Construction cost indices have often been used to represent changes in house prices because no data on housing transactions, suitable for computing a broad-based index of house prices are available. In this article a technique for isolating price changes from estimates of market value is developed and used to derive a price index for nonfarm one-family houses of all types (excluding mobile homes) in the United States for the years 1947-64. Theoretical and empirical evidence is also presented to show that, in general, an index of construction cost would not reflect house prices accurately.