preciation (DEP), and new debt financing (EF). Thus, the budget con-
straint is:
R&D + CE + DIV + LIQ = NI + DEP + EFResearch and development expenditures are modeled in terms of in-
vestments, dividends, and new capital issues (to reflect the imperfect mar-
kets hypothesis) and previous research and development expenditures (to
serve as a surrogate for previous patents and R&D activities). We use
Compustat R&D data as well as NSF/Census R&D data in this chapter.
We also use a three-year lag on the R&D variable, as was done in Guerard,
Bean, and Andrews (1987). There is little difference between the three-year
lag structure and the use of a one-year lag, as was done in Chapter 4. This
result is consistent with the Guerard, Bean, and McCabe (1986) results in
which the authors found no significant differences between using contem-
porary and distributed lag variables of investment, dividends, and R&D.
The investment equation (CE) uses the rate of profit theory (Tinbergen
1939; Dhrymes and Kurz 1967) in which net income positively affects in-
vestment. The accelerator position on investment is also examined through
the two-year growth in sales (DSAL) variable. Depreciation is normally in-
cluded in the investment analysis because depreciation describes the deteri-
oration of capital in the productive process. This study uses cash flows
(CF) to incorporate both net income and depreciation effects; moreover,
other noncash expenditures are included in the firm’s cash flows. The in-
vestment, dividend, and external financing equations used here were dis-
cussed in Chapter 6. The variables are again divided by assets to reduce
heteroscedasticity.
It is expected that the price of common stock (PCS) would be posi-
tively correlated with research and development expenditures, patents,
lagged patents, book value of equity, and investment (Ben-Zion 1984).
Financial Decision Estimation Results
Guerard and McCabe (1992) used a sample of 303 very large U.S. firms
drawn from 12 industries to model a firm’s financial decisions. This
database covered the 1971–1982 time period and was constructed from
Compustat and Patent Office data. Cross-sectional regressions were used
to estimate the model. Industry dummy variables, based on Standard In-
dustrial Code (SIC) classifications, were developed to examine industry