Previous models of residential energy demand have emphasized either (i) short-run continuous adjustments in the utilization of a fixed capital stock or (ii) long-run discrete decisions about the capital embodied in new housing. This paper examines medium-run adjustments to the existing housing stock, focusing on discrete energy conservation "retrofits" such as insulation and storm windows. Individual household data are employed in a two-level nested logit model to estimate a translog indirect utility function. Simulations reveal considerable sensitivity of the demand for retrofits to their own prices, to relative energy prices and to changes in real incomes.