cases it appears that the firm’s actual financial decisions resulted in de-
pressed stock prices in 1982 relative to its asset base. Given its assets, it
was underinvesting in R&D and capital projects, paying lower dividends,
and issuing more new debt than the industry average. However, given its
asset base, the new debt decision was closer to optimal than the other deci-
sions. The most striking contrast between the two cases involves dividend
policy. Clearly, in the case of the Compustat data, stock prices in this in-
dustry are much more dependent on dividend policy than in the NSF/Cen-
sus case. The comparisons are summarized in Table 7.11.
Summary and Conclusions
An examination of R&D data for matched pairs of firms has shown that
Compustat data and NSF/Census data are comparable for a diverse sam-
ple of firms. However, on a year-by-year basis, the NSF/Census data ap-
pear to be significantly different from the Compustat data in 1979 and,
probably, in 1978. In general, it appears that firms overreported R&D ex-
penditures in the Compustat studies relative to NSF/Census studies. This
Summary and Conclusions 197
TABLE 7.10 Sample Firm—Results of Optimization Procedure Using 1982
NSF/Census Data ($ millions)
Firm’s Actual Optimal Levels Under/OverR&D 68 189 $–121
Capital Expenditures 770 882 –112
Dividends 42 139 –97
New Debt 577 540 36
TABLE 7.11 Sample Firm—Optimal 1983 Decisions Based on 1982 Actual and
Industry Levels ($ millions)
1982 1982 Optimal Optimal
Actual Industry 1983 1983
Decision Variables Levels Levels Compustat NSF/Census
R&D 68 178 199 189
Capital Expenditures 769.5 798.5 924.1 881.5
Dividends 42 117 278 139
New Debt 577 231 475 540