Wolff, 1994, for more details. The results specification 5 do show a very strong vintage effect the coefficient of AGEKCHG is negative and significant at the one
percent level. Moreover, the coefficient of the growth in mean education is positive and now significant at the 10 level. Moreover, when initial education is included,
its coefficient, while still negative, is no longer statistically significant results not shown. In the final specification, I included RD intensity, though this variable is
available only for 1960 and later. In this case, the coefficient of the growth in mean education becomes insignificant.
14
5. Interactions with technical change
There is now a voluminous literature supporting the argument that the rate of productivity growth of a country is strongly related to the RD intensity of its
production see, for example, Griliches, 1979, a review of the literature. Moreover, the Arrow and Nelson-Phelps models suggest that there may be interaction effects
between the educational level of the work force and the RD intensity of a country. I introduce the interaction effect into the model as follows:
lnRGDPW
1
RGDPW t
1
− t
= b
+ b
1
RGDPW +
b
2
INVRATE + b
3
EDUC + b
4
RDGNP +
b
5
RDEDUC + o 4
I use two measures for RD intensity. The first is the average ratio of RD expenditure to GNP over the period RDGNP, and the second is the average
number of scientists and engineers engaged in RD per 10 000 of population over the period SCIENG. For the first measure, the coefficient b
4
is usually interpreted as the rate of return to RD.
An interaction term is included between EDUC and RD, because, according to the Arrow and Nelson-Phelps models, a more educated labor force should be more
successful in implementing the fruits of the RD activity. For example, it is frequently argued that the Japanese economy is successful in adapting new technol-
ogy to direct production because of the high level of education of its workforce. In this sense, of two countries with the same RD intensity but different education
levels, the one with the more educated labor force should adopt new technology more quickly and effectively and this should show up in higher measured productiv-
ity growth.
15
This formulation is admittedly crude and specification problems might arise if, for example, the variability in the education variable is low enough to
cancel out the variability in the RD variable. In this case, the interaction term might also show low explanatory power.
Results for all OECD countries over the 1960 – 1990 period are shown in Table 12 note that these results differ somewhat from those of Table 8, whose regressions
14
This set of results remains virtually unchanged even when country dummy variables are included in the various regression equations results not shown.
15
This assumes, of course, that the output of new inventions is directly proportional to RD activity.
465 E
.N .
Wolff Structural
Change and
Economic Dynamics
11 2000
433 –
472
Table 12 Regressions of the annual growth of real GDP per worker RGDPW on initial RGDPW, the investment rate, RD intensity, schooling, and the
interaction between schooling and RD, all OECD countries, 1960–1990
a1
RD EDUCRD R
2
Adjusted R
2
Standard error Samp size
Education INVRATE
Education Relative
variable RDGPW
65
variable 0.67
0.006 23
− 0.021
d
0.100
d
0.57
c
0.71 5.92
3.19 2.42
0.66 0.006
23 SCND-ENRL
65–91
− 0.023
d
0.092
c
0.007 0.58
c
0.72 5.02
2.41 0.61
2.69 0.64
0.006 23
SCND-ENRL
65–91
− 0.021
d
0.098
c
− 0.002
0.02 0.008
0.72 3.61
0.44 0.02
0.09 2.61
0.68 0.006
22 UNIV-ENRL
65–91
− 0.026
d
0.116
d
0.035 0.55
0.74 5.53
2.18 1.43
3.51 0.67
0.006 22
UNIV-ENRL
65–91
− 0.027
d
0.111
d
0.056 0.74
− 0.009
0.75 0.33
4.30 3.03
0.82 1.15
0.66 0.007
21 SCND-ATTN
1970
− 0.018
d
0.103
d
− 0.012
0.67
c
0.73 3.83
2.51 1.02
3.05 0.022
0.77 0.70
0.006 21
SCND-ATTN
1970
− 0.047
b
− 0.015
d
− 0.09
0.135
d
1.58 2.97
3.71 2.08
0.19 0.66
0.006 21
UNIV-ATTN
1970
− 0.019
d
0.097
c
− 0.032
0.59
c
0.73 3.86
2.24 0.91
2.77 0.72
0.006 21
UNIV-ATTN
1970
− 0.015
d
0.129
d
− 0.169
c
− 0.02
0.081 0.79
1.43 3.30
3.68 2.36
0.05 0.67
0.006 23
MEAN-EDUC
1975
− 0.023
d
0.105
d
0.001 0.52
c
0.73 5.91
2.16 1.11
3.34 0.001
0.73 0.66
0.006 23
MEAN-EDUC
1975
0.001 −
0.023
d
0.11 0.110
d
5.31 0.35
0.09 0.22
3.13 0.66
0.006 23
BL-EDUC
1970
− 0.019
d
0.100
d
− 0.001
0.55
c
0.72 3.45
2.29 0.57
3.15 0.001
0.72 0.64
0.006 23
BL-EDUC
1970
− 0.001
− 0.018
d
0.20 0.108
d
2.87 0.67
0.43 0.22
3.01
a
Note: The absolute value of t-ratios are shown in parentheses below the coefficient estimate. A constant term is included in the equation but its coefficient is not shown. The dependent variable is lnRGDPW
90
RGDPW
60
30.
b
significant at the 10 level, two-tail test.
c
significant at the 5 level, two-tail test.
d
significant at the 1 level, two-tail test.
cover the 1950 – 90 period. In specifications without an interaction effect, RDGNP is positive and significant at the 5 level in all cases. Moreover, the coefficients are
generally in the range of 0.50 – 0.60, suggesting extraordinarily high returns to RD investment.
16
However, the interaction term is insignificant in all cases, casting some doubt, at least, on this interpretation of the Arrow and Nelson – Phelps
models.
17
Another striking result is that the coefficient of every educational variables is statistically insignificant in this set of regressions.
Another interpretation of the two models is that an educated labor force might make the adoption andor adaptation of foreign technology easier and thus
expedite the international transfer of technology Richard Nelson suggested this interpretation to me in a private conversation. This argument suggests that there
might exist an interaction effect between the educational level of the workforce and the technology gap, as reflected in the initial productivity level of the country
relative to the U.S. However, using a variety of measures of educational level, I found no case in which this interaction variable proved significant.
6. Concluding remarks