Monte Carlo evidence is presented which supports the conclusion that when a block grants variable in an ordinary least squares (OLS) model is endogenously constructed due to the closed-ended nature of matching aid, estimates of the propensity to spend lump-sum aid are biased upward. The findings suggest that OLS estimates lead to erroneous conclusions regarding the existence of a flypaper effect and support the use of maximum likelihood models in situations where budget constraints are piecewise-linear.