Milk supply response was estimated for Pennsylvania using three different levels of structural aggregation. The base level involved the estimation of milk production in a single equation. Under the second method, production was the product of two equations: milk per cow and number of milk cows. The third method factored production into three equations: milk per cow, number of dairy farms, and number of cows per farm. As expected, the greater the degree of disaggregation the more was learned about the structural aspects of milk production. At the same time, predictive accuracy generally decreased, but the differences among models was slight.