Byline: Vincent Arel-Bundock, James Atkinson, Rachel Augustine Potter How does the institutional design of a state's bureaucracy affect foreign policy? We argue that institutions can moderate bureaucrats' incentives to act in accordance with an Executive's diplomatic preferences. Where the Executive can influence budgets or career paths, bureaucrats face incentives to adopt her diplomatic goals as their own. Where agencies are shielded from Executive influence, bureaucrats are free to act independently in a bid to enhance their autonomy and their reputation for competence. To test these expectations, we develop a new measure of bureaucratic independence for the 15 aid-giving agencies in the US government. We analyze how independence affects foreign aid allocation patterns over the 1999-2010 period. We find that in "dependent" agencies, foreign aid flows track the diplomatic objectives of the president. In "independent" agencies, aid flows appear less responsive to presidential priorities and more responsive to indicators of need in the recipient country. Our results highlight limits on the diplomatic use of foreign aid and emphasize the importance of domestic institutional design. Our findings yield insight into a broad range of policy domains-including international finance, immigration, and the application of economic sanctions-where multiple government agencies are in charge of implementing foreign policy. CAPTION(S): Table S1. Models from Table [Relationsh] with recipient and agency fixed effects. Table S2. Models from Table [Relationsh] with agency and year fixed effects. Table S3. Models from Table [Relationsh] with recipient and year fixed effects. Table S4. Models from Table [Relationsh], but excluding agency outliers. Table S5. Models from Table [Relationsh], but excluding country outliers - Afghanistan, Colombia, Egypt, Iraq, Israel, Israel, Pakistan. Table S6. Probability of obtaining U.S. ODA, and value of ODA receipts in the recipient subsample. Logit and ordinary least squares models. Agency-specific time trends omitted from the table. Table S7. Models from Table [Relationsh], but controlling for all foreign policy motives in each equation. Table S8. Models from Table [Relationsh] with additional control variable for divided government in the U.S. Table S9. Models from Table [Relationsh] with additional control for the occurence of intersate/intrastate intermediate armed conflicts or wars (UCDP). Table S10. Models from Table [Relationsh], but using a binary Independence measure. Table S11. Models from Table [Relationsh], but using an alternative measure of agency Independence from Selin (Forthcoming). Table S12. Alternative measures of development need - income per capita, % of AIDS in the Log (Population), % of Log (Population) under 5$ per day, and life expectancy. Table S13. Models from Table [Relationsh] with alternative measures of U.S. diplomatic interest - Troops stationed in the recipient country, Distance from the U.S., net FDI outflow from the U.S. to the recipient, and bilateral tax treaty. Table S14. Bootstrapped PCA model with method of composition. Table S15. Models from Table [Relationsh], but with all regressors lagged by one year. Figure S1. Distribution of ODA receipts by country and agency (1 of 2). Figure S2. Distribution of ODA receipts by country and agency (2 of 2).