An attempt is made to compare and evaluate polynomial and rational lag models, both theoretically and empirically, by estimating the same lagged relationship using both procedures. Empirical comparison is achieved by eliminating the unknown nature of the lag by using generated data with a known double lag structure.