Should donors who are interested in the effectiveness of developmental programs rely on conditional budget support or on project aid? To answer this question, we present a model in which only a subset of the developmental expenditures can be subject to conditionality. We show that budget support is preferable to project aid when donors and recipients' preferences are aligned, and when assistance is small relative to recipients' resources. Then, we test our model estimating a modified growth model for a panel of developing countries, and find evidence in support of our predictions.

Should donors interested in the effectiveness of developmental programs rely on conditional budget support or on project aid? To answer this question, we present a model in which only a subset of developmental expenditures can be subject to conditionality. We show that budget support is preferable to project aid when the preferences of donors and recipients are aligned, and when assistance is small relative to the recipient's own resources. Furthermore, when donors cannot observe the recipient's type, they may impose higher levels of conditionality to separate committed governments from uncommitted ones.

We develop a theoretical model to compare the two major foreign aid modalities: project aid and budget support. These two modalities have a different impact on the production of 'developmental goods'. Firstly, conditionality can be associated with budget support, but only a subset of the developmental expenses - the observable ones - can be subject to conditionality. Secondly, when using project aid, the donors control the overall allocation of the aid resources. However, we consider that, because of limited harmonisation and coordination, project aid can be associated with a cost of imperfect fit. We develop a unified framework to compare these two modalities where we allow the simultaneous utilisation of both instruments. We show that all the aid should be given via budget support, no matter whether conditionality is used or not. Furthermore, we show that the optimal use of conditionality depends on the recipient's developmental preferences, the productivity of the inputs and the level of aid compared to the recipient's budget: when these parameters are relatively high, conditionality should be enforced. Otherwise, the optimal aid allocation is such that all the aid is given through unconditional budget support. We conclude that conditionality does not always improve the aid effectiveness.

Assessing Aid determines that the effectiveness of aid is not decided by the amount received but rather the institutional and policy environment into which it is accepted. It examines how development assistance can be more effective at reducing global poverty and gives five mainrecommendations for making aid more effective: targeting financial aid to poor countries with good policies and strong economic management; providing policy-based aid to demonstrated reformers; using simpler instruments to transfer resources to countries with sound management; focusing projects oncreating and transmitting knowledge and capacity; and rethinking the internal incentives of aid agencies.

This paper considers approaches towards improving the predictability of aid to low income countries, with a special focus on budget support. In order to accelerate progress towards the Millennium Development Goals, the donor community is increasing aid flows while pushing for more coordination and tighter performance-based selectivity. However, these factors may increase the unpredictability of aid from current levels, which are already high enough to impose significant costs. Predictability is a particular challenge in the area of budget support, which will continue to increase in importance as aid is sought to underpin longer-term recurrent spending commitments. Budget support reduces transactions costs and drains on capacity, but it tends to be more vulnerable to fluctuations than multi-year project support. Poor predictability raises the threat of a low-level equilibrium: countries, budgeting prudently within a medium-term fiscal framework, will discount commitments; donors will see few funding gaps, so pledges will fall. With some countries discounting aid commitments in formulating budgets, some already see signs of this happening. To improve predictability, donors must extend their funding horizons. However, even if this can be done, several major issues will remain at country level. First, how can countries deal with residual short-run volatility of disbursements relative to commitments? Second, can donors lengthen commitment horizons to individual developing countries without excessive risk of misallocating aid? Third, within a country's overall aid envelope, how should donors set the shares of project aid and budget support? Finally, the paper considers the other main approach to budget support, the output or outcome-driven approach of the European Union. The paper concludes that many of these issues can be addressed. Simple spending and savings rules built around a buffer reserve fund of 2-4 months of imports can help smooth public spending. Aid can be pre-committed several years ahead with only small efficiency losses, using a strategy of "flexible pre-commitment." Guidelines can be set to limit the volatility of budget support while keeping it performance-based, and past experience can be used more systematically to develop "outcome" norms to better guide aid allocation.