is not surprising, given the exclusion of engineering and technical services,
as well as research in the behavioral sciences, from NSF’s definitions of
R&D. In the analysis of six specific industries, the R&D expenditures
data were significantly different in only one, petroleum, where the sample
size was relatively small.
When the NSF/Census R&D data were substituted into a series of re-
gression equations used to test for relationships associated with the perfect
markets hypothesis, support for the interdependence of R&D and other fi-
nancial management decisions became progressively weaker as:
The time frame of the data series was compressed.
The diversity of the firms in the sample was reduced.When the NSF/Census R&D data was substituted for the Compustat
data in the more homogeneous sample in the compressed time frame, the
regression results were not altered significantly. Thus, it appears that the
NSF/Census data were equally useful in testing the perfect markets hypoth-
esis within this sample.
Suggestions for Future Research
A larger, more diversified sample of firms should be drawn from the
NSF/Census data set to reexamine the perfect markets hypothesis. The hy-
pothesis is important to national R&D policy concerns because it ad-
dresses the question of wealth creation and its underlying causes. The
larger, more diversified sample drawn from the Compustat data provided
relatively strong support for the concept of simultaneous determination of
the financial decisions within a firm that influence wealth creation. More-
over, the level of R&D funding was an important determinant of wealth
creation. A less diversified sample, covering fewer firms and a shorter time
frame, provided weaker support for the interdependencies among these de-
cisions, regardless of whether NSF/Census data or Compustat data were
used. Since it is well known that technology flows through the economy by
interindustry transfer processes (Scherer 1982), it is necessary to include
technology-using firms that may not perform R&D, as well as technology-
generating (R&D-performing) firms, in studies of the role of R&D in
wealth creation. If the NSF/Census data set does not include non-R&D-
performing firms, it becomes increasingly important to understand the pros
and cons of linking this data set to others that are more representative of
the economy as a whole.