The accompanying table shows BEA's estimates of the total R&D capital stock, and of selected components, together with estimates that others have published. In addition, it shows an alternative set of BEA estimates that is based on 11-percent per year geometric depreciation (rather than straight-line depreciation). The upper panel of the table contains estimates of R&D stocks for selected years. The lower panel shows BEA's estimates less the estimates by others. The comparison should be viewed as rough, because it was necessary to convert most of the other estimates to 1987 dollars from other base periods by using the ratios of BEA's R&D deflator in various base years to its 1987 value of 100. Because of weight shifts over time, the conversion factors yield only approximations of what would be the actual values of rebased deflators.
The alternative BEA stock estimates are not very different from the BEA estimates in the satellite account, and the two series show no tendency to diverge over time. Estimates made by John Kendrick [32] are increasingly higher than the BEA estimates over time. Estimates made by Robert Eisner [24] begin slightly higher than the BEA estimates and become increasingly higher over time. Estimates made by the Office of Management and Budget (OMB)which appeared in the Analytical Perspectives volume of the fiscal year 1995 Budget of the United States Government [40]begin at about the same level and become increasingly higher.
A major reason for these divergences is that BEA's methodology depreciates basic research capital, while the others' methodologies treat it as immortal and do not depreciate it. The rebasing of prices may also explain some of the differences between the Kendrick and Eisner estimates and the BEA estimates. Other differences result from different methodologies for calculating depreciation and from the others' use of NSF estimates of R&D expenditures rather than BEA's estimates; Eisner's use of a 20-year life for other R&D capital also contributed to the differences.
Estimates of the federally financed R&D capital stock made by OMB are increasingly larger than those produced using a rough BEA approximation of BEA's constant-dollar expenditures with geometric depreciation. This divergence reflects OMB's assumption that basic research capital is immortal. In addition, OMB's estimates assume a 10-percent rate of depreciation for other research, somewhat lower than the 11-percent depreciation rate underlying BEA's alternative estimates. Other differences arise because OMB used Federal outlays on a fiscal year basis, whereas BEA's estimates are primarily based on performers' reports of expenditures on a calendar year basis.
Estimates of industry R&D capital stock from a study by the Bureau of Labor Statistics (BLS) [11] are increasingly lower than the corresponding BEA estimates from 1960 to 1965 and are roughly the same amount lower thereafter. The principal reason for the lower values is that the BLS study did not include development expenditures in their capital formation estimates. Other differences arise from the BLS study's assumption that basic research capital is immortal, its lower10 percent per yearrate of depreciation for applied research, its longer gestation lags, and its different method of deflation.
Estimates of industry R&D capital stock made by Nadiri and Prucha [38] are somewhat higher than BEA's corresponding estimates for 1965. Thereafter, their estimated capital stocks grow a little more slowly, on average, and are modestly lower in 1985. The initial difference may be due to the assumed seed value that begins their capital stock estimates. Thereafter, the slower growth reflects a 12-percent per year estimate for the rate of depreciation, somewhat higher than BEA's effective rates of depreciation.