Abstract: We ask what specification of preferences can account for the changes in the expenditure shares of broad sectors that are associated with the process of structural transformation in the U.S. since 1947. Following the tradition of the expenditure systems literature, we first calibrate utility function parameters using NIPA data on final consumption expenditure. We find that a Stone-Geary specification fits the data well. While useful, this exercise does not tell the researcher what utility function to use in a model that posits sectoral production functions in value added form. We therefore develop a method to calculate the value added components of consumption categories that are consistent with value added production functions, and use these data to calibrate a utility function over sectoral consumption value added. We find that a Leontief specification fits the data well. Interestingly, the two specifications display very different properties: for final consumption expenditure income effects are the dominant force behind changes in expenditure shares whereas for consumption value added relative price effects are dominant.