The traditional literature on agency models predicts that, for zero liabili
ty contracts, it is optimal for the principal to pay for the information he
cannot observe. However, this principle is not valid for a set of contract
s mostly used by government agencies whose distinguishing feature is repres
ented by a stringent budget constraint for the principal. This paper shows
that in this environment the principal will either choose a structure exibi
ting pooling or a bargaining solution. The bargaining solution represents t
he analytical proof to the intuition of the difficulty in implementing proc
urement contracts stated by Laffont and Tirole (1993).