Throughout the US dairy farm industry, observed rates of dairy cow rep
lacement consistently exceed the rates prescribed as optimal by dairy
economists. We attempt to uncover the causes of this discrepancy by es
timating a series of dynamic discrete choice models of dairy cow repla
cement using historical farm-level data. We also advance the methods u
sed to estimate dynamic discrete choice models by demonstrating how or
thogonal polynomial projection methods can be effectively combined wit
h standard maximum likelihood techniques to estimate structural models
of dynamic decision making.