utilizes a somewhat idiosyncratic, though quite natural, use of terms, which is spelt out first. What is usually called "revenue effort"9 in the tax literature is referred to as "revenue performance" here, measured by the share of revenue in GDP. The common usage implicitly assumes that the input (which is here called effort) is exactly measured by the output-where a low level of achieved revenue is taken to reflect a proportionately low level of effort to raise that revenue. Without wishing to deny that there is a positive relation between the two, it seems implausible that this relationship should be linear, rather it probably reflects some sort of diminishing returns. We make the plausible assumption, instead, that increased effort would lead to increased revenue, at least over a range, but at a diminishing rate.'0 Devoting an extra 10 percent of resources to collection, or raising rates by 10 percent, would increase revenue, but by less than 10 percent. It also seems plausible, and evidence supports this claim, that diminishing returns set in more sharply if the attempt to raise performance is rapid rather than gradual, at least if this is attempted by rate rises. Conversely, lowering rates, widely perceived as unjustifiably high, can improve collections by increasing the system's legitimacy and hence compliance. However, in Uganda, there has already been a sustained sequence of reforms to the tax system during the 1990s; although there is scope for further design improvements, it does not seem likely that these would of themselves generate substantial gains in revenue, legitimacy, or compliance. Even in the literature, which presupposes a linear relationship between effort and performance, it is widely acknowledged that the slope will differ between countries, reflecting differences in the practical difficulties of collecting taxes. (See the discussion of the paper by Ghura above.) Once these difficulties have been taken into account (typically by regression analysis, using proxies for the various recognized dimensions of "tax-collectability"), a normalized revenue performance can be computed for a country that acknowledges its 9 The discussion is equally valid for the closely similar but somewhat narrower concepts of tax effort and tax performance. B' Beyond that range, the converse will be true. The idea that there is a revenue-maximizing rate for any tax, so that any further rate rise actually reduces revenue, has been made familiar in the concept of the "Laffer curve. 6  idiosyncratic characteristics. Any discrepancy between the normalized and actual performance is then ascribed to differences in its effort relative to the average in the sample. In particular, a low ratio of actual to normalized revenue is taken to imply a proportionately low level of effort. There are two problems with this procedure. First, it assumes that the residual category "effort" is uncorrelated with the tax-collectability parameters. But this seems unlikely, and a- significant correlation would obviously lead to misleading inferences. To illustrate, suppose low-income countries, facing severe difficulties in raising revenues, systematically tried harder than richer ones. Then the coefficient on per capita income would be biased downward; a low-income country whose effort was high relative to the overall average but low relative to the low-income group would be categorized as low effort relative to both. If the correlation had the opposite sign, effort levels would again be biased, and the regression would now exaggerate the benefits of income growth for revenue. Second, the residual category is not really effort at all, but a composite of tax design, tax administration, compliance, and corruption. A low level of "effort" could therefore reflect unduly low tax rates or excessive exemptions; it could reflect lax and inefficient tax administration; it could reflect a culture of non-compliance; or it could reflect a high level of corruption. The latter two difficulties are often treated as two sides of the same coin, but there is a material difference between them. Poor compliance will give a tax administration problems even if it is completely free from corruption. Most tax administrations depend on a high degree of voluntary compliance by taxpayers coupled with a reliable system of enforcement, which itself requires an appropriate culture. The effort calculation-for what it is worth-suggests low effort in Uganda. This poses a choice. It would be possible to discard the calculation, and argue that Uganda's structure and circumstances make it appropriate to settle for a low level of revenue generation for the time being. This argument would have to rest not on the proposition that a country in these circumstances could not raise substantially more-since there are counter examples- but on the proposition that such a country should not raise more, on the grounds that the damage inflicted to private sector activities, income, and growth would be excessive. By extension, other countries in similar circumstances but with higher revenue collections should be encouraged to reduce their revenue effort. The alternative is to accept the calculation, and consider ways of improving performance. Since the structure does not seem to be characterized by unduly low rates, the explanation must lie in one or more of the other components, and the different strategies to address them. Particularly difficult to tackle is a culture of non-compliance. Also, while the perception that revenue performance is poor seems naturally to imply that revenue effort should be increased, it is far from clear how large an increase in effort would be justified. Indeed, it is unclear how far and how fast the revenue-to-GDP ratio could be raised, even if 7  the government was to devote all available energy to raising it. In principle, it would be possible to carry out a cost-benefit analysis, weighing the marginal costs (including those borne by the private sector) of increased efforts against the benefits, but the information for such an analysis is lacking. Given the present circumstances, where donors are prepared to fund Uganda government operations very generously, the pressing issue may be one of absorptive capacity rather than the contribution of domestic revenue. In other words, even given the low level of domestic revenue mobilization, donor supplementation could be carried to the point of severely diminishing returns to government spending. In this context, raising domestic revenue may be a low priority. If it were possible to raise revenue performance reliably and quickly by adopting certain clear cut measures to increase effort, then there would be a strong case for leaving the revenue picture alone to concentrate on other matters until such time as absorptive capacity ceased to be a problem, and/or donor flows began to subside. However, there does not appear to be any quick fix of this type. Prudence suggests that the attempt to improve revenue performance should not be delayed until funding shortfalls have already become an issue. 3. Expenditures 3.1 Operation of the Medium-Term Expenditure Framework The BFP/MTEF was introduced in Uganda in 1992/93 during a time of fiscal crisis. The first objective was to obtain and maintain macroeconomic control and disinflation, and 3 years seemed an appropriate time horizon to achieve this. In these early years attention had shifted to the medium term, but the focus remained on aggregate spending rather than its composition. While some compositional issues were important from the start, the design of comprehensive expenditure allocations by broad sector and by ministry was first seriously addressed in 1995/96. These allocations already reflected objectives embodied in 'priority sectors." More recently, the Poverty Action Fund (PAF) has become a vehicle within the MTEF for targeting earmarked funds into the poverty specific parts of the budget. These arrangements were therefore designed cumulatively to ensure four things: that aggregate spending is consistent with the finance available ex ante; that aggregate spending is held to what can actually be financed ex post; that the budgeted composition of spending reflects a considered view of priorities; and that this composition is implemented in the out-turn, and especially that spending allocations to priority areas are protected in the event of revenue shortfalls or other shocks. 8  How successful was the budget apparatus in delivering on these four objectives? The question becomes particularly pointed as a number of donors have shifted to general budget support and others might be persuaded to do so. Some confidence that the MTEF can deliver on these goals is obviously an important consideration in these decisions. We now briefly consider each objective in turn. Aggregate Spending Ex Ante This matter deals with the accuracy of the revenue and other forecasts used in the forward budget. Forecasts of aid inflows often bedevil budget planning. In spite of the long duration of Uganda's adjustment efforts and the good relations enjoyed with donors, the problems of instability and unpredictability remain. These issues are considered further in section 4. Here we focus on domestic revenue. In fact the relation between the revenue forecast and out-turn has been quite close (figure 1). The two major exceptions in recent years have been 1996/97 and 1999/00.1" In both years the out-turn was well below the forecast, but for somewhat different reasons. Figure 1. Tax Revenue in Uganda: Budget and Out-Turns, 1990/91-1999/00 0 4. 2.0 Source: Ministry of Finance, Planning and Economic Development. In 1996/97, revenue actually grew as a share of GDP, but it did not grow as fast as the overoptimistic forecast (the one percentage point per annum growth in share had begun to decelerate sharply). In 1999/00, when the share was forecast to rise by the more modest half a percentage point, it actually fell sharply, which the authorities attributed to an adverse movement in the terms of trade, an exchange rate depreciation, and a consequent slow-down in imports. Hence 1999/00 compounds a predictable forecasting error (the now mythical "efficiency gains" in tax administration) with an adverse shock. The 1996/97 event is more purely in the forecast error category. There was also a somewhat smaller shortfall in 1997/98. 9  Thus there have clearly been shortcomings in the past role of the MTEF as a predictor of the resource envelope, and to an extent these errors have been systematic. The current conservative view of GOU that the revenue share will be nearly stationary-until there is hard evidence to the contrary-seems an appropriate response to this history. Aggregate Spending Ex Post The two revenue shortfalls noted above posed a similar problem-how to adjust expenditure to accommodate the shortfall-but since they had different characteristics, the appropriate responses also differed. In 1996/97, the shortfall had involved an overoptimistic view of revenue growth: the required revision in spending was permanent, so it was appropriate to cut expenditure one-for-one. In 1999/00, as noted, the revenue shortfall was partly of this permanent type, and partly temporary (the adverse shock). It was therefore appropriate to cut spending less than one-for-one, with some financial accommodation as well. A brief discussion of this point is given in section 5. While the later case is still in progress, all the accumulated evidence shows that the GOU has continued to run a tight fiscal ship so that neither poor ex ante forecasts nor adverse shocks have been allowed to destabilize the macroeconomic impact of the budget. Budgeted Composition of Spending Successive budgets have shown a remarkable switch in the composition of spending toward the social sectors, generally, and those components believed to impact most on poverty reduction, in particular. How did these budgeted shifts get translated into actual expenditures? We proceed immediately to that aspect. Achieved Composition of Spending One of the concerns about the operation of the MTEF a few years ago was that aggregate expenditure control had been achieved at the expense of wide variations between sectoral (and ministry) budgets and actual expenditures. When Moon compared the budgets and out-turns for the years 1995/96 and 1996/97 with the corresponding averages for 1987/88-1994/95,22 he found very marked and durable discrepancies between units that consistently spent over budget (e.g., State House) or under budget (e.g., Agriculture). These large discrepancies could not be explained by movements in aggregate spending. In other words, there was a free for all within the (firmly contained) aggregate, with some units routinely spending resources which had been allocated by the budget process to other units. Thus, until a few years ago, the MTEF was unable to deliver on its compositional prospectus. While this problem has not been fully resolved, tables I and 2 demonstrate that the GOU was able to shift expenditures more or less according to plan. There has been a 12 A. Moon, Uganda's Budget Framework, World Bank, December 1997. 10  substantial and sustained increase in the share going to education; the share going to roads has also risen, but in a more volatile way. The attempted reduction in the share of security spending was initially successful (down from 19.6 percent in 1994/95 to 14.8 percent in 1997/98), but lost ground in 1998/99 (back up to 19.8 percent), and then mostly recovered in 1999/2000 (15.5 percent). The plan is to lower it further, to 13.3 percent in 2002/03. Meanwhile, the share of health seems to have fallen, partly as a matter of unambitious budgeting reinforced by a moderate failure to deliver that budget. This seems to reflect a perception in the central financial agencies that the health sector is less reliable in its use of funds than the education sector. Perhaps more important, it has also reflected the differential dependence of the two sectors on donor finance. The share of health expenditures financed by donors has typically been much higher than that in education (see section 4). Naturally the MoH looked to donors rather than the MFPED for budgetary salvation. However, the increasing primacy of the MTEF as an allocation device is now leading line ministries to exercise greater effort and caution when participating in the budgetary process. The combined share of education and health in the out-turn has risen remarkably from 27.8 percent of total spending in 1994/95 to 32.4 percent of total spending in 1999/00 (and is budgeted to rise further to 36.7 percent by 2002/03), with the largest share for education (19.8, 26.0, and 26.6 percent). Thus the rise in the share of education spending is an accomplished fact, while rise in share of health spending (8.0, 6.3, and 10.0 percent) remains an intention (table 1). Table 1. Sectoral Composition of Government Expenditure, 1994/95-2002103 (wage, nonwage, and domestic development expenditures) (percent of total expenditure) 1994/95 1995/96 1996/97 1997/98 1998/99 1999/00 2000/01 2001/02 2002/03 Out-turn Budget Projected Projected Roads and works 4.4 4.3 6.8 4.9 6.2 8.2 9.2 8.9 8.6 Agriculture 2.6 1.5 1.4 1.1 1.0 1.4 1.6 1.6 1.5 Education 19.8 18.8 22.0 26.0 26.9 26.0 26.8 26.3 26.6 Health 8.0 9.9 7.2 6.6 6.7 6.3 7.6 8.4 10.0 Defense 19.6 18.8 18.4 14.8 19.8 15.5 14.0 13.7 13.3 Law and order 8.9 9.7 8.7 8.9 7.2 7.3 6.3 6.1 5.9 Economic functions 8.7 6.1 6.4 4.6 4.0 5.9 8.7 8.6 8.7 Public administration 20.1 22.3 22.0 25.4 21.3 21.7 18.8 19.2 18.0 Interest payments due 8.0 8.6 7.0 7.6 7.0 7.7 7.1 7.2 7.3 All sectors 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Education and health 27.8 28.6 29.3 32.6 33.5 32.4 34.4 34.7 36.7 Note: Data exclude externally financed development expenditures, arrears, and promissory notes. Source: Data from the Ministry of Finance, Planning and Economic Development (MFPED), 13 October 2000. 11  The case of education in particular demonstrates a capacity to plan and deliver major shifts in the composition of spending, but the data continue to demonstrate difficulties in holding sectors to their budgets. This was especially true for the first two years covered in table 2. For example, in 1997/98, total spending came in close to budget, with a shortfall of less than I percent (table 2). However, the story is much more variable at the sectoral level. Spending for security and defense, education, and law and order were up by a few percentage points, and public administration-a large category-was up by 10.6 percent, whereas spending for roads and works, and agriculture, were down by more than 18 percent, and health about half that. Again in 1998/99, total spending was fairly close to budget, this time up by 4 percent. Spending for security came in 24 percent over budget, education and public administration were also well over budget, while the other sectors were well below. The aggregate picture is very different for 1999/2000, with total spending constrained to fall 4 percent below budget. This reduction was distributed more evenly, with roads, agriculture, education and health all taking the cut in rough proportion, defense being cut rather less, while law and public administration escaped altogether (in fact, they logged increases of 4 and 6 percent, respectively). The only sector to be radically out of line with this general picture is economic and social services, which came in 25 percent below budget. However, this was more due to implementation delays than asymmetric treatment in the process of releases. (See the discussion in section 3.2 on the PAF.) All in all, budget management in 1999/00 seems to have achieved a far closer correspondence between sectoral budgets and out-turns than in previous years. This is all the more remarkable, since it took place during a year of sharp expenditure cuts (unlike the previous two years). Indeed, for much of the year as decisions were being made, it seemed that the reduction in resources would be substantially greater than eventually proved to be the case. This must have made budget management markedly more difficult than the eventual overall cut of 4 percent might suggest. These improvements only partially address the problem stressed by Moon. Sometimes there are good reasons for differences between budgets and out-turns; for instance, they can represent efficient responses to unforeseen contingencies of various kinds. If so, it is important to provide a detailed justification to stakeholders in the process so they can assess whether the discrepancies are indeed justified. At present no systematic reporting on these changes occurs, so the suspicion must remain that they often represent a breakdown of budgetary scrutiny, rather than an efficient response to contingent events. 12  Table 2. Budget Performance: Wage, Nonwage Recurrent and Domestic Development, 1997/98-1999/00 (in billions of shillings) FY 1997/98 FY 1998/99 FY 1999/2000 Variation Variation Variation percent percent percent Budget Out-turn of budget Budget Out-turn of budget Budget Out-turn of budget Security and defense 117.45 120.77 2.83 163.26 203.06 24.38 193.86 191.00 -1.48 Wage 33.64 33.92 0.81 67.80 74.93 10.52 107.85 107.78 -0.06 Nonwage recur 73.81 85.32 15.59 78.26 120.73 54.26 78.90 78.86 -0.05 Develop 10.00 1.54 -84.56 17.20 7.40 -56.98 7.10 4.36 -38.59 Roads and works 48.89 39.99 -18.20 76.64 63.06 -17.73 105.59 100.74 -4.59 Wage 0.85 0.96 14.10 2.39 1.28 -46.31 1.70 1.32 -22.35 Nonwage recur 14.86 11.13 -25.06 24.24 25.01 3.18 28.31 28.79 1.70 Develop 33.19 27.89 -15.96 50.01 36.77 -26.49 75.58 70.63 -6.55 Agriculture 11.34 9.21 -18.77 15.50 9.75 -37.06 18.37 17.75 -3.38 Wage - 1.29 1.46 12.86 4.40 1.40 -68.15 4.34 2.71 -37.56 Nonwage recur 5.99 4.18 -30.23 3.86 3.61 -6.47 5.99 5.93 -1.00 Develop 4.05 3.57 -11.94 7.23 4.74 -34.47 8.04 9.10 13.18 Education 204.53 212.42 3.86 247.41 275.01 11.16 341.47 321.69 -5.79 Wage 130.69 140.61 7.60 158.56 158.06 -0.32 165.52 152.41 -7.92 Nonwage recur 67.13 64.69 -3.63 78.93 81.20 2.88 103.75 104.12 0.36 Develop 6.72 7.11 5.94 9.92 35.75 260.40 72.20 65.16 -9.75 Health 59.89 54.15 -9.58 72.37 68.22 -5.73 82.92 78.32 -5.55 Wage 13.42 12.65 -5.75 24.66 22.63 -8.21 31.93 23.51 -26.37 Nonwage recur 38.62 34.30 -11.19 35.07 35.58 1.48 36.77 38.07 3.54 Develop 7.84 7.20 -8.22 12.65 10.01 -20.89 14.22 16.73 17.66 Law and order 71.39 72.84 2.03 76.05 73.40 -3.49 87.23 90.52 3.77 Wage 29.37 27.62 -5.96 31.26 30.77 -1.59 34.37 34.27 -0.29 Nonwage recur 39.79 42.32 6.34 34.80 36.01 3.46 42.05 48.74 16.00 Develop 2.22 2.90 30.55 9.99 6.63 -33.65 10.81 7.51 -30.53 Economic and social 39.22 37.62 -4.08 43.74 40.76 -6.81 98.14 72.88 -25.74 Wage 3.10 3.38 9.19 3.76 3.21 -14.84 3.46 2.99 -13.58 Nonwage recur 10.39 7.49 -27.94 5.10 4.70 -7.86 13.71 13.14 -4.16 Develop 25.73 26.75 3.96 34.88 32.86 -5.79 80.97 56.74 -29.92 Public administration 188.16 208.01 10.55 200.94 218.10 8.54 252.39 267.55 6.01 Wage 35.56 40.06 12.66 46.31 45.29 -2.20 54.11 52.01 -3.88 Nonwage recur 150.57 163.31 8.46 143.09 162.29 13.42 184.08 203.93 10.78 Develop 2.02 4.63 129.17 11.55 10.52 -8.89 14.21 11.61 -18.30 Interest payments due 76.19 62.41 -18.09 82.37 72.13 -12.44 89.63 94.70 5.66 Contingency: wage 7.93 - 3.61 - 4.10 - Contingency: nonwage - - 2.03 - 7.63 - Contingency: develop - - - - - - Total wage bill 255.86 260.68 1.88 342.76 337.57 -1.51 407.37 377.02 -7.45 Total nonwage recurrent 477.36 475.16 -0.46 487.75 541.26 10.97 590.84 616.28 4.31 Total development 91.77 81.59 -11.09 153.42 144.67 -5.70 283.13 241.85 -14.58 Total expenditures 824.99 817.43 -0.92 983.94 1,023.50 4.02 1,281.33 1,235.14 -3.60 Note: Data exclude externally financed development expenditures, arrears. and promissory notes. Source: Data from the Ministry of Finance, Planning and Economic Development (MFPED), 13 October 2000. 13  In the past, these concerns were exacerbated by the problem of arrears (see table 6 below). Up to the mid-1998s, arrears had stabilized at 30 percent of total spending (i.e., they had been rising in real terms), despite sustained efforts by MFPED to reduce them. This was the clearest evidence of a breakdown in budget discipline. In early 1999, the GOU introduced a commitment control system (initially restricted to nonwage recurrent spending) designed to clear all domestic arrears within three years. This innovation seems to be working across the bulk of spending ministries, as the stock of arrears was down to 17 percent of 1999/2000 spending in June 2000. The discussion has so far concentrated on the functional composition of spending. Table 2 also provides information on how plans relate to outcomes in terms of the economic composition, both within sectors and overall. A somewhat similar conclusion holds here as well. That is to say, the GOU has recently budgeted for major changes in the economic composition of the budget, and has been broadly successful in bringing these about. Specifically, it budgeted to increase the share of domestic development spending (at the expense of the nonwage recurrent budget) from 11-16 percent to 22 percent over the three years 1997/98 to 1999/2000. While the out-turns are somewhat lower (with an implementation rate on the domestic development budget averaging a little under 90 percent, itself an impressive figure by the standards of comparable countries), the intended doubling of share was achieved (10, 14, and 20 percent). Domestically financed development is only one component in the total development budget, however. The larger picture is considered further in section 4. Overall, the story of budget composition seems to be one of considerable, but not wholly reliable, achievement, with signs of real gains in the last year. 3.2 The Poverty Action Fund This subsection examines the evolving scale of operations covered by the PAF, and then turns to some wider considerations concerning its future role. The Share of the PAF in Government Expenditure The Poverty Action Fund (PAF) has become the major vehicle for directing funds into the priority sectors,'3 as table 3 illustrates. Assessing the magnitude of activities in the PAF is, however, far from straightforward. Presentations of budgets and out-turns by the GOU have stressed the element of additionality which the PAF represents. The justification is, presumably, that donors like to see that funds for debt relief, and other resources committed in parallel, are targeted toward social sector activities, and not substituting for resources which GOU would have itself deployed there. Addressing this issue directly would require '1 Indeed, the BFP asserts that in the medium term, the concept of Priority Program Areas will be fully subsumed into the PAF, since the coexistence of the two is confusing. 14  construction of a hypothetical "counterfactual" of what government would actually have done if things were different, but it is not feasible to do this in a way that would carry much conviction. Table 3. The Share of PAF Programs in Sectoral and Total Expenditure, 1998/99'-2002/03 (percent) 1997/98 1998/99 1999/00 2000/01 2001/02 2002/03 Out-turn Out-turn Budget Out-turn Budget Projectiotis Projections Roads and works 21.1 32.0 23.2 24.3 22.1 22.5 22.9 Agriculture 5.9 2.2 33.3 24.3 34.3 34.8 35.4 Education 56.6 61.8 65.1 65.6 68.2 67.0 66.9 Health 8.0 30.4 34.8 27.0 53.5 59.3 67.0 Economic and social 11.5 31.5 39.5 36.7 50.6 57.5 60.9 Public administration 1.8 3.5 5.6 5.1 7.6 8.5 10.5 Total2'3 17.3 22.7 26.2 24.4 30.5 31.0 33.2 Note: 1. The PAF dates from 1998/99. The 1997/98 figures are given for those expenditure categories that were subsequently absorbed in the PAF. 2. The total includes some spending in the "law and order" sector, but this never amounts to more than I percent of that sector's spending so is not reported in the sectoral breakdown. 3. Total expenditure as defined in table 1. Source: Own calculations on data from the Ministry of Finance, Planning and Economic Development, 26 September 2000. We present the PAF using a baseline of 1997/98, the last year before it came on line. Expenditure on a qualifying program is then measured by the excess of current spending over the level achieved in 1997/98. Implicitly, this relies on the counterfactual that, in the absence of the donor funds flowing into the PAF, these expenditures would have been frozen, rather than share in any accretion of general resources available to the GOU. Consequently, the procedure exaggerates the real additionality due to the device. However, in another sense, it underestimates the real scale of activities by discounting the 1997/98 component. Matters are further confused by the fact that the "menu" of activities covered by the PAF is not static, but has tended to expand over time. To avoid further exaggeration of additionality, it is necessary to go back to the 1997/98 figures, and include expenditures on programs not eligible within the PAF initially, but which have become so. Paradoxically, this may have the reverse effect. If the range of PAF eligible programs increases, it is obviously true that its significance has risen, but this is somewhat discounted by the current reporting procedure. In short there is no method of documenting activities under the PAF in a way that permits unbiased assessments of its impact. Table 3 abstracts from these rather metaphysical considerations by reporting total spending on all programs currently within the PAF. This seems the proper measure to concentrate on, namely the total resources reaching favored areas, regardless of whether they are financed by dedicated donor flows or from the GOU's other resources. Out-turns are reported for the three years 1997/98-1999/00; budget figures are given for 1999/00-2000/01, and projections for 2001/02-2002/03. The share of the PAF in total spending rose rapidly from 17 percent in 1997/98 to 24 percent in 1999/00, and is projected to reach 33 percent by 15  2002/03. In other words, the intention is to double the pre-PAF share of spending on these priority areas by the end of the current MTEF horizon. This is a very major and rapid switch in spending pattern, but not out of line with what has been achieved so far. One apparent anomaly is that the PAF out-turn share for 1999/00 was 24.4 percent against a budget share of 26.2 percent. Part of the philosophy for this type of prioritization is to offer the favored sectors greater protection from cuts during spending squeezes, such as the ones experienced last fiscal year. At first sight, therefore, it appears that the ring-fencing provided by the PAF did not succeed in this objective. However, this is incorrect; the shortfalls within the PAF programs were caused by implementation delays, not by lack of available funds. We return to this issue after discussing sectoral composition. Table 3 also gives the sectoral breakdown. In fact, this breakdown should be treated with caution as some of the allocations are rather arbitrary, even if sometimes necessarily so. For example, unconditional grants to districts are allocated to public administration, while equalization grants are allocated to economic and social services. The restocking program is not allocated to agriculture but to economic and social services because it comes under the Prime Minister's Office. In any event, the pattern displayed at the sectoral level is very diverse. The contribution of PAF funds to roads and works is stable (1998/99 apart) at 20-25 percent, both in the recent past and in future projections. PAF contributions to agriculture, health and economic and social services has risen sharply and is projected to go on doing so. Education is a little anomalous, since it shows a very high and relatively stable contribution throughout. This is to some extent an artefact of the recent inclusion of primary education teachers' wages within the PAF, which is a major component of the total education budget. "Backward dating" this category, as is appropriate, raises the figures for all years. We conclude by returning to the implementation issue. For 1999/00, out-turn expenditures under the PAF were only 89 percent of budget (cf 96 percent for the budget as a whole-see table 2); hence the fall in share noted above. However, this shortfall was almost entirely due to a handful of implementation delays. In particular, delays in starting the very large restocking project meant releases only started in the final quarter, and reached only 32 percent of budget. Difficulties in recruiting additional workers in agricultural extension and in primary health care for whom wages had been budgeted meant that only 14 percent of these budgeted wages were paid in the year. For the remainder of expenditures under the PAF, the implementation rate was 99 percent. Hence while the shortfall may signal problems of capacity or project design, it does not indicate a lack of budgeted funds. Future Operation of the PAF The PAF seems to work well in budgeting for growth of spending in the target areas, making the associated finance available, and ensuring that budgeted funds are indeed reserved for these areas. However, it is unclear what the long-run role of the PAF is given that the MTEF is the comprehensive vehicle for budget planning, and since in principle all relevant 16  funds are supposed to pass through the budget, and be monitored accordingly. The PAF owes its origins to the need for a mechanism for implementing the Multilateral Debt Fund, and was extended to handle the flows associated with HIPC (Heavily Indebted Poor Countries Debt Initiative). But it is far from clear that institutional duplication within a consistent framework is desirable. At first sight, it is either redundant, or, if not, risks subverting the integrating features of the MTEF. One argument for its retention might be that it is better able to ring- fence priority expenditures; but that would imply a failure of the MTEF itself, which should be charged with this role. Another rationale for the PAF might be to deny/prevent fungibility within the budget, but that would be odd given the view of the MTEF as fully reflecting overall spending decisions, including those in the priority sectors. There is also the possibility of dissonance between the two mechanisms. For instance, the general budget precept is "use it or lose it." However, the HIPC allocation via PAF to Water appears to have been made not in the belief that it will all be spent this year, but that it will act as a pressure to increase absorption in this sector subsequently. It is unclear whether this means that an unspent balance within PAF can be rolled over for use in the same sector in a later year. There is some discussion in the BFP of such a savings mechanism, justified by the new process for reporting and release of PAF conditional funds, particularly to districts. The argument revolves around incentives, with the GOU anxious that funds should only be released when programs are performing well, but not lost to the program indefinitely in the event of temporary problems. This again raises the issue of the special place of the PAF within the overall budget mechanism. Why should there be special arrangements for monitoring and accountability for funds originating within the PAF? The short answer, presumably, is because there are earmarked funds within PAF for the purpose. While this is all to the good, it does imply differential degrees of scrutiny. If the PAF process is simply thought of as a pilot for arrangements which will subsequently be extended to other sectors, that is fine. But if, as seems all too plausible, it results in a two-tier system-particularly in the context of decentralization and the associated problems of institutional capacity-it may prove confusing and costly to have different systems operating in parallel. 3.3 Decentralization Uganda is in the midst of a shift toward a more decentralized system of government. A common difficulty encountered when implementing decentralization is the lack of capacity at the district and sub-county levels. Such difficult transitional issues are further exacerbated by the shift toward an increased role for budget support discussed elsewhere in this paper. In this subsection, we first look at the scale and type of fiscal transfers to the districts and then at some organizational issues. 17  Scale and Type of Fiscal Transfers The two main features of fiscal transfers are shown in the summary statistics given in tables 4 and 5. The evolving pattern of the share of transfers in total government spending is shown in table 4. As expected, in this context, the share rises strongly over time, from 14 percent in 1994/95 to a projected 34 percent in 2002/03. However, the rise is not steady. It rose rapidly and continuously over the years 1994/95-1997/98, doubling to 28 percent. It then remained at roughly this level for three years (through 1999/2000), and is budgeted to jump to a new plateau around 33 percent this year and for the next two years. Table 4. Fiscal Transfers to Districts, 1994/95-2002/03 1994/95 1995/96 1996/97 1997/98 1998/99 1999/00 2000/01 2001/02 2002/03 Out-turn Out-turn Out-turn Out-turn Out-turn Out-turn Budget Proj. Proj. Billions of Uganda shillings Total expenditures 592.1 632.0 791.3 817.4 1,023.5 1,235.1 1,515.8 1,688.1 1,872.7 of which central government 508.4 515.7 615.0 592.2 734.9 899.7 1,005.1 1,127.4 1,238.9 transfers to districts 83.7 116.3 176.3 225.2 288.6 335.5 510.7 560.8 633.8 As a percentage Total expenditures 100 100 100 100 100 100 100 100 100 of which central government 86 82 78 72 72 73 66 67 66 transfers to districts 14 18 22 28 28 27 34 33 34 Note: Net of arrears, promissory notes and externally financed development. Data refer to releases from the central government to districts and do not fully reflect districts' total resources. Source: Own calculations using data from the Ministry of Finance, Planning and Economic Development, 13 October 2000. Table 5 demonstrates the very heavy dependence within overall transfers on the device of conditional grants. These accounted for 90 percent of the total transfer in 1994/95 and are projected at 87 percent of the total in 2002/03. However, the share has not been stable in the period, falling sharply (to 76 percent) in the initial period of high growth in the total transfer, but then recovering somewhat (to 80 percent) during the first "plateau." They are budgeted to jump back to 85 percent this year, since virtually all the substantial rise in the shilling value of transfers will be in the conditional form. Thus, while the initial increased devolution of funds to the districts saw a rise in the unconditional component, the subsequent continuation has seen the reverse. This has implications to which we now turn. 18  Table 5. Conditional and Unconditional Transfers to Districts, 1994/95-2002/03 1994/95 1995/96 1996/97 1997/98 1998/99 1999/00 2000/01 2001/02 2002/03 Out- Out- Out- Out- Out- Out- turn turn turn turn turn turn Budget Proj. Proj. In billions of Uganda shillings Total transfers to districts 83.7 116.3 176.3 225.2 288.6 335.5 510.7 560.8 648.1 of which unconditional grants 8.4 16.0 37.5 54.3 65.3 68.6 75.6 79.6 83.8 conditional grants 75.3 100.3 138.8 170.9 223.3 270.2 439.0 485.5 569.1 As a percentage of total district transfers Total transfers to districts 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 of which unconditional grants 10.0 13.7 21.3 24.1 22.6 20.4 14.8 14.2 12.9 conditional grants 90.0 86.3 78.7 75.9 77.4 79.6 85.2 85.8 87.1 Note: Net of arrears, promissory notes and externally financed development expenditure. Unconditional grants include equalisation grants. Data refer to releases from central government to districts and do not reflect districts' total resources. Source: Own calculations on data from the Ministry of Finance, Planning and Economic Development, 13/10/2000. Organizational Issues There is a tension between the need to ensure proper monitoring, accountability and control in the local use of centrally provided funds, and the desire to increase local autonomy. In the immediate future, taking due note of the first consideration is bound to impede the implementation of the second. As capacity in the lower tiers is developed, this tension will gradually ease. However, it will not be feasible to develop this capacity uniformly across districts, so, unless more highly developed districts are to be held back, it will probably be necessary to envisage a two (or more) track system, with some districts acquiring effective autonomy sooner than others. The Ministry of Finance now relies heavily on conditional grants to deploy transfers to the lower tiers of government. There are currently 22 conditional grants, 11 falling within the PAF, and 11 outside it.14 Each conditional grant requires a separate bank account, and those within the PAF have separate and specific requirements for the quarterly release of funds. However successful this system proves to be in enforcing accountability, it has the broad effect of reducing local government to the role of an implementing agency for central government. Since the bulk of the funds are routed through the PAF, and since the means for resourcing the PAF (such as the HIPC) lay such heavy stress on monitoring and accountability, there seems little alternative to this type of arrangement in the short term. However, there needs to be careful consideration of how the system can be "liberalized" in the medium term. As noted above, it may very well be necessary to have fast and slow track 14 Conditional grants falling within the PAF account for the lion's share of the total transfer. 19  arrangements, with fast track districts acquiring more autonomy over spending patterns as well as more responsibility for accounting for funds at an earlier stage than others. Given the major role of aid flows in financing precisely those expenditures that are to be the responsibility of local government, this transition must be designed in a way that addresses legitimate donor concerns and generally minimizes donor discomfort. Buying into a mature system of budget support already requires donors to accept some substitution of the host government's procedures in place of their own. Extension to an autonomous decentralized system is a substantial additional step, as it must rely on the adequacy of local government's capacities and procedures. These considerations imply some difficult issues in organizational design. It is important not to confuse appearance with reality. It is all too clear that even tightly specified formal procedures for procurement, reporting and the release of funds accompanying donor- financed projects often failed to deliver appropriate and adequate services. Existing arrangements which embody detailed donor supervision may therefore still fail to ensure true value for the donor's money. Even if there are concerns about procedural adequacy in moving to a system of decentralized budget support, this could not, and should not, be characterized as a move from a safe and efficient existing system, which would needlessly jeopardize satisfactory existing delivery. The balance of risks is indeed a balance. The increase in "informal" accountability associated, for example, with transparency at the local level coupled with local civil society might be more than enough to compensate for some deterioration in "formal" accounting. This is the perspective suggested by the present enthusiasm for outputs and outcomes. However, it may require more than diffuse enthusiasm for donors to dismantle (some of) their existing procedures and buy-into the budgetary process. Improving capacity at the district and sub-county level would be important even if these tiers were to be used only as implementing agencies; and it will be absolutely crucial to the much fuller exercise in devolution currently under way in Uganda. The Ministries of Finance and of Local Government have been very actively engaged with the relevant line ministries in providing regional training workshops and frequent opportunities for dialogue in the preparation of the annual local government budget framework papers (LGBFPs). While the LGBFPs in the second round were still uneven and incomplete, there appears to be real commitment to the process and evidence of a steep learning curve as they move into the third cycle. It is also worth noting that increased transparency seems to account for a huge change for the better in delivering funds to their intended destination. 20  Table 6. Domestic Stock of Arrears, 1996/97-1999/00 1996/97 1997/98 1998/99 1999/00 Stock Flows Stock Flows Stock Flows Stock Flows Stock Jun-96 Incurred Paid Jun-97 Incurred Paid Jun-98 Incurred Paid June-99 Paid June-00 In billions of Uganda Shillings Total arrears 201.1 81.1 48.8 233.4 79.5 73.4 239.5 147.7 126.1 261.2 144.7 211.2 Goods and services 70.7 74.0 45.7 99.0 62.4 53.2 108.2 123.5 80.0 151.7 105.6 110.5 Wages and pensions 14.7 7.1 0.0 21.8 12.0 5.3 28.5 18.2 20.9 25.8 17.0 8.8 Compensation 90.7 0.0 3.1 87.6 4.7 14.8 77.5 2.5 22.1 57.9 21.7 66.6 Subs to Int. Orgs 25.0 0.0 0.0 25.0 0.3 0.0 25.3 3.5 3.1 25.8 0.5 25.3 As a percentage of public expenditure Total arrears 31.8 10.2 6.2 29.5 9.7 9.0 29.3 14.4 12.3 25.5 11.7 17.1 Goods and services 11.2 9.4 5.8 12.5 7.6 6.5 13.2 12.1 7.8 14.8 8.5 8.9 Wages and pensions 2.3 0.9 - 2.8 1.5 0.6 3.5 1.8 2.0 2.5 1.4 0.7 Compensation 14.4 - 0.4 11.1 0.6 1.8 9.5 0.2 2.2 5.7 1.8 5.4 Subs to nt. Orgs 4.0 - - 3.2 0.0 - 3.1 0.3 0.3 2.5 0.0 2.0 Memotandum items Total public expenditure 632.0 791.3 791.3 791.3 817.4 817.4 817.4 1,023.5 1,023.5 1,023.5 1,235.1 1,235.1 Note: 1. These figures are subject to revision, particularly for the later years, which are estimates. 2. Total public expenditure is defined to include wage, nonwage recurrent and domestically financed development expenditure only. Source: Data from Ministry of Finance, Planning and Economic Development (MFPED) at 22 September 2000. 21  3.4 Arrears Several points emerge from table 6, which provides a breakdown of stocks and flows of arrears over 1996/97-1999/00. Prior to the recent reforms, the problem of arrears was serious in magnitude but stationary in a relative sense. That is, the stock in the early years was roughly constant at 30 percent of total expenditure. Of course, this still meant that arrears were rising in real terms, with newly incurred arrears systematically exceeding those paid off in any period. On average, new arrears were being incurred at 11-12 percent of total spending. As the table indicates, the bulk of these (nearly 10 percent of the total, or 15 percent of the nonwage budget) were incurred in purchasing goods and services, rather than in wages. The 1999/00 figures show a major improvement, with these (new arrears on goods and services) falling to less than 3 percent of total spending. The commitment control system responsible for this improvement only has purchase on items in the budgetary process. Some arrears arise in extrabudgetary ways (though of course they have budgetary consequences), in that they involve recognition of liability for compensation for past events, rather than current excess commitments. This is the "compensation" row in the table, showing pension arrears for former employees of the East African Community, which collapsed more than twenty years ago. Unlike the other components, this category of arrears is still rising in real terms (though not as a share of spending). Plans to clear the other categories of arrears over the next three years will be possible even with a falling rate of repayment of past arrears, since the problem has been substantially addressed at source. 4. The Development Budget and Donor Financing In an economy such as Uganda's, the shape of the development budget as a whole may be very different from that of domestically financed development spending. This section examines the relative contributions of these two sources, briefly considers donor program finance, and concludes with an informal comparison of implementation rates. The Financing of the Development Budget Table 7 gives the budgeted/projected financing mix for both the total development budget and the main sectors for 1999/2000-2002/03. (It is not yet possible to examine out- turns for 99/00 for donor projects.) Overall, the shares are stable, with domestically financed expenditures accounting for 35 percent of the total, and donor projects for 65 percent of the total. Given the recent and 22  rapid expansion in the domestic component, this relative stability does not reflect stagnation; rather, it is a consequence of the (roughly) equally rapid expansion in both components. The stability of relative contributions carries down to the sector level. The notable exception is the health sector, where the donor contribution is projected to fall very substantially over the MTEF horizon, and the domestic contribution is projected to rise rather faster than average. The consequence is a startling shift in the projected share, with the domestic component rising from 11 percent in last year's budget to over 46 percent in 2002/03. Table 7. Budgeted Financing of Sectoral Development Expenditures, 1999/2000-2002/03 (billions of Uganda shillings and percent) 1999/2000 2000/01 2001/02 2002/03 Budget Budget Projections Projections Security Domestic 7.1 11.1 11.7 12.2 Donor - - - - Percent domestic 100 100 100 100 Roads and works Domestic 75.6 99.4 104.3 109.6 Donor 127.3 176.9 191.8 207.2 Percent domestic 37.3 36.0 35.2 34.6 Agriculture Domestic 8.0 11.8 12.4 13.0 Donor 54.5 78.3 84.9 91.7 Percent domestic 12.8 13.1 12.7 12.4 Education Domestic 72.2 90.6 99.6 104.3 Donor 31.2 42.1 45.7 49.3 Percent domestic 69.8 68.3 68.5 67.9 Health Domestic 14.2 22.0 28.1 38.8 Donor 114.7 135.1 82.3 44.4 Percent domestic 11.0 14.0 25.4 46.6 Law and order Domestic 10.8 11.3 11.8 12.4 Donor 2.0 4.7 5.I 5.5 Percent domestic 84.4 70.6 69.8 69.3 Economic and social Domestic 81.0 105.6 115.7 128.4 Donor 186.7 319.5 346.3 374.1 Percent domestic 30.3 24.8 25.0 25.6 Public administration Domestic 14.2 20.3 27.9 29.6 Donor 6.0 12.4 13.4 14.5 Percent domestic 70.3 62.1 67.6 67.1 Total development Domestic 283.1 372.1 421.4 463.2 Donor 522.4 779.3 780.6 798.6 Percent domestic 35.1 32.3 35.1 36.7 Note: Total development expenditure includes some items, which are not allocated to sectors. Source: Own calculations on data from the Ministry of Finance, Planning and Economic Development. 13 October 2000. 23  The other feature of note is the very discrepant shares in different sectors. It is no surprise to find limited donor enthusiasm for funding defence, law and order, or public administration, with much greater involvement in roads, agriculture, health and economic and social services. But it is interesting to note that the development budget for education has as high a proportionate reliance on domestic finance as that of public administration, and twice as high as the other "donor favoured" sectors Program Support Because the GOU is heavily dependent on general program support-both in the traditional guise of import support and increasingly in the form of budget support-the regularity and predictability of these very substantial flows is of enhanced importance. Unpredictability through disbursement delays, for example, is also a problem in project finance, but tends to be self-limiting. If finance for the project is delayed then the project is delayed (of course, the causality can run the other way); but the knock-on effect into other budgetary operations is limited. In the case of program finance, fungibility within the budget means that an adverse shock has to be allocated across sectors and "managed," making the whole program potentially vulnerable. Table 8 gives a series for these flows in millions of $US during 1992/93-1998/99 for out-turns and during 2000/01-2002/3 for projections. As is readily apparent, both series are unstable, so that a steady flow of funds can neither be incorporated in the budget, nor (on the basis of past evidence) will such a flow materialize in practice. This has two implications. First, it is highly desirable that donors -achieve smoother flows in future. Second, it is desirable, to the extent they do not, for the GOU to explore-with the donors-ways to insulate expenditure from this type of shock. Table 8. Budget/Import Support, 1992/93- 2002/2003 ($US million) 92/93 93/94 94/95 95/96 96/97 97/98 98/99 99/00 00/01 01/02 02/03 O/Turn O0Turn O0Turn O/Turn O/Turn O/Turn O/Turn O/Turn Proj. Proj. Proj. Grants 111.4 75.5 95.7 86.3 118.8 171.1 113.6 166.0 244.9 214.4 230.3 Loans 83.9 119.2 124.4 49.9 45.9 45.0 123.6 38.5 235.4 170.8 155.6 Total 195.3 194.7 220.1 136.2 164.7 216.1 237.2 204.5 480.3 385.2 385.9 Source: Data from the Ministry of Finance, Planning and Economic Development. 23 September 2000. Implementation Rates The GOU's implementation rate for the domestically financed development budget is nearly 90 percent. The implementation rate for donor projects has also risen from an average of 75 percent during 1993/94-1996/97 to a little under 90 percent during 1997/98-1998/99. The analogous ratio between the level of program finance in the budget and the level received 24  is also fairly similar for 1998/99, at 86 percent. A similar level of realization in the current year would represent a substantial jump in program finance received. From table 8, the projected budget total is $US 480 million, and 86 percent of that, over $US 410 million, is more than twice the realization in 1999/00. 5. The Budget in its Macroeconomic Context This section focuses briefly on one strategic choice and a related presentational matter. It addresses the conceptual issues, illustrating them by making limited reference to the current numbers. Different Deficit Concepts Should budget deficits be measured before or after grants, and which is the appropriate gauge for fiscal prudence? If we accept that grant flows to well-governed countries are pretty durable, then the deficit after grants is the appropriate measure. But this argument is complicated by the fact that soft loans have a large grant component, but are nonetheless booked entirely "below the line." Consequently, a change in the proportionate grant element in a country's loans will alter its fiscal position but not be picked up in the conventional calculation of the deficit after grants. Hence, there would be merit in reporting deficits measures not only before and after actual grants, but also a deficit measure after "augmented" grants (i.e., grants plus the grant element in soft loans). For example, over the current MTEF period, the deficit before grants is projected to average 8.9 percent of GDP, that after grants 2.5 percent. However, the net inflow of loans is projected to average 3.5 percent. If we assume a grant element of 70 percent in these loans'5 then the deficit after augmented grants would be zero (70 percent of 3.5 = 2.5). Another aspect of "which deficit" is appropriate is that the overall deficit needs to be disaggregated into domestic and external components. In effect, the overall deficit is not a very interesting concept: it is the components that matter. What is appropriate for the external deficit depends partly on the soft loan issue raised in the previous paragraph, partly on the HIPC initiative, and partly on donor attitudes. Broadly speaking, it makes sense for a country like Uganda to accept any soft loan that is offered, provided it is sufficiently soft for the rate of interest to be below a conservative forecast of the growth rate, and provided there are no onerous side conditions. In other words, the external deficit-provided it involves soft loans-should be as high as donors will countenance (subject to any practical limits on absorption). What is appropriate for the domestic deficit, on the other hand, depends on a 15 This is probably rather conservative, since it is not only the grant element in new loans that is relevant, but the relationship between this and the grant element in loans in the process of amortization or forgiveness. Thus even if net borrowing were zero (just enough new loans to finance amortization) there could still be a positive "grant element," provided the new loans were on softer terms than the old. 25  detailed review of the money supply, inflation targets, private domestic credit, foreign exchange reserves, and the evolving velocity of circulation. Projected Budget Balance during the MTEF Over the current MTEF period, the domestic budget balance is projected to have a surplus averaging 1.1 percent of GDP. This includes the provision for eliminating arrears noted above. Since these belong to a species of government debt that is not formally accounted for in the government's recorded use of domestic credit, the associated reduction in government's use of the latter falls only by 0.5 percent of GDP each year. In effect, the scope for total domestic credit expansion consistent with prudent monetary expansion (given anticipated real growth of 7 percent and an inflation target of 5 percent) and maintenance of foreign exchange reserves (at 5 months of imports) is not only being made available to the private sector in its entirety, it is being augmented by these government savings. As noted earlier, continued growth is probably going to require substantial increases in private investment and it is important to ensure that adequate credit is available.. Even so,, it would be useful to study the evolution of private credit closely to ensure that provision on this scale is necessary and likely to be prudently absorbed. If not, there is a case for a somewhat more relaxed domestic fiscal stance, perhaps involving a balanced domestic budget, or even a small deficit. Total expenditure is projected to average 20.7 percent of GDP, with external finance accounting for nearly half (10 percent). It might be that this level of expenditure is the most that government can efficiently absorb. But if not, and assuming that Uganda has reached the funding boundary of donors, then there may still be limited scope for a further spending increase financed in this way. This is certainly not an argument for relaxed vigilance on the inflation front, nor for faster monetary expansion. It is simply an argument to ensure that the scope for non-inflationary expansion of domestic credit is used effectively, whether by private or public agents. The Past Fiscal Year It is one thing to plan ahead on the basis of best projections, but of courseevents may unroll differently. As noted, the current year is an acute example of this, with revenue collections down 12 percent on budget. The substantial temporary component in this adverse out-turn made it inappropriate to cut spending programs to the same extent. In the event,, private sector borrowing was also well below what had been projected, in part, presumably, because of the same events that lowered revenue, such as low import levels. The latter also reduced the level of foreign exchange reserves consistent with five months cover. Hence, borrowing from domestic banks was used to partly offset the negative revenue shock. This required a relaxation in the Poverty Reduction and Growth Facility Program, with which the IMF acquiesced. Consequently, the GOU was able to reduce the depth of the expenditure cuts and, in particular, fully protect the priority sectors. Aggregate spending as a share of GDP 26  actually rose markedly relative to 1998/99, since there was a substantial increase in the net flow of aid. 6. Some General Procedural Issues Out of a wide variety of ongoing procedural issues, this section highlights four. Institutional Arrangements with Donors It is important for bilateral donors to increasingly accept the MTEF as the organizing principle for their interactions with government. They will doubtless continue to have manifold direct dealings in their aid programs, but they should see these as being framed by the MTEF. Ideally, more donors will sign up to the MTEF as the actual vehicle through which their aid is disbursed. To stand any chance of this happening, regular meetings will be needed to allow, in a reverse sense, donors to share "ownership" of the process. These should probably be quarterly, and involve discussion of substantive, ongoing issues, rather than being limited to dissemination of what has already been decided elsewhere. The obvious existing vehicle for this role would be the macroeconomic working group. However, it should have a much wider function than simply bringing relevant donors " on board" the ongoing MTEF process. It has previously been possible for both donors and to an extent government to abdicate from choosing the macroeconomic stance of the budget. In the past, the relative non-negotiability of the IMF's position-often fully justified by a lack of real room for macroeconomic maneuvering-has preempted the macro policy debate. In Uganda's case, this has ceased to be valid for some years, given the cumulative payoff from a consistent policy of fiscal prudence. The new element in the picture is the IMF's shift toward recognizing the desirability of greater flexibility in fiscal targets coupled with greater national "ownership" of policy. This confers increased scope but also greater responsibility on government to examine the options and choose between them. While the primary dialogue on fiscal matters will remain one between government and the Fund, it should also involve other donors, especially to the extent that they commit to budget support or otherwise embrace the MTEF as the organizing principle. Increasingly, they will need to become comfortable not only with issues of budget composition, as previously, but also with the stance of fiscal policy. For example, if the Fund is no longer insisting on a particular number for the budget deficit, so that this is not pre-empted from consideration by anyone else, it becomes a matter for choice and dialogue involving the weighing of alternatives, so donors need to take an intelligent interest in it. There is also the question of how best to integrate the activities of donors wedded to traditional mechanisms of project finance. In principle, this should not present insuperable difficulties provided two conditions are met. The first relates to informational flows. For an integrated budget process to work, it is vital that the government not only obtains detailed ex 27  post information on spending, but also that it is able to factor planned project expenditures into the process at a sufficiently early stage.'6 The crucial point is that the MTEF, coupled with the PEAP, should decide on program priorities and the associated composition of spending. Within that framework, the principle of fungibility should rule. So a donor could directly finance a set of projects provided they were already (or potentially) "within the system." What would be unsatisfactory would be for project finance de facto to reshape the budget allocation. There remains the problem, of course, that separate project finance is also likely to impose wasteful duplication of procedures. Once again, the macroeconomic working group could provide an appropriate forum within which reporting could take place. These meetings could be analogous to the quarterly meetings of the PAF, which offer a means of regular reporting and monitoring. The difference would be that in the present case, donors would be reporting on their previously extrabudgetary operations, and these would be progressively brought more completely on budget. Mechanisms within the MTEF The potentially uncomfortable role of the PAF within the MTEF has already been noted, particularly in the context of extensive decentralization and the desirability of organizing a streamlined and unitary set of procedures. The PAF may have had a useful role in providing a sort of visible semi-guarantee that funds (such as HIPC) would indeed be spent in the ways intended: indeed, the presentation of the PAF accounts goes out of its way to stress additionality. However, given the presence of fungibility in government spending, the only real guarantee that the funds at government's disposal will be spent in ways that are comfortable to donors is for an overarching device like the MTEF to stipulate the composition overall. As long as the composition elaborated in the MTEF is acceptable, and that composition is duly delivered, the problem is solved. There is also a more general difficulty with the notion of priority areas in general, whether or not they are embedded in a separate funding mechanism. It seems that there could be three rather different types of rationale for delineating priority areas. One would be in recognition that there was a disagreement between donors and government on the desired composition of spending. Another would be if government itself felt that the composition of spending was inappropriate, but that it would take time to reshape it. In both cases, a properly articulated MTEF seems to be the better device, whereby a compromise is reached, or a trajectory determined. The third rationale is that a priority sector is one that should be protected against cuts. But this protection could be required in two rather different circumstances. One is where 16 In theory, this objective should be met by the process of annually updating the PIP. However, that hope has not been generally realized. 28  politically powerful sectors seek to undo the previously agreed budget allocation, at the expense of the weaker. As we have seen, there is some evidence of this happening in Uganda. A clear commitment to protect the weaker sectors may then be valuable as a supplement to the budget process, but it is a sign of weakness in that process all the same. The other circumstance is where the need for cuts originates in revenue shortfalls. But the argument then would be to protect sectors where interruption would be particularly costly, rather than sectors that were in some abstract sense "more deserving." There seems to be some confusion as to why priority sectors are so classified, and further discussion of this might clarify what would be the best budgetary response. It is obviously insufficient to assert that they are just more important, since if that were generally agreed, and there were no problems of adjustment, then their importance would be rewarded by a bigger relative allocation, until they ceased to be more important, at the margin. Costing and Monitoring Programs Chapter 7 of the Poverty Eradication Action Program provides detailed initial information on costing and implementation. The bottom line is that the current 2000/01 budget provision is around 63 percent of the PEAP requirement for all programs taken together. However it is only around 54 percent for the priority areas (roads, agriculture, education, health, water and law), as opposed to 83 percent for other MTEF programs. On the face of it, this has two implications. The first is that a further shift in budget composition toward the priority sectors is indicated. The second is that the affordability of the PEAP is going to depend on rapid real growth in resources. Much effort has gone into choosing indicators, but until the monitoring process is seriously under way it will be difficult to judge their ability to deliver clear signals that can be interpreted and acted upon. In this respect, Uganda suffers the disadvantage of being a pioneer Long-Run Issues It is well-known that separating the budget into development and recurrent components can lead to a mismatch between the two, with investments being undertaken that cannot subsequently be operated at their design capacity. It is also well understood that donor- financed projects have often exacerbated this, adding a second problem to the first. There is growing support for the idea that the adoption of an MTEF within an integrated Ministry (such as MFPED) will resolve the first problem: and that growth of program support, in general, and devices for budgetary support (such as the PERC), in particular, will resolve the second. However, the capital/recurrent problem is more difficult to resolve than these views imply because of the lengths of the relevant time horizons, of which there are two. The first is the horizon over which the cumulative investment program will take place. For example, if a 29  program to build enough classrooms is to be consistent with a given (enhanced) secondary school enrolment target, over how many years will the first, catch-up phase of the program run? Of course, population growth will require further investment subsequently, in a second phase, but at a reduced proportional rate, so the issue of budget sustainability is likely to be a problem of the first rather than the second phase. The second horizon is given by the lifetime of the component investments (the classrooms in the previous example). The recurrent cost consequences of a program stretch over the sum of these two horizons, which will be many multiples of the three-year MTEF horizon. We are not suggesting that the MTEF horizon be extended, but rather that the MTEF be supported by longer term, necessarily speculative, calculations. Since it is natural to assume that the expenditure-to-GDP ratio will be fairly stable in the long run, however it is financed, the key consideration on the resources side is the growth of GDP. The implicit cumulative expenditure profile of any program would have to be set alongside this. DFID recently commissioned some work that looked at this in the Uganda context, focusing on public sector pay reform. A wider set of investigations could usefully be undertaken by sector working groups. 7. Conclusions The MTEF and ancillary policymaking instruments in Uganda have been very successful in delivering macroeconomic stability in a fairly turbulent environment. It has also been successful in steering expenditure composition toward social sectors and economic infrastructure, most notably in favor of education, and in protecting priority sectors against cuts. Of course, this means that these have been concentrated on other sectors, which may also be damaging. Ensuring that budget allocations translate reliably into actual expenditures remains problematic, whether or not there has been a shortfall in available resources in aggregate. This problem has been exacerbated by that of arrears, which the GOU has recently taken steps to control. The cumulative achievement to date has been impressive, but evidently more remains to be done. Issues that need further attention: The MTEF needs to better reflect the macroeconomic framework within which budgetary decisions are located, including alternative profiles for domestic and donor resources, as well as financing options, interrelations with the domestic private sector, and issues of absorptive capacity. Also needed is a more systematic consideration of how to manage volatile resource flows and other shocks. In part this means moving beyond the familiar, tried and tested mechanism of the cash flow toward a more sophisticated but still prudent fiscal strategy. A strategic view needs to include an appropriate and practicable profile for domestic revenue in the medium term, and a plan for implementation. 3()  The long-term goals of the PEAP and the associated costs of achieving them need to be more actively integrated with the MTEF and associated budget allocations to ensure that budgets accurately reflect priorities. If they do not, then programs of transition between current and desired allocations shou]d be elaborated. The relation between budgetary allocations and out-turns seems recently to have been markedly tightened, but still remains imperfect. It is currently difficult for outside - observers (for example donors or civil society) to assess whether these discrepancies are efficient responses to unforeseen contingencies or whether they reflect poor budget management. A systematic system of reporting the nature of these discrepancies needs to be put in place to explain the rationale and, where appropriate, the steps taken to rectify the problem. Something of this sort already takes place within the PAF quarterly meetings. On the monitoring/evaluation/output/outcome fronts, it will be important to maintain the momentum already achieved to ensure that expenditures are not only under control, but also usefully deployed. To the extent that the low-revenue collection reflects a problem of compliance, the perceived legitimacy of taxes can only be enhanced if (i) the revenues are believed to translate into budgeted expenditures and (ii) these expenditures are perceived to be useful. Although the PAF has attracted donor funds, helped reallocate the budget toward priority sectors, and improved budget procedures, it sits uncomfortably within the MTEF and could be phased out, while perhaps extending its best practice, over, say, a three-year horizon. Also care must be taken to ensure that the PRSC itself does not add to these intrabudgetary anomalies. Conditional grants to local government may be a suitable short-term expedient to deal with capacity problems, but are inimical to the intention of achieving a more devolved system. Active consideration should be given to how to progress from here, perhaps at two or more speeds, depending on local capacity. Donors must move toward fuller integration of donor financed projects into the budgeting process, both ex ante in the MTEF process, and ex post in tracking out-turns. They also need to provide greater stability and predictability in donor flows, not least for program support. 31  I  Africa Region Working Paper Series ARWPS 1 Progress in Public Expenditure Management in Africa: Evidence from World Bank Surveys, January 1999, by Christos Kostopoulos ARWPS 2 Toward Inclusive and Sustainable Development in the Democratic Republic of the Congo, March 1999, by Markus Kostner (out of stock) ARWPS 3 Business Taxation in a Low-Revenue Economy: A Study on Uganda in Comparison with Neighboring Countries, June 1999, by Ritva Reinikka and Duanjie Chen ARWPS 4 Pensions and Social Security in Sub-Saharan Africa: Issues and Options, October 1999, by Luca Barbone and Luis-Alvaro Sanchez B. (out of stock) ARWPS 5 Forest Taxes, Government Revenues and the Sustainable Exploitation of Tropical Forests, January 2000, by Luca Barbone and Juan Zalduendo ARWPS 6 The Cost of Doing Business: Firms' Experience with Corruption in Uganda, June 2000, by Jacob Svensson (out of stock) ARWPS 7 On the Recent Trade Performance of Sub-Saharan African Countries: Cause for Hope or More of the Same, August 2000, by Francis Ng and Alexander J. Yeats ARWPS 8 Foreign Direct Investment in Africa: Old Tales and New Evidence, November 2000, by Miria Pigato ARWPS 9 The Macro Implications of HIV/AIDS in South Africa: A Preliminary Assessment, November 2000, by Channing Arndt and Jeffrey D. Lewis ARWPS 10 Revisiting Growth and Convergence: is Africa Catching Up?, December 2000, by Charalambos G. Tsangarides All Papers can be accessed via the internet at http://www.worldbank.org/afr/wps/index.htm  ARWPS 11 Spending on Safety Nets for the Poor: How Much, for How Many? The Case of Malawi, January 2001, by William James Smith ARWPS 12 Tourism in Africa, February 2001, by lain T. Christie, Doreen E. Crompton ARWPS 13 Conflict Diamonds, February 2001, by Louis Goreux ARWPS 14 Reform and Opportunitv: The Changing Role and Patterns of Trade in South Africa and SADC, March 2001, by Jeffrey D. Lewis ARWPS 15 The Foreign Direct Investment Environment in Africa, March 2001, by Miria Pigato ARWPS 16 Choice of Exchange Rate RegimesforDeveloping Countries, April 2001, by Fahrettin Yagci ARWPS 17 Export Processing Zones: Has Africa Missed the Boat? Not yet!, May 2001, by Peter L. Watson ARWPS 18 Rural Infrastructure in Africa: Policy Directions, June 2001, by Robert Fishbein. ARWPS 19 Changes in Poverty in Madagascar: 1993-1999, July 2001, by Stefano Paternostro, Jean Razafindravonona, and David Stifel. ARWPS 20 Information and Communication Technology, Poverty, and Development in sub- Saharan Africa and South Asia, August 2001, by Miria Pigato. ARWPS 21 Handling Hierarchy in Decentralized Settings: Governance Underpinnings of School Performance in Tikur Inchini, West Shewa Zone, Oromia Region, September 2001, by Navin Girishankar, Abebaw Alemayehu and Yusuf Ahrnad. ARWPS 22 Child Malnutrition in Ethiopia: Can Maternal Knowledge Augment The Role of Income?, October 2001, by Luc Christiaensen and Harold Alderman. ARWPS 23 Child Soldiers: Preventing,Demobilizing and Reintegrating, November 2001, by Beth Verhey. All Papers can be accessed via the internet at http:flwww.worldbank.org/afr/wps/index.htm   Headquarters 1818 H Street, N.W. Washington, D.C. 20433, USA Telephone: (202) 477-1234 Facsimile: (202) 477-6391 RCA 248423 WORLDBK Cable Addr.: INTBAFRAD WAHINGTONDC Working Paper Series web address: http://www.worldbank.org/afr/wps/index.htm European Office 66, avenue d'1ena 75116 Paris, France Telephone: (1) 40.69.30.00 Facsimile: (1) 40.69.30.66 Telex: 640651 Tokyo Office Fukoku Seimei Building, 10th Floor 2-2-2 Uchisaiwai-Cho Chiyoda-Ku, Tokyo 100-0011 Japan Telephone: (81-3) 3597-6650 Facsimile: (81-3) 3597-6695 Telex: 26838 The World Bank