THE NONDURABLE CONSUMPTION variable used in Tables 8 and 9 is the annual expenditures on perishables, services, and semi- durables as estimated by Swanson and Williamson ( 1972). For 1929-41, the Swanson and Williamson data is that of the United States Department of Commerce. For 1919-28, Kuznet's ( 1961) data were revised to conform to the 1966 Department of Commerce definitions. The Swanson and Williamson revisions are largely those suggested by Kendrick ( 1961) except that Kendrick used the 1954 Department of Commerce data, whereas Swanson and Williamson used the 1966 Department of Commerce estimates.
Alternative consumption variables were tried to see if the regression results were sensitive to the choice of data. Depreciation on consumer durables was estimated crudely, and a series for nondurable consumption plus use value from durables was constructed. No drastic changes in parameter values (and, therefore, residual patterns) resulted. This alternative consumption variable was abandoned because of the low quality of the estimate for the use value from consumer durables--inadequate data forbade accurate determination. Total expenditures by consumers was tried as the consumption variable with some, often minor, quantitative changes in the regression results. This variable is inappropriate because theory suggests that durable and non- durable expenditure decisions are not made in the same manner. Nevertheless, results using this variable are shown in Tables 8 and 9 for comparison with the conceptually preferable results.
The disposable income data also came from Swanson and Williamson. Their income series is based on computations suggested by