use http://fmwww.bc.edu/EC-P/data/rvaluni3.dta
You should set mem 5m when accessing these data. There are 41,702 firm-year observations.
Please contact your GSA for any questions about these datasets.
This is the documentation for the dataset used in B. H. Hall,
"The Stock Market's Valuation of R&D Investment During the
1980s," AER 83 (1993): 259-264, and B. H. Hall, "Industrial
Research During the 1980s: Did the Rate of Return Fall?,"
BPEA Micro (1993): 289-343.
There are 41,702 ohservations on the dataset, containing data
on about 3000 firms in an unbalanced panel from 1959-1991.
Variable definitions:
cusip 6-digit cusip from Compustat (CNUM, more or less)
year two-digit year of data.
sic 4-digit sic from Compustat (DNUM).
sic2d 2-digit SIC category.
gmsmp dummy=1 if this firm was in the original (Griliches-
Mairesse) 157-firm sample.
logv log of market value (debt+equity+prefstk+stassts).
loga log of tangible assets (p&e, inventory, and other).
inva inventory/assets ratio.
otha other/assets ratio.
cfa cash flow/assets ratio.
fyr month of fiscal year end.
rnda rnd/assets ratio.
rsta stock of rnd/assets ratio.
adva advertising/assets ratio.
investa investment/assets ratio.
logs log of sales.
debta longterm debt/assets ratio.
kgam constructed variable (ignore)
kfgam constructed variable (ignore)
netinc net income of the firm (before xtry items).
q Tobin's q; exp(logv-loga). Trimmed at 5 and 95 percentile.
All values are nominal; dollar figures are millions of dollars.
Stocks (including R&D stocks) are beginning of year.
This dataset differs from the NBER-provided data in the construction of the q and sic2d variables.
Please contact your GSA for any questions about these datasets.