This paper empirically estimates cost functions for two milking technologies, stanchion and parlor, using farm level data from New York dairy farms for the years 1993 through 2002. A translog cost function was estimated along with input cost share equations for each milking technology by Iterative Seemingly Unrelated Regression. Any pair of inputs among feed, hired Labor, and cows had some degree of substitutability except for a pair of feed and hired labor evaluated by the Allen elasticity, and that of hired labor and feed evaluated by the Morishima elasticity. Additionally, economies of scale were found to exist over the entire range of output levels of the samples. The cost of stanchion technology was lower than that of parlor technology over the sample range of output levels of stanchion technology, but because parlor using farms were larger and costs continually decline, parlor using farms eventually experience lower costs than farms milking with stanchions.