Predictability of aid: aid recipients and donors alike frequently voice concerns about the lack of predictability of development aid. The existing literature often treats predictability and volatility as closely related, but this paper shows that they are conceptually and empirically distinct. Using two main data sources, we demonstrate that, contrary to common belief, lack of predictability typically involves managing both aid shortfalls and windfalls, and hampers aid management even in countries with stable implementation of macroeconomic policies. Although regression analysis of the sources of low predictability for a large panel picks up two indicators that could be seen as justifying unexpected revisions in aid disbursements, a large unexplained residual remains for which we cannot identify a link between low predictability and aid effectiveness concerns by donors. Using detailed data from IMF programmes, we demonstrate the significant costs of low predictability of budget aid in relatively well performing recipient countries. Deviations of disbursed from expected budget aid of more than 1% of GDP on average are absorbed asymmetrically: aid shortfalls lead to debt accumulation and cuts in investment spending, whereas aid windfalls help reduce debt but also lead to additional government consumption. Lack of predictability thus shifts government spending from investment to consumption activities.