You are on page 1of 29

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/264999671

The Effect of R&D Capital on Firm Productivity

Article

CITATIONS READS

0 56

2 authors:

Ebbe Krogh Graversen Michael Spejlkavik Mark


Aarhus University Nordic Institute for Studies in Innovation, Research and Education
35 PUBLICATIONS   281 CITATIONS    14 PUBLICATIONS   56 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Sustainable path creation for innovative value chains for organic waste products (SusValueWaste) View project

Return on investment for workplace training View project

All content following this page was uploaded by Ebbe Krogh Graversen on 20 November 2014.

The user has requested enhancement of the downloaded file.


The Danish Centre for Studies in
Research and Research Policy
University of Aarhus Finlandsgade 4 DK-8200 Århus N.
Phone (+45) 8942 2394 Fax (+45) 8942 2399 www.cfa.au.dk

The Effect of R&D Capital on Firm Productivity

By Ebbe Krogh Graversen and Michael Mark

Working paper from


The Danish Centre for Studies in
Research and Research Policy
2005/3

Abstract:
The present report analyses the importance of R&D for Danish private firm
productivity. The aim is to analyse whether R&D capital increases the firms’ value
added, i.e. productivity. The analyses are the most complete and covering so far for
the Danish private sector and give results that can be generally used for research
and business policy purposes. The analyses support and confirm results from
previous studies and show a clear effect on firms’ value added from their R&D
activities. It is found that the value added per employee is 40 percent larger among
R&D active firms than among R&D inactive firms and that the value added increases
1 percent for each 10 percent increase in R&D capital among the R&D active firms.
The authors have performed the analyses on a fully representative and complete firm
specific R&D database from CFA merged by firm specific economic and personnel
data from Statistics Denmark.1

1
The paper builds on results reported in Danish in Graversen and Mark (2004) in a report requested by
the Danish Ministry of Science, Technology and Innovation. Eventual misinterpretations and other
discrepancies are naturally the authors’ own responsibility.
Table of Contents

1. Introduction ............................................................................................................. 3
2. Theoretical and empirical modelling ....................................................................... 4
2.1 The production function.................................................................................... 5
2.2 Empirical firm specific data for 2001 ................................................................ 5
2.3.1 Different R&D capital measures ................................................................. 6
3. Firm characteristics................................................................................................. 7
3.1 Value added in 2001 ........................................................................................ 9
4. Results on productivity and R&D capital............................................................... 11
4.1 The main model ............................................................................................. 12
4.1.1 Sensitivity to definitions of R&D capital .................................................... 13
4.2 Interaction effects........................................................................................... 14
5. Conclusion ............................................................................................................ 18
5.1 Research policy implications.......................................................................... 19
References ............................................................................................................... 19
Appendix 1 Construction of R&D Capital measures ................................................. 22
Appendix 2 Definitions and additional empirical results............................................ 24

2
1. Introduction
Private sector firms are today competing on markets with high international
competition. Denmark as well as the other West European countries experience an
increasing outsourcing of production to Eastern Europe or Asia, where especially the
wage costs are lower and where the supply of well educated and other demanded
labour is large and increasing. Hence, sufficient value added in Danish firms is a key
element for their employment and competitivity on the medium sight, i.e. their
survival as firms located in Denmark. Increased value added can be obtained by
reduced costs, i.e. basically wage competition, or by unique and advanced
knowledge intensive innovations, i.e. competition on complexity and innovativeness
in the production.

The experiences from the last couple of decades on outsourcing of production to


cheaper countries have shown that wage competition is not the strategy to follow.
Conversely, the experiences show that employees that become unemployed when
production is moved relatively quickly find other jobs and often jobs with a higher
knowledge demand. This dynamism in the labour market secures a continuous
renewal of the production and services in the national labour market, so the new jobs
are in areas where Danish firms are competitive.

Considerable and sufficient innovative knowledge imbedded in the firms’ production


and services demands a basis of basic new knowledge through R&D activities that
can be spread and used broadly in all firms’ innovative activities. Hence, R&D
supports and benefits the entire business sectors innovation, such that new
knowledge intensive production and services secures the employment on sight.

In the present report it is tested whether R&D and other forms of innovation together
with knowledge and the ability to use knowledge, i.e. absorption capacity, in Danish
firms improve their value added, i.e. productivity. The potential for economic growth
is often explained by the ability to be innovative and dynamic. Firms investing in R&D
or more broadly in different forms for innovation are assumed to generate more
economic growth than comparable firms. The community return from investments in
R&D is similarly theoretically much larger than the firm specific return since the
generated knowledge typically can be shared and used by many firms; even if it is
protected by patents or kept secret in the short run. The community return from
investments in R&D is also generally found to be much larger than the firm specific

3
return, c.f. for example Smith (2002) for a discussion.2 In the literature, return above
20 percent at the firm level and 50 percent at the community level is referred. Hence,
the community has a large interest in supporting the frames and conditions for
private sector R&D.3

The report identifies the return to R&D capital and the R&D related factors that
influence growth in the firms. The firms in the analysis are from the data material
collected for the official Danish R&D Statistic 2001 from The Danish Centre for
Studies in Research and Research Policy, c.f. AFSK (2003). The data set is the most
comprehensive and representative sample of Danish firms’ R&D activities so far
used in this form for analysis’. The data material for 2001 includes more than 2200
firms and represents broadly all Danish private sector firms with more than 9
employees.

2. Theoretical and empirical modelling


The analysis is based on a modified model for firm production in form of a Cobb-
Douglas production function in the original form specified in Solow (1956) and
modified as in for example Hall and Mairesse (1995) and Smith et al (1999, 2000)
among others. Hence, present value of firms accumulated R&D investments together
with interaction effects are included directly in the production function, such that the
influence from R&D on production is directly testable in the empirical modelling.4
Similarly, a number of firms specific R&D related and general characteristics is used
to identify observable differences in return to and level of firms’ value added and
productivity. The empirical model is estimated sequentially to test whether the
parameter estimate to the R&D capital is robust, and the model is estimated for all
firms as well as for R&D active firms solely.

2
Olesen-Larsen (2003) criticise these results and the conclusions based on them to be less general for
the Danish private sector. The analyses are mainly foreign and performed on specific subsamples of the
private sector and typically of older date. Never the less it can be assumed that at least some of the
results are globally usable due to the theory and demand of equal return of investments in a global
market with (almost) free capital movements.
3
The difference between the social or community return and the firm specific return is the maximum
that the community logically can support the firms’ R&D activities.
4
Firms’ accumulated R&D capital and R&D personnel are included in their assets and number of
employees. To prevent double counting the R&D capital is subtracted from the assets and the R&D
personnel is subtracted from the number of employees in the production function. Both Hall and
Mairesse (1995), Wakelin (2001) and Dilling-Hansen et al. (1999) point at the problem of double
counting in the innovation indicators in the model as an important problem that changes the results
significantly. Double counting is for example when a researcher counts both under labour as an
employee and under R&D capital through his salary. See also Schankerman (1981) and Cuneo and
Mairesse (1984) for the first explorations of this specific bias problem in the empirical approach.

4
2.1 The production function
Initially a production function as defined in Solow (1956) is set up and estimated. In a
logarithmic version of this Cobb-Douglas production function, where production is
estimated as a function of R&D capital, other assets and labour, the estimation
coefficient to R&D capital equals the output elasticity of R&D capital. The production
or output is in the analysis the firm specific value added. A split of the assets and
labour in a part that is caused by R&D and a part that is not modifies the production
function. Furthermore, level differences and interaction variables are included in the
model in order to improve the explanatory power and to test for these differences.5
Hence, the empirical modeling of the interaction effects with R&D capital on
production equals a number of other models in similar analyses; cf. Hall and
Mairesse (1995), Dilling-Hansen et al (1999), Smith et al (2000) among others6. Hall
and Mairesse (1995) conclude that among several different empirical specifications,
the direct productions function approach is preferable.

2.2 Empirical firm specific data for 2001


The primary data set in the analyses is the survey data collected to the private sector
R&D Statistics 2001, c.f. AFSK (2003). Additional information comes from the
corresponding material from 1991, 1993, 1997, 1998 and 1999, c.f. descriptions in
AFSK (2003). The data material includes survey data on private sector firms’ R&D
and other innovation activities in the respective years. The initial data set consists of
the firms that are included in the 2001 data collection and it is weighted to represent
the firm population of 18381 firms in 2001. The data set is the most comprehensive
and representative sample of Danish firms’ R&D activities used so far for this type of
analyses. The data represents broadly all Danish private sector firms with more than
9 employees.7

5
The interaction variables are for example business sector indicators, size indicators etc. Especially
business sector indicators are of interest in this study since it is the first study where firms from all
private business sectors are included. Usually only firms from the industry sectors or part of it have
been used.
6
The modeling of growth rates are usually set up in an extended standard production function written
as Y=AKαNβCγeε and in logarithmic form y=a+αk+βk+γn+ε, where y measures output, n measures
labour input, k measures assets, and c measures the R&D capital. This modelling is the basis for a
number of empirical studies of the relation between innovation and productivity. The studies uses the
production model with various adjustments in definition of output and specification of how innovation
influences the production, cf. Hall and Mairesse (1995), Singh and Trieu (1996), Dilling-Hansen et al.
(1999), Smith et al. (2000), Wakelin (2001), Criscuolo and Haskel (2002), Rouvinen (2002), Medda et
al. (2003).
7
A weight variable makes the firms representative for more than 18000 firms, approximately
corresponding to all private sector firms in Denmark with 2 or more employees in the IT, R&D and other
knowledge service business sectors, 50 or more employees in the primary industry, retail trade and
lawyers’ business sectors and 6 or more employees in the remaining business sectors, c.f. AFSK
(2003).

5
These figures correspond to an original data set on 2465 unweighted firms. The R&D
data has been merged with economic data and information on employees in the
firms using firm specific data from Statistics Denmark. It was possible to match 2278
firms with valid values in the firm statistics at Statistics Denmark. The missing
matches are caused by differences in firm unities in the registers used to collect R&D
statistics and economic information. The firm units in the R&D statistics are identified
as firms delivering publicly available accounts to a governmental firm department
while the value added is calculated from access restricted accounts made for tax
payment purposes. The firm units in the two registers may not match exactly, since
they, by law, have different information purposes. A few of the firms could not be
found in the official registers at Statistics Denmark at all. Furthermore, 50 firms had
none or a negative value added, so the data set usable for the analyses was
reduced to 2228 firms. The 2228 firms represent 2929 firms with a positive R&D
capital and 14025 firms without.8 The present value of their R&D capital is found as
described in appendix 2 based on the data material for all the previous years.

A number of variables in the survey data have empty or missing values. Some of
them are substituted by logical values such as 0 on number of R&D employees if the
firm is R&D inactive. Other empty values are caused by missing or inconsistent
answers in the survey. The missing values have been estimated by the use of other
forms of information in the statistics or other sources. For example that a firm has a
positive R&D capital if it earlier but not any longer has R&D costs.

As mentioned above, the firm specific data for 2001 have been matched with
information on value added, employee growth and employee education level among
others at Statistics Denmark. According to the rules for anonymity this causes that
the final data set exclusively is placed at Statistics Denmark and that the analyses
are performed at the research facilities at Statistics Denmark.

2.3.1 Different R&D capital measures


The firm specific R&D capital is constructed in the same way as most studies in the
area. However, in order to test the sensitivity in the results with the used definition
and construction of the R&D capital, six other R&D capital measures have been
constructed and used in the estimations. Basically the R&D capital measures are
constructed from the firms’ R&D costs in the past. The R&D costs are accumulated,
deflated and depreciated as described in details in appendix 1.

As done in Hall and Oriani (2003) we also uses depreciation rates on 15 percent and
accumulates over 5 and 10 years respectively, K5_15t and K10_15t. We also use a

8
The data represents 3099 and 15282 firms with and without R&D capital in the original 2001 sample.

6
20 percent depreciation rate as in Smith et al (1999, 2000 and 2001) and Dilling-
Hansen (1999), K5_20t and K10_20t. Furthermore, we use different depreciation
rates for different R&D cost types, cf. table A.1, K5at and K10at. Finally, we use the
results from Bayer (2003) with a depreciation rate over three years and an immediate
depreciation on 40 percent in period 0, c.f. Bayer (2003).

As the averages of the R&D capital for the firms presented in table A.2 shows, the
differences are very small. Naturally the R&D capital average is larger when the
accumulation period is 10 years compared to 5 years but only around 5-10 percent.
Similarly, the average is larger when the depreciation rate is 15 percent instead of 20
percent, but again only 10-15 percent. The detailed construction of the R&D capital
with different depreciation rates for different R&D cost types do not deviate much. It
gives basically the same average values as the method with a common 15 percent
depreciation rate. Hence, the average R&D capital fluctuates less than 10 percent,
except for the capital measure constructed as in Bayer (2003), which is one-third the
size of the others.

As shown in section 4.1.1 the estimation results using the different capital measures
are also more or less equal as earlier found in Hall and Mairesse (1995). Hence, a
short conclusion is that it does not matter for the model results whether R&D capital
is accumulated over 5 or 10 years, whether a 15 or 20 percent depreciation rate is
used or whether or not R&D cost type specific depreciation rates is used. Hall and
Mairesse (1995) conclude that the longer the accumulation period the better is the
quality and robustness of the results but that the choice of depreciation rate does not
matter much.

3. Firm characteristics
This section gives descriptive statistics on the R&D active and inactive firms included
in the analysis. One of six firms were R&D active in 2001, cf. table 1. One out of four
was innovative. One-third of the R&D active firms had their own permanent R&D
department. Naturally, the shares deviate over business sectors and firm size
groups. For example are more than half of the firms in medico and more than half of
the large firms R&D active in 2001. These firms have also more often research-
educated employees employed, cf. table 1.9

9
Research-educated employees are defined as employees with a Ph.D. degree or something similar.

7
Table 1 Danish private sector firms in 2001
R&D Research
active; Otherwise Own R&D educated Share of Share of
Business sector 2001 innovative department employees firms employees
Industry ------------ Share of firms in percent ------------ -- Percent --
Low tech 11 18 3 0 17 19
Medium-low tech 13 13 5 1 11 9
Medico 55 10 29 24 1 3
ICT industry 21 17 15 2 4 4
Other high and
35 12 16 3 9 12
medium-high tech
Service
ICT Service 36 13 16 6 6 7
Other knowledge
17 15 7 3 17 13
service
Other sectors 6 10 2 0 35 32
Firm size
Under 50 employees 12 12 5 1 84 25
50-99 employees 20 18 7 2 9 13
100-249 employees 33 20 18 5 5 16
Over 249 employees 53 19 33 20 3 46
All firms 15 13 6 2 100 100
Note: Number of unweighted observations is maximum 2228.

There is a large consistence between the R&D active firms in 2001 and the firms with
a positive R&D capital in 2001. The latter group is a bit larger as it can be seen in
table 4 since a few firms not R&D active in 2001 have been R&D active in at least
one of the ten previous years. Hence, there is almost no difference in firm
characteristics in the two overlapping groups and there will not be distinguished
between them in the following. Hence, the term R&D active firms cover both groups
in the remaining sections. This also means that differences that relates to large
business sector and firm size can be found in tables 2 and 3. It is clear that there is a
large difference in the levels between the sectors and firm groups, which indicate a
need for these variables to be included in the production model estimation.

8
Table 2 Firms with R&D capital; numbers, percent and 1000 DKK
R&D
current R&D
R&D R&D costs/ Capital/ Share
R&D personnel costs/ R&D R&D of Share of
Business sector personnel share turnover personnel personnel firms employees
Industry ---- Average conditional on positive R&D capital --- -- Percent --
Low tech 5 3 1 269 1895 13 22
Medium-low tech 5 6 2 281 1889 11 11
Medico 71 32 15 514 2465 3 7
ICT industry 24 23 7 412 2109 7 6
Other high and
13 16 6 352 2314 20 24
medium-high tech
Service
ICT Service 13 35 25 564 1916 16 12
Other knowledge
13 49 22 302 1464 19 9
service
Other sectors 3 9 2 334 1201 11 9
Firm size
Under 50
4 29 13 314 1126 65 10
employees
50-99 employees 8 11 6 351 1343 13 8
100-249 employees 16 10 8 554 2100 13 18
Over 249
74 9 4 519 2415 9 65
employees
All firms 12 23 11 364 1996 100 100
Note: Number of unweighted observations is maximum 662.

3.1 Value added in 2001


The concept of value added is used to measure production in the present analyses.
The gross value added, i.e. production minus input in the production, is used.10 The
basic results in this section show that R&D active firms have a 40 percent larger
value added per employee than R&D inactive firms. The R&D active firms have close
to 50 percent of the total value added in the private sector although they only
constitute 17 percent of all firms included in the data.

The average value added per employee is found and given in table 3. Differences in
average values over business sectors and firm size groups indicate level differences
between these groups. Similarly do differences in value added per employee among
R&D active and inactive firms indicate whether the R&D active firms have a relative
higher average value added. The numbers in table 3 do not show whether firms have
a positive return from R&D capital, only whether they have a higher value added per
employee. This can be caused by other factors such as differences in capital
intensity between business sectors, differences in employee education levels etc.

10
Hence, neither salaries nor use of real capital is subtracted from the production measure.

9
As the numbers in table 3 shows, there are considerable differences telling that the
medico and ICT business sectors and the large firms have the highest value added
per employee. The R&D active firms have in all sectors and groups a higher value
added per employee than among the corresponding R&D inactive firms. Generally it
indicates that R&D capital correlates positively with the firms’ value added per
employee.

Table 3 Average value added per employee in 2001 by business sector and
firm size, 1000 DKK and shares in percent

R&D inactive
R&D active firms All firms
firms
Value Differ Sha R&D Value
Value added added per ence re capital per added per
1) 2)
Business sector per employee employee employee employee
Industry
Low tech 361 << 477 32 14 38,5 416
Medium-low tech 382 < 412 8 17 50,0 397
Medico 353 <<< 922 161 64 554,5 909
ICT industry 423 << 546 29 29 341,6 490
Other high and medium- 397 < 434 9 39 178,7 427
high tech
Service
ICT Service 600 < 706 18 43 222,3 676
Other knowledge service 325 <<< 486 50 19 292,5 372
Other sectors 366 < 425 16 6 41,7 373
Firm size
Under 50 employees 424 < 448 6 13 216,0 428
50-99 employees 374 < 437 17 26 112,3 390
100-249 employees 355 < 420 18 47 147,7 385
Over 249 employees 319 <<< 564 77 60 176,8 461
All firms 371 << 517 39 17 170,6 431
Note: Number of unweighted observations is maximum 2228.1) Percent higher average value added per
employee at R&D active firms. 2) Share of firms that have a positive R&D capital.

The relative small difference in value added per employee among the small firms
with and without R&D capital is caused by relatively many ICT firms with a high value
added per employee increasing the average for R&D inactive firms in this group.
Similarly, it seems to matter for the ability to increase value added that firms are
small (and young) when they do not have R&D activities to support the growth. The
correlation between value added per employee and firm size is small and
insignificantly negative for R&D inactive firms but positive although insignificant for
R&D active firms. For these firms there is instead a significant positive correlation on
0.4 between value added per employee and R&D capital per employee. Hence, the
basic statistics supports the results in earlier studies in for example the DISKO-
project, finding that there is several types of firms; both small and large passive firms

10
with low value added per employee, a group of small firms on their way up with a
large value added per employee and a group of large firms with a large activity and a
large value added per employee.

Less than 4 percent of the employees in the R&D inactive firms are highly educated,
while more than 9 percent in the R&D active firms were highly educated.11 The
overall average share is 6 percent highly educated among the employees in the
firms.

4. Results on productivity and R&D capital


The influence of R&D capital on productivity is empirically tested on the firms
included in the analyses. The estimation results, the variation in the results and their
robustness is presented in this section. The major results are that the return to R&D
capital is positive, large and significant and that the return to R&D capital increases
when the firms have research educated employees. The results support a long list of
earlier empirical results finding a positive correlation between R&D investments and
productivity growth, cf. for example Griliches (1998) , The Economic Council (1997),
Smith et al. (2000, 2001), Smith (2001, 2002) and Salter and Martin (1999).

Section 4.1 shows results from estimation of the main model and section 4.1.1
shortly refers the robustness in the results from estimation of the model using
different R&D capital measure. Finally section 4.2 refers results from estimation of
models with interaction and level effects. Generally, the parameter estimates are
very robust towards changes in the model specification. 12 The parameter estimates
to R&D capital varies very little when the different R&D capital measures is used, cf.
section 4.1.1 and tables A.5 and A.6. However, there is a difference when the model
is estimated on the industry sector solely. The parameter estimate of R&D capital is
smaller for the industry sector than for the service sector when the model is
estimated for all firms in the two sectors separately. The difference almost
disappears and become insignificant when the model is estimated on R&D active
firms alone. A relatively larger share of R&D active firms in the service sector and the
fact that R&D active firms in general have a larger productivity primarily cause the
difference. Another general difference is that the parameter estimate of R&D capital
increases by a factor five when the model is estimated on the R&D active firms
alone.

11
Highly educated are defined as employees with a bachelor degree or more.
12
A truncation of the data set that removes the upper and lower five percent of the firms sorted by
value added per employee does not change the results in any significant way. Hence there is no
indication on extreme values or firms in the data set. Similarly, the model is estimated in a version
where the R&D cost (investments) in 2001 is subtracted from the R&D capital in order to test whether
this less clear causality between cost and productivity in the same period influence the results. This is
not the case, so it is not pursued further in this paper.

11
4.1 The main model
In the calculation of the R&D capital the present value of the firms’ salaries to R&D
personnel is included. For year 2001 it is also included in the R&D capital, which
results in a double counting of R&D personnel through the variable labour and the
R&D capital variable, cf. model 2 in table 4. Hence the labour variable is fully
reduced by the R&D personnel, cf. model 3 in table 4.13 In the analysis where the
R&D personnel is a separate variable also the salary costs for 2001 is removed from
the R&D capital in order to prevent a similar double counting, cf. model 4 in table 4.

Table 4 below shows the estimation results from a sequential extension of the
standard production function. It is estimated for all firms and for the R&D active firms
respectively.14 The model 1 is the standard model with a single dummy variable for
the firm being R&D active or not. The extension in the remaining models 2, 3 and 4 is
shortly explained above. Model 3 is the model that is used in the further analysis in
sections 4.1.1 and 4.2.

From table 4 it can be seen that the coefficients to labour and other assets are very
stable and robust. In column 1 under all firms the coefficient to the dummy variable
measures the productivity difference between R&D active and inactive firms; here
the coefficient means that the R&D active firms have a 20 percent higher productivity
on average than R&D inactive firms. The results also show that the correction for
double counting between model 2 and 3 only increases the coefficient to the R&D
capital slightly. Including labour reveals a high significance of the R&D personnel
factor. In the case of all firms, the inclusion does not change the coefficient to the
R&D capital, but in the case of R&D active firms alone, the coefficient to the R&D
capital is reduced by a factor five. Hence, when R&D personnel is seen as a short
run R&D indicator and the R&D capital measure in model 4 as a long run R&D
indicator, there seems to be a large short run effect of R&D in the firm productivity
and a smaller although still highly significant long run effect of R&D. In model 3 the
R&D capital measures both the short and long run effect.

13
Except in the Bayer case where R&D capital is measured as described in appendix 1. Here, only the
share that is not immediately depreciated reduces the labour variable.
14
Similar model as is estimated in table 4 is given in table A.3 in appendix 2 with the only difference
that assets are exchange with fixed assets only. The results deviate somewhat but the major
conclusions are the same.

12
Table 4 Estimation of the production function15
All firms R&D active firms
Model
11) 22) Basis3) 44) 11) 22) Basis3) 44)

Constant 9.308 9.366 9.664 9.813 8.294 8.102 8.846 10.390


Labour 0.626 0.625 0.644 0.642 0.504 0.517 0.573 0.585
R&D labour - - - 0.226 - - - 0.209
Assets 0.289 0.286 0.264 0.256 0.384 0.314 0.242 0.234
R&D capital - 0.016 0.018 0.019 - 0.094 0.114 0.023
Indicator for
0.205 - - - - - - -
R&D capital > 0
Number of firms 2228 2228 2228 2228 662 662 662 662
R2 0.90 0.90 0.89 0.90 0.91 0.92 0.92 0.92
Note: All variables are measured in logarithms except the ”indicator variable for R&D”. R&D capital is
accumulated over ten years and depreciated by 20 percent per year. Parameters in bold is significant
on a 1% level. 1) Simple model with an indicator for R&D capital>0. 2) Incl. R&D capital, but without
3)
correction for double counting. Incl. R&D Capital and correction for double counting in assets. This
model is the basis model is Section 4. 4) Incl. R&D capital and R&D personnel and correction for double
counting in both assets and labour.

The coefficients to the R&D capital in model 3 in table 4 can be converted to return
by multiplying it with value added stock and divide it with the R&D capital stock. This
gives a return rate of R&D capital of 11 percent among all firms and 34 percent
among the R&D active firms alone. These return rates are considerably higher than
the calculated return to other assets, especially for the R&D active firms.

4.1.1 Sensitivity to definitions of R&D capital


As it is described in appendix 1, seven different R&D capital measures were
constructed. Results from estimating the basis model with these different R&D
capital measures is shown in tables A.5 and A.6 in appendix 2 for all firms and for
R&D active firms respectively. In order to make the results more independent of the
business sector differences, interaction effects among these are included. Basically,
the results show that the coefficient to R&D capital is more or less independent of
how the R&D capital is constructed, cf. table 5, although there is a level difference
depending on the accumulation horizon for the R&D active firms, cf. table A.6. The
coefficient to the R&D capital constructed as in Bayer (2003) halves the coefficient,
but only for the R&D active firms in table A.6.

15
The results show the effect of R&D capital for all firms and R&D active firms solely. Hence the
coefficient tells the percent change in productivity (value added) caused by a one percent change in
R&D capital in a random firm (out of all firms) or in an R&D active firm (out of the R&D active firms
solely). The results show that the effect of increased R&D capital is larger when it is given to an R&D
active firm than when it is given randomly.
The parameter coefficient is also an elasticity coefficient. Hence, return to R&D capital is calculated as
the elasticity coefficient multiplied by the value added and divided by the R&D capital stock.

13
4.2 Interaction effects
Tables 5 and A.4 in appendix 2 show results from estimation of the basis model for
R&D active and all firms respectively with level effects and R&D interaction effects
included. The estimated models in table 5 and A.4 are equivalent and the only
difference between them is the sample size, where the model in table 5 is estimated
for R&D active firms solely and the models in table A.4 is estimated for all firms in the
sample.

Commenting all the results is overwhelming so the following will resume the main
results, differences as well as similarities and the corresponding implications. There
are several of the models in table 5 where the interaction coefficient is positive and
the corresponding level effect is negative. Basically, this just means a higher payoff
per R&D capital unit to the firms with this characteristics and not a lower level for the
same firms, since the level tells the crossing at zero R&D capital where no R&D
active firms is situated. Most R&D active firms have an R&D capital far above zero
and the level coefficient cannot be treated as a level difference between firms having
the specific characteristic and firms not having it. Similarly, in the estimation based
on all firms in table A.4 the same holds. This time there is R&D inactive firms having
zero R&D capital and the level effect partly covers their level together with the
technical level difference between R&D active firms having or not having the specific
characteristic. Hence, in the following we will focus on the interaction effects.

As model 1 and 2 (column 1 and 2) in tables 5 and A.4 shows, neither a business
sector nor a size effect exist among the R&D active firms. There is a higher effect on
the residual group called other sectors, but this is caused by a few atypical firms in
this small group of R&D active firms. The level coefficient for the same group is also
negative. Similarly, a negative level effect is found among the small firms in column
two in table 5, indicating an increasing productivity level with firm size. There are no
interaction effect differences across firm size. However, among all firms in table A.4,
there is a higher coefficient to R&D capital in the medico sector, in the other
knowledge service sector and among small firms. This means that increased R&D
capital in these sectors have an above average increase in productivity in general.
The results in column 1 also indicate that productivity is higher in the service sector
while the results in column 2 indicates a clear ranking towards higher productivity the
larger the firms are.

The differences in results between the model estimated on R&D active firms solely
and all firms continues in the following models. There is no additional influence from
foreign ownership, R&D department and employing researchers (columns 3 and 4)
among the R&D active firms, but among all firms there is a significant additional
effect supporting earlier results that R&D active firms are more productive and
relative more often foreign owned. Having an R&D department and employing

14
researchers are highly correlated. The result concerning R&D department and
employing researchers seems robust since it is similar to the result in column 9
where there is controlled for business sector level differences.

Similarly, firms having external financing of R&D activities have a higher coefficient
to R&D capital, cf. column 5. Only among all firms the additional coefficient for public
external financing is significant and negative. This may be explained looking at the
results in column 6 where applied R&D and for all firms also to development R&D
increases the coefficient to the R&D capital. Hence, there seems to be a correlation
between external public finance and basic R&D where the productivity effect of
additional R&D capital is lower. These results are also similar to the result in column
12 where there is controlled for business sector level differences.

The most impressive results are found in columns 7 and 8 (similar to the results in
columns 10 and 11) where the share of R&D personnel and the share of researchers
are included as interaction variables. The share of R&D personnel reduces the
common coefficient to R&D capital by 50 percent for the R&D active in table 5 and to
an insignificant coefficient for all firms in table A.4. Simultaneously it marginally
increases the explanatory power of the model. This indicates that the R&D personnel
variable is a useable substitute for R&D capital. The model estimated in columns 8
and 11 shows a similar significant additional effect of having researchers employed.
The effect of R&D competent employees seems to be robust and increasing by the
personnel type intensity.

The inclusion of business sector and firm size level indicators in columns 9 to 12
shows contrary to the results in column 1 and 2 that there are significantly higher
productivity in the medico sector and the two knowledge service sectors and that
productivity increases disproportional with firm size.16

16
The labour and other capital variable shall explain the proportional increases in productivity. Hence,
the indicator variables for small, medium, large and very large explain jumps in the level of productivity.

15
Table 5 The basis model estimated with different interaction coefficients on R&D capital and level; R&D active firms solely
Basis model with different interaction and level indicators
1 2 3 4 5 6 7 8 9 10 11 12
Constant 8.772 10.055 8.879 9.118 9.099 9.118 9.455 9.780 9.582 9.757 9.453 9.710
Labour 0.585 0.497 0.566 0.577 0.570 0.563 0.675 0.576 0.511 0.623 0.517 0.500
Assets 0.249 0.229 0.245 0.237 0.246 0.242 0.234 0.253 0.244 0.240 0.254 0.247
R&D capital 0.103 0.092 0.109 0.100 0.091 0.104 0.049 0.103 0.093 0.043 0.087 0.088
Interactions with R&D capital
Low tech 0.017
Medium-low tech -0.038
Medico -0.006
ICT industry -0.019

Industry
Other high and medium-high tech Ref.
ICT service 0.032
Other knowledge service -0.009
Other sectors 0.059
< 50 employees 0.041
50-99 employees -0.011
100-249 employees -0.052
> 249 employees Ref.
Foreign owner 0.023
R&D Department 0.005 0.016
Researchers employed 0.033
Extern financed R&D 0.106
Public financed R&D -0.024
Basic R&D -0.001 -0.001
Applied R&D 0.061 0.057
Development R&D 0.005 0.007
Share of R&D personnel 0.056 0.049
Share of researchers 0.136 0.116
Continues....
Note: R&D capital is accumulated over the ten years and depreciated by 20 percent per year. Parameters in bold is significant on a 1% level. Parameters in italic
and bold is significant at a 10% level. Labour, Assets and R&D capital are measured in logarithms.

16
Table 5 continued
Basis model with different interaction and level indicators
1 2 3 4 5 6 7 8 9 10 11 12
Level interactions
Low tech -0.183 0.018 0.027 0.016 0.002
Medium-low tech 0.601 0.056 0.043 0.048 0.082
Medico 0.320 0.215 0.175 0.227 0.149
ICT industry 0.249 -0.052 -0.031 -0.049 -0.065

Industry
Other high and medium-high tech Ref. Ref. Ref. Ref. Ref.
ICT service -0.335 0.127 0.126 0.116 0.142
Other knowledge service 0.347 0.198 0.166 0.163 0.186
Other sectors -0.820 -0.026 -0.022 -0.032 -0.017
< 50 employees -1.088 -0.401 -0.259 -0.379 -0.409
50-99 employees -0.175 -0.236 -0.142 -0.220 -0.250
100-249 employees 0.580 -0.155 -0.101 -0.145 -0.161
> 249 employees Ref. Ref. Ref. Ref. Ref.
Foreign owner -0.214
R&D Department -0.134 -0.276
Researchers employed -0.335
Extern financed R&D -1.524
Public financed R&D 0.389
Basic R&D 0.191 0.219
Applied R&D -0.979 -0.966
Development R&D -0.072 -0.022
Number of firms 662 662 662 662 662 662 662 662 662 662 662 662
R2 0.92 0.92 0.92 0.92 0.92 0.92 0.92 0.92 0.92 0.93 0.92 0.92
Note: R&D capital is accumulated over the ten years and depreciated by 20 percent per year. Parameters in bold is significant on a 1% level. Parameters in italic
and bold is significant at a 10% level. Labour, Assets and R&D capital are measured in logarithms.

17
5. Conclusion
The present analyses have documented the positive influence of R&D on firm
productivity in the most extensive analysis so far on Danish firm specific R&D data.
The immediate conclusion is that R&D active firms have a much higher value added
per employee than R&D inactive firms. Similarly, there is a significant positive return
to productivity from R&D capital in analyses of all firms as well as R&D active firms
solely.

The conclusions are that


• R&D active firms have a 40 percent higher value added per employee.
• R&D active firms have 50 percent of the total value added in the private sector
even though they only cover 17 percent of all the firms.
• The return to R&D capital is 34 percent among the R&D active firms and 11
percent among all firms. The return is significantly higher than the return to the
other firm specific capital.
• The higher the share of R&D employees or researchers is in the firms, the higher
is the output elasticity of R&D capital.

A simulation of the effect of additional R&D capital to the firms shows that17
• 1 percent increased R&D capital among the R&D active firms will increase their
value added by 214 mill. DKK in year 1.
• 1 percent increased R&D capital among all firms will increase their value added
by 70 mill. DKK in year 1.
• 1 percentage-point increase in the share of R&D personnel causes a 1.6 bio.
DKK increase in the value added per year among the R&D active firms
(conditional on unchanged R&D capital and unchanged total number of
employees).
• 1 percentage-point increase in the share of research educated personnel causes
a 2.8 bio. DKK increase in the value added per year among the R&D active firms
(conditional on unchanged R&D capital and unchanged total number of
employees). However, this simulation builds on a thin data set and is rather
imprecise.

The estimated effects are minimum estimates for the social or community effect,
since there is a positive multiplication effect of such increases in R&D activities.
These conclusions concerning the Danish private sector are also found in other more
or less comparable studies. In Smith et al (2000) it is concluded that the return to

17
A 1 percent increase in R&D capital corresponds to 620 mio. DKK. A 1 percentage point increase in
the share of R&D personnel or research educated employees corresponds to 3633 persons.

18
R&D capital is positive and large, in the DISKO report no. 7 from 1999 it is concluded
that increased competitivity causes more innovation, but often decreases the
employment growth and last but not least in the IDA report from 2004 it is concluded
that innovation increases growth and that employment of highly educated employees
increases the innovation probability in the firms.

5.1 Research policy implications


The results found in the analyses give some explanations on how R&D activities
influence the private sector productivity. The results show that the Danish private
sector has a group of firms that are more competitive and create a higher value
added because they are R&D active. The present research policy in Denmark aim to
increase the private sector competiveness through an increased knowledge stock
imbedded in the production. This seems to be an effective strategy since there
obviously is an economic benefit from R&D activities and because there in total is no
loss of jobs in Denmark even though the private sector cannot compete on wages
compared to firms in Eastern Europe or Asia.

In the light of the conclusions above, the future development in the Danish research
policy must be to further support the sectors possibilities to become and stay
innovative or even better R&D active and to increase the amount of R&D activities.
Furthermore, the private sector should be supported in their aims to increase the
knowledge imbedded in their production. The analyses in this report show that an
effective way to support both intentions is to increase the education level in the firms
by employing R&D competent employees or by up-qualifying the competences
among the existing stock of employees in the firms.

References
AFSK. 2003. R&D in the Private Sector. Statistics for 2001. (In Danish:
Erhvervslivets forskning og udviklingsarbejde. Forskningsstatistik 2001.) R&D
Statistics Series 2003, The Danish Institute for Studies in Research and
Research Policy. 2003.
Bayer M. 2003. The Market Valuation of R&D intensive Firms and the Capitalization
of R&D Expenditures (In Danish: Aktiemarkedets prisfastsættelse af FoU-
intensive virksomheder, herunder aktivering af FoU-udgifterne). AFSK Report
2003/1, The Danish Institute for Studies in Research and Research Policy.
Bloch C. 2003. The Effect of R&D Expenditures on Stock Market Returns for Danish
Firms. AFSK WP 2003/6, The Danish Institute for Studies in Research and
Research Policy.
Chan L.K.C., J. Lakonishok and T. Sougiannis. 2001. The Stock Market Valuation of
Research and Development Expenditures. Journal of Finance, 56, 2431-56.

19
Criscuolo C. and J. Haskel. 2002. Innovations and Productivity Growth in the UK.
CeRiBA Discussion Paper.
Cuneo P. and J. Mairesse. 1984. Productivity and R&D at the firm level in French
manufacturing. In Z. Griliches (ed.) R&D, patents, and productivity. University of
Chicago Press.
Dilling-Hansen M., T. Erikson, E.S. Madsen and V. Smith. 1999. The Impact of R&D
on Productivity: Evidence from Danish Manufacturing Firms. AFSK WP 1999/1.
The Danish Institute for Studies in Research and Research Policy.
Danish Economic Council. 1997. Danish Economy, Spring 1997. (In Danish:
Vismandsrapport).
Graversen E.K. and M. Mark. 2004. The Effect of R&D on Productivity and
Employment. (In Danish: Forskning og Udviklingsarbejdes påvirkning af
produktivitet og beskæftigelse). CFA Report 2005/1, The Danish Centre for
Studies in Research and Research Policy.
Griliches Z. and J. Mairesse (1984). Productivity and R&D at the Firm Level, in
Griliches Z. (ed.), R&D, Patents and Productivity, University of Chicago Press,
Chicago.
Griliches Z. 1998. R&D and Productivity: Economic Results and Measurement
Issues, in Paul Stoneman (ed.), Handbook of the Economics of Innovation and
Technological Change, Blackwell, 52-89.
Hall B. 1993. The Stock Market’s Valuation of R&D Investment During the 1980’s.
American Economic Review, vol. 83, pp. 259-264.
Hall B. and J. Mairesse. 1995. Exploring the relationship between R&D and
productivity in French manufacturing firms. Journal of Econometrics, Vol. 65.
Hall B. and R. Oriani. 2003. Does the Market Value R&D Investment by European
Firms? Evidence from a Panel of Manufacturing Firms in France, Germany and
Italy. DRUID Conference paper 2003.
IDA. (Rambøll Management og Ingeniørforeningen i Danmark). 2004. The Value of
High Educated Employees in Small and Medium Sized Enterprises. (In Danish:
Højtuddannedes værdi for små og mellemstore virksomheder). IDA Report.
Lundvall B.-Å. 1999. The Danish Innovation System – A research based discussion
document concerning the Challenges and Action Possibilities of Innovation
Policy. (In Danish: Det danske innovationssystem - et forskningsbaseret oplæg
om innovationspolitiske udfordringer og handlemuligheder) The DISKO-project.
Summarizing Report, DISKO report no. 9.
Medda G., C.A. Piga og D. Siegl. 2003. On the Relationship Between R&D and
Productivity: A Treatment Effect Analysis. Working Paper 02/2003, Nottingham
University Business School.
Olesen-Larsen P. 2003. Does it Matter? Payoff from R&D in the Public Sector. (In
Danish: Kan det betale sig? Udbytte af forskning i den offentlige sektor).
Økonomi & Politik nr. 4-2003.
Rouvinen, P. 2002. R&D-Productivity Dynamics: Causality, Lags and “Dry Holes”.
Journal of Applied Economics, Vol. V, no. 1, pp. 123-156.

20
Salter A.J. and B.R. Martin. 1999. The economic benefits of publicly funded basic
research: A critical review. SPRU Electronic WP series no. 34.
Schankerman M. 1981. The effect of double counting and expensing on the
measured returns to R&D. Review of Economics and Statistics, Vol. 63.
Singh N. and H. Trieu. 1996. The Role of R&D in Explaining Total Factor Productivity
Growth in Japan, South Korea and Taiwan. Working Paper no. 361, Department
of Economics University of California, Santa Cruz.
Smith N. 2001. Investment in R&D. (In Danish:Investering i forskning). Chapter 6 in
The Danish Reseach Commision’s Recommendations. Vol.2.
Smith N. 2002. Public Investments in R&D – To little, To Much or Misguided? (In
Danish: Offentlige investeringer i forskning og udvikling - for lidt, for meget eller
forkert?). Økonomi & Politik nr. 3-2002.
Smith V., M. Dilling-Hansen, T. Eriksson og E. Strøjer Madsen. 1999. The Impact of
R&D on Productivity. (In Danish: Påvirker omfanget af virksomhedernes
forskning og udviklingsarbejde deres produktivitet?). AFSK Report 1999/7, The
Danish Institute for Studies in Research and Research Policy.
Smith V., M. Dilling-Hansen, T. Eriksson og E. Strøjer Madsen. 2000. R&D and
Productivity in Danish Firms: Some Empirical Evidence. AFSK WP 2000/4, The
Danish Institute for Studies in Research and Research Policy.
Smith V., M. Dilling-Hansen, T. Eriksson og E. Strøjer Madsen. 2001. The Efficiency
and R&D Investments among Danish Enterprises. (In Danish: Danske
virksomheders effektivitet og investeringer i forskning og udvikling). AFSK WP
2001/4, The Danish Institute for Studies in Research and Research Policy.
Solow .1956. A Contribution to the Theory of Economic Growth. Quarterly Journal of
Economics, 70:65-94.
Wakelin, K. 2001. Productivity Growth and R&D Expenditure in UK Manufacturing
Firms. Research Policy 30, pp. 1079-1090.

21
Appendix 1 Construction of R&D Capital measures
1. R&D capital is the present value of all firm specific R&D costs in the past. In
reality the R&D capital equals the present value of the last ten years R&D costs
back to time period t-10.18
a. Alternatively period t is not included in the sum, i=0,1,... is changed to
i=1,2,... under point 4.a. Hereby a clear time dependent causality from
R&D costs to productivity is secured.

2. R&D Costs in the years previous to 2001. The costs are collected in real
values, i.e. 2001-prices. The firms own given R&D costs are used if the firms
have reported such in the specific years. Otherwise, R&D costs are estimated for
the firms in the years the R&D costs are not reported; given that the firms existed
these years. Hence, exact or estimated R&D costs exist for all firms in all years
back to 1991 in the sample of firms in 2001. The exact procedure is as follows:
a. The firms own reported R&D costs are used if it exists.
1. Values in years between two valid observations are linearly
approximated as the average of the two values. This is primarily
done for the years where no R&D data has been collected.
(2000, 1996, 1994 and 1992).
b. In case of missing information on R&D costs in previous years it is
assumed that:
1. There is no R&D cost in previous periods if there is no R&D cost
in 2001 or in the first valid observation after a missing
observation.
2. Sector specific growth rates are calculated from the valid data in
the respective years. The R&D costs in previous years are
consecutively calculated as R&Dt-1 = R&Dt * (1+rt)-1 for each firm,
where rt is the sector specific growth rate.
3. The estimated R&D costs are corrected for non-existence of the
firms in the previous years.

3. R&D costs is assumed to have a real depreciation rate, dt-i, which may depend
on the R&D type, c.f. table A.1. Hall and Oriani (2003) uses depreciation rates on
15 percent and accumulates over 5 and 10 years respectively. They do not
consider R&D type. Dilling-Hansen et al. (1999) and Smith et al (1999, 2000 and
2001) uses 20 percent while Bloch (2003) and Chan et al (2001) use a linear
depreciation on 20 percent per year in four years (20, 40, 60 and 80 percent).

18
The method in Griliches and Mairesse (1984), Hall (1993) and Hall and Oriani (2003) is closely
followed.

22
Bloch tests the sensitivity of the definition of depreciation method by a
comparison with the normal method, but uses here a rate on 15 percent.

Table A.1 R&D cost type and depreciation rates used in K5a and K10a
R&D cost type Depreciation rate, dt
Operational costs: 15%
a. Salary 14%
a. Other operational costs 20%
Investment costs: 10%
a. Buildings 7%
a. Apparatus and instruments 12%

4. R&D capital, Kt, is calculated as the present value of the R&D costs in the
firms:19
a. As Kt = Σi Σu R&Du,t-i * Πi (1-du,t-i), i = 0, 1, ...., J, u = R&D type.
b. Simpler, it is just the sum of the present value of all the annual R&D costs
inflated to 2001-prices.

5. Different measures of R&D capital, Kt. It the paper seven different R&D capital
measures is calculated and used to test the parameter stability in the empirical
model.
a. K5at, K10at sums over 5 and 10 years respectively with depreciation rates
as shown in table A.1.
b. K5_15t, K10_15t sums over 5 and 10 years respectively with depreciation
rates on 15 percent.
c. K5_20t, K10_20t sums over 5 and 10 years respectively with depreciation
rates on 20 percent.
d. K_Bayert with depreciation rate over three years and an immediate
depreciation on 40 percent in period 0, c.f. Bayer (2003).

19
Alternatively, the R&D capital could be estimated from firm output such as firm value of patents, HR,
among others. This information is normally difficult to measure and is seldom available right away.
Furthermore, the depreciation rates could be estimated on sector level by the methods used in Bayer
(2003). He estimates a model on the ordinary operational result explained by material assets, sales
costs and R&D costs together with three lags of R&D costs. He finds decreasing depreciation rates over
the three years and an immediate depreciation on 40 percent in period 0. Unfortunately, the
depreciation rates are relative (sums to 1) and not absolute. Bayer’ results are used in the analysis.

23
Appendix 2 Definitions and additional empirical results

Table A.2 Average R&D capital for different R&D capital measures, 1000 DKK
Variable Average Minimum Maximum
R&D capital 1) conditional on positive 20421 6 4228230
Other assets (exclusive R&D capital1)) 128809 6 55889929
R&D capital 2) conditional on positive 19291 6 4228230
Other assets (exclusive R&D capital2)) 131320 14 55899123
R&D capital 3) conditional on positive 23156 7 4478391
Other assets (exclusive R&D capital3)) 128732 6 55876888
R&D capital 4) conditional on positive 21017 7 4478391
Other assets (exclusive R&D capital4)) 131323 14 55890352
R&D capital 5) conditional on positive 23799 7 4523817
Other assets (exclusive R&D capital5)) 128004 5 55954436
R&D capital 6) conditional on positive 21284 7 4523817
Other assets (exclusive R&D capital6)) 130400 5 55954436
R&D capital 7) conditional on positive 6846 0 1646856
Other assets (exclusive R&D capital7)) 137132 23 55939668
Maximum number of unweighted firms: 2228
Note: 1) Accumulated over ten years and depreciated by 20 percent per year. 2) Accumulated over five
3)
years and depreciated by 20 percent per year. Accumulated over ten years and depreciated by 15
percent per year. Accumulated over five years and depreciated by 15 percent per year. 3)
4)
4)
Accumulated over ten years and depreciated as described in table A.1. Accumulated over five years
and depreciated as described in table A.1. 7) Accumulated over four years and depreciated as estimated
in Bayer (2003).

Table A.3 Estimation of the production function where assets equal fixed
assets solely
All firms R&D active firms
Model 1) 2)
1 2 33) 4 4)
1 1)
22) 33) 44)

Constant 11.897 11.929 12.751 12.724 11.693 10.848 10.860 11.543


Labour 0.804 0.801 0.866 0.847 0.757 0.725 0.727 0.725
R&D labour - - - 0.247 - - - 0.102
Assets 0.107 0.105 0.037 0.043 0.143 0.108 0.101 0.096
R&D capital - 0.021 0.025 0.010 - 0.105 0.112 0.066
Indicator for
0.273 - - -
R&D capital > 0
Number of firms 2228 2228 2228 2228 662 662 662 662
R2 0.87 0.87 0.86 0.87 0.91 0.92 0.92 0.92
Note: All variables are measured in logarithms with the exception of the ”indicator for R&D”. R&D capital
is accumulated over ten years and depreciated by 20 percent per year. Parameters in bold is significant
on a 1% level. 1) Simple model with an indicator for R&D capital>0. 2) Incl. R&D capital, but without
correction for double counting. 3) Incl. R&D Capital and correction for double counting in assets. This
4)
model is the basis model is Section 4. Incl. R&D capital and R&D personnel and correction for double
counting in both assets and labour.

24
Table A.4 The basis model estimated with different interaction coefficients on R&D capital and level; all firms
Basis model with different interaction and level indicators
1 2 3 4 5 6 7 8 9 10 11 12
Constant 9.549 10.318 9.753 9.838 9.843 9.902 9.872 9.740 9.825 9.913 9.836 9.794
Labour 0.668 0.609 0.643 0.642 0.649 0.640 0.670 0.648 0.637 0.689 0.637 0.640
Assets 0.264 0.255 0.259 0.254 0.253 0.250 0.242 0.259 0.263 0.249 0.264 0.258
R&D capital 0.013 0.005 0.017 0.015 0.014 0.015 -0.000 0.017 0.012 -0.002 0.012 0.012
Interactions with R&D capital
Low tech -0.007
Medium-low tech -0.008
Medico 0.020
ICT industry 0.004

Industry
Other high and medium-high tech Ref.
ICT service 0.008
Other knowledge service 0.013
Other sectors -0.009
< 50 employees 0.015
50-99 employees 0.007
100-249 employees 0.001
> 249 employees Ref.
Foreign owner 0.014
R&D Department 0.010 0.018
Researchers employed 0.168
Extern financed R&D 0.309
Public financed R&D -0.187
Basic R&D -0.024 -0.032
Applied R&D 0.055 0.059
Development R&D 0.088 0.078
Share of R&D personnel 0.080 0.073
Share of researchers 0.195 0.169
Continues.....
Note: R&D capital is accumulated over the ten years and depreciated by 20 percent per year. Parameters in bold is significant on a 1% level. Parameters in italic
and bold is significant at a 10% level. Labour, Assets and R&D capital are measured in logarithms.

25
Table A.4 continued
Basis model with different interaction and level indicators
1 2 3 4 5 6 7 8 9 10 11 12
Level interactions
Low tech -0.003 -0.011 -0.014 -0.020 0.003
Medium-low tech 0.029 0.014 0.013 0.007 0.039
Medico -0.008 0.176 0.103 0.185 0.075
ICT industry -0.074 -0.052 -0.064 -0.061 -0.054

Industry
Other high and medium-high tech Ref. Ref. Ref. Ref. Ref.
ICT service 0.219 0.256 0.211 0.237 0.255
Other knowledge service 0.187 0.212 0.173 0.196 0.225
Other sectors 0.027 0.028 0.011 0.017 0.043
< 50 employees -0.432 -0.187 -0.171 -0.209 -0.091
50-99 employees -0.297 -0.111 -0.129 -0.132 -0.026
100-249 employees -0.177 -0.075 -0.078 -0.088 -0.011
> 249 employees Ref. Ref. Ref. Ref. Ref.
Foreign owner -0.041
R&D Department -0.134 -0.207
Researchers employed -2.550
Extern financed R&D -4.476
Public financed R&D 2.868
Basic R&D 0.587 0.745
Applied R&D -0.999 -1.091
Development R&D -1.261 -1.107
Number of firms 2228 2228 2228 2228 2228 2228 2228 2228 2228 2228 2228 2228
R2 0.90 0.90 0.90 0.90 0.90 0.90 0.91 0.90 0.90 0.91 0.89 0.90
Note: R&D capital is accumulated over the ten years and depreciated by 20 percent per year. Parameters in bold is significant on a 1% level. Parameters in italic
and bold is significant at a 10% level. Labour, Assets and R&D capital are measured in logarithms.

26
Table A.5 The basis model estimated with seven different R&D capital
measures and with interaction coefficients on business sector
Basis model with different R&D capital measures
1 2 3 4 5 6 7
Constant 9.622 9.618 10.314 10.166 9.395 9.393 9.383
Labour 0.661 0.662 0.711 0.701 0.646 0.647 0.648
Assets 0.264 0.264 0.213 0.224 0.281 0.281 0.282
R&D capital 0.011 0.011 0.012 0.012 0.009 0.010 0.010
Interactions with
R&D capital
Low tech -0.008 -0.008 -0.008 -0.009 -0.007 -0.007 -0.008
Medium-low
-0.006 -0.007 -0.007 -0.007 -0.006 -0.006 -0.006
tech
Industry

Medico 0.019 0.019 0.022 0.021 0.015 0.015 0.017


ICT industry -0.001 -0.003 -0.000 -0.003 -0.000 -0.003 -0.003
Other high and
medium-high Ref. Ref. Ref. Ref. Ref. Ref. Ref.
tech
ICT service 0.021 0.020 0.023 0.023 0.020 0.020 0.022
Other knowledge
0.025 0.024 0.026 0.024 0.025 0.024 0.027
service
Other sectors -0.007 -0.008 -0.006 -0.007 -0.006 -0.007 -0.008
Number of firms 2228 2228 2228 2228 2228 2228 2228
2
R 0.90 0.90 0.89 0.89 0.90 0.90 0.90
Note: Note: All variables are measured in logarithms. Parameters in bold is significant on a 1% level,
1)
parameters in italic and bold is significant on a 10% level. Accumulated over ten years and
2)
depreciated by 20 percent per year. Accumulated over five years and depreciated by 20 percent per
3) 4)
year. Accumulated over ten years and depreciated by 15 percent per year. Accumulated over five
3)
years and depreciated by 15 percent per year. Accumulated over ten years and depreciated as
4) 7)
described in table A.1. Accumulated over five years and depreciated as described in table A.1.
Accumulated over four years and depreciated as estimated in Bayer (2003).

27
Table A.6 The basis model estimated with seven different R&D capital
measures and with interaction coefficients on business sector for R&D
active firms
Basis model with different R&D capital measures
1 2 3 4 5 6 7
Constant 8.721 8.639 8.167 8.102 8.160 8.096 8.273
Labour 0.578 0.571 0.538 0.535 0.537 0.533 0.534
Assets 0.258 0.252 0.307 0.301 0.307 0.302 0.344
R&D capital 0.097 0.111 0.087 0.100 0.087 0.100 0.042
Interactions with
R&D capital
Low tech 0.003 0.005 0.002 0.004 0.002 0.004 -0.001
Medium-low
0.005 0.006 0.004 0.005 0.003 0.005 -0.000
tech
Industry

Medico 0.013 0.012 0.015 0.014 0.008 0.007 0.012


ICT industry -0.003 -0.006 -0.003 -0.006 -0.003 -0.006 -0.006
Other high and
medium-high Ref. Ref. Ref. Ref. Ref. Ref. Ref.
tech
ICT service 0.010 0.009 0.009 0.023 0.010 0.009 0.014
Other knowledge
0.015 0.014 0.016 0.007 0.015 0.015 0.018
service
Other sectors 0.000 0.000 -0.001 -0.001 -0.001 -0.001 -0.008
Number of firms 662 662 662 662 662 662 662
2
R 0.92 0.92 0.93 0.93 0.93 0.93 0.92
Note: Note: All variables are measured in logarithms. Parameters in bold is significant on a 1% level,
parameters in italic and bold is significant on a 10% level. 1) Accumulated over ten years and
depreciated by 20 percent per year. 2) Accumulated over five years and depreciated by 20 percent per
year. 3) Accumulated over ten years and depreciated by 15 percent per year. 4) Accumulated over five
3)
years and depreciated by 15 percent per year. Accumulated over ten years and depreciated as
described in table A.1. Accumulated over five years and depreciated as described in table A.1. 7)
4)

Accumulated over four years and depreciated as estimated in Bayer (2003).

Table A.7 Business sector definitions


Business sector NACE code
Other high and medium-high tech industry 24, 29-35.3 (except medico and ICT industry)
Medium-low tech industry 23, 25-28
Low tech industry 15-22, 35.4-37
Knowledge service 64, 72-74, 92.2 (except ICT service)
Other sectors 1-14, 40-41, 45, 51-52, 60-63, 65-67, 75.25, 90
(except ICT trade)
Medico (industry) 24.4, 33.1
ICT industry and trade 30.01-30.02, 31.3, 32.1-32.3, 33.2-33.3, 51.43.2,
51.64.1, 51.84, 51.86
ICT service 64.2, 71.33, 72.1-72.6

28

View publication stats

You might also like