Faced with having to justify programs to offices of management and budget, government agencies generate numbers which describe expected program impacts. But the assumptions or data on which these numbers are based are frequently suspect, as is the utility of relying on counts and modeling techniques for evaluating the achievement of program aims. The result is that agencies often create “concrete fictions,” hard numbers with feet of (soft) clay. Offices of management and budget are able to make their methodology “hegemonic” because agencies usually have to secure their approval to get funding. But imposing this methodology encourages agencies to use research staffs more to defend against the budget office than to help create effective programs, creates differences between the expectations of government and the public, and fosters the overrepresentation of particular interests.