Corporate R&D and Productivity

1
IPTS-2009-J03-16-NC

Corporate R&D and productivity:
Econometric tests based on microdata"

Final Report

IN

2
INDEX
Executive non-technical summary p. 7
1. Introduction: background, hypotheses and objectives of the study p. 15
2. Theoretical background and previous empirical evidence p. 18
2.1 The transatlantic productivity gap p. 18
2.2 The link between R&D and productivity at the firm and sectoral level p. 24
2.3 The role of embodied technological change and the peculiarities of the
medium and low-tech sectors p. 27
2.4 The business cycle and the regional peculiarities p. 30
3. Data and methodology p. 33
3.1 The data source p. 33
3.2 The construction of the dataset p. 34
3.3 The econometric specification and descriptive statistics p. 42
4. Econometric analysis p. 49
4.1 Overall results; EU vs. US p. 49
4.2 A sectoral breakdown p. 51
4.3 The US/EU comparison: crossing the geographical and the sectoral
dimensions p. 53
4.4 A regional breakdown p. 57
4.4.1 Macro-regions p. 58
4.4.2 R&D-intensive regions vs. non R&D-intensive ones p. 62
4.5 The business cycle p. 68
5. Conclusions and policy implications p. 72
References p. 79
Appendix p. 89

3
LIST OF FIGURES AND TABLES

List of figures

Fig. 1: Real GDP growth in the US and the EU15: 1990-2008 p. 19
Fig. 2: Labour productivity growth in the US and the EU15: 1990-2008 p. 20
Fig. 3: TFP growth in the US and the EU15: 1990-2004 p. 21
Fig. 4: GERD/GDP in the US and the EU15: 1990-2007 p. 22
Fig. 5: Private R&D (BERD)/GDP in the US and the EU15: 1990-2007 p. 23

4
List of tables

Tab. 1: Distribution of firms and observations across countries in the final
version of the dataset p. 41
Tab. 2: Correlation table: correlation coefficients p. 43
Tab. 3: VA/E (Value Added/Employees) p. 45
Tab. 4: K/E (R&D Stock/Employees) p. 46
Tab. 5: C/E (Physical Capital Stock/Employees) p. 47
Tab. 6: E (Employees) p. 48
Tab. 7: Whole sample, US and EU p. 50
Tab. 8: Sectoral decomposition: Manufacturing (High-tech + Other) and
Service sectors p. 51
Tab. 9: Sectoral decomposition: High-tech and Other manufacturing sectors p. 53
Tab. 10: US versus EU: Manufacturing sectors p. 54
Tab. 11: US versus EU: Service sectors p. 54
Tab. 12: US versus EU: High-tech manufacturing sectors p. 56
Tab. 13: US versus EU: Other manufacturing sectors p. 56
Tab. 14: European macroareas: North (Denmark, Finland, Sweden) + UK,
Other EU countries, Other EU countries without South p. 58
Tab. 15: European macroareas: Manufacturing sectors p. 60
Tab. 16: European macroareas: Service sectors p. 60
Tab. 17: European macroareas: High-tech manufacturing sectors p. 61
Tab. 18 : European macroareas: Other manufacturing sectors p. 61
Tab. 19: European NUTS R&D intensities (BERD/GDP) p. 63
Tab. 20: European NUTS: Innovative NUTS versus Weakly innovative NUTS
(Regional BERD/GDP >= 1.8% is the threshold) p. 65
in High-tech manufacturing sectors p. 66
Tab. 22: European NUTS: Innovative NUTS versus Weakly innovative NUTS in
Other manufacturing sectors p. 66
Tab. 23: European NUTS: Innovative NUTS versus Weakly innovative NUTS in
Manufacturing sectors p. 67
Tab. 24: European NUTS: Innovative NUTS vs. Weakly innovative NUTS in
Service sectors p. 67

5
Tab. 25: Whole sample, Recessions and Expansions p. 69
Tab. 26: US versus EU: Recessions p. 69
Tab. 27: US versus EU: Expansions p. 70

6
Appendix

Tab. A7: Whole sample, US and EU (complete) p. 90
Tab. A8: Sectoral decomposition: Manufacturing (High-tech + Other) and
Service sectors (complete) p. 91
Tab. A9: Sectoral decomposition: High-tech and Other manufacturing
sectors (complete) p. 92
Tab. A10: US versus EU: Manufacturing sectors (complete) p. 93
Tab. A11: US versus EU: Service sectors (complete) p. 94
Tab. A12: US versus EU: High-tech manufacturing sectors (complete) p. 95
Tab. A13: US versus EU: Other manufacturing sectors (complete) p. 96
Tab. A14: European macroareas: North (Denmark, Finland, Sweden) + UK,
Other EU countries, Other EU countries without South (complete) p. 97
Tab. A15: European macroareas: Manufacturing sectors (complete) p. 98
Tab. A16: European macroareas: Service sectors (complete) p. 99
Tab. A17: European macroareas: High-tech manufacturing sectors (complete) p. 100
Tab. A18 : European macroareas: Other manufacturing sectors (complete) p. 101
Tab. A20: European NUTS: Innovative NUTS versus Weakly innovative NUTS
(Regional BERD/GDP >= 1.8% is the threshold) (complete) p. 102
in High-tech manufacturing sectors (complete) p. 103
in Other manufacturing sectors (complete) p. 104
Tab. A23: European NUTS: Innovative NUTS versus Weakly innovative NUTS in
Manufacturing sectors (complete) p. 105
Tab. A24: European NUTS: Innovative NUTS vs. Weakly innovative NUTS in
Service sectors (complete) p. 106
Tab. A25: Whole sample, Recessions and Expansions (complete) p. 107
Tab. A26: US versus EU: Recessions (complete) p. 108
Tab. A27: US versus EU: Expansions (complete) p. 109

7

Executive non-technical summary

8
Introduction
This summary presents the findings and conclusions of the project Corporate R&D and
productivity: Econometric tests based on microdata. The main project objective is understanding
the link between R&D and firms performance, taking into account possible implications in terms
of competitiveness and economic growth, and therefore, for European policy making.
With the exclusion of the short-term turbulence associated with the current recession,
starting from the early '90s, a persistent divide both in terms of economic growth, labour
productivity growth and total factor productivity growth can be found between the US and the EU.
The literature converges in pointing out the important role that R&D and innovation have in
explaining productivity differences within the industrialized countries. Indeed, the role of corporate
R&D has been recognized as an engine for productivity growth at the macro and microeconomic
level.
However, although the underinvestment in R&D plays a crucial role in the interpretation of
the transatlantic divide in terms of productivity performance and economic growth, the overall
European productivity delay might also be explained by a lower capacity by European firms to
translate R&D investment into productivity gains.
The report intends to analyze why European firms productivity reveals to be lower
comparing with their US counterparts. The project jointly investigates the two aspects that can
originate this gap: the lower level of public and corporate investment in Europe and the lower
capacity to translate R&D investment into productivity gains.
A second set of hypotheses of this work are related to sectoral comparisons and structural
differences. Following this approach, the comparison between the role of R&D and physical capital
in fostering productivity in high-tech and non-high-tech sectors is explored, as well as the
comparison among manufacturing and services.
Another set of hypotheses are regarding the regional disparities in the EU; in this regard, the
potential differences in the R&D/productivity link across EU regions are explored in detail.
Hence, the aim of this report is to investigate the link between R&D and labour productivity,
taking into account the time, regional and sectoral dimensions of the phenomenon and the

9
interactions of R&D with other sources of innovation such as the embodied technological change
incorporated in physical capital.
The empirical analyses conducted in this report are based on a longitudinal unbalanced panel
of microdata extracted from a variety of sources, including companies annual reports. The sample
is constructed by companies that account R&D expenditures belonging to US and EU 27. The final
panel comprises 1,809 companies for the period 1990-2008. We used labour productivity (value
added over total employment) as dependent variable, while our pivotal impact variables are the
R&D stock per employee and the physical capital stock per employee.

Key Findings
The results of the project were obtained through an econometric analysis using panel
methodologies, taking into account the sectoral, country and time dimensions of the available
microdata. In particular, the following research issues were investigated.

1) To see whether significant differences emerge in the link between R&D and productivity
between the US and the EU, in order to shed some light on the interpretation of the transatlantic
productivity gap (Section 4.1).
Consistently with the previous literature, we found robust evidence of a positive and significant
impact of R&D on productivity.
However, although uniformly positive and statistically significant, the R&D impacts for the US
firms turn out to be consistently larger than the corresponding impacts for the European firms
(about 60% of their US counterparts). We interpreted these unambiguous results as a clear evidence
of the better ability of US firms in translating R&D investments into productivity gains and as a
signal of a gap of efficiency that European firms and European policy have to deal with.

2) To see whether further support can be found to the hypothesis that R&D should be clearly
and significantly linked to productivity in the high-tech sectors, while a weaker impact should
emerge in the other sectors of the economy (Section 4.2).
This hypothesis was also confirmed by the microeconometric estimates: the R&D impacts related to
the high-tech manufacturing sectors always turned out to be larger than the corresponding ones for

10
the other manufacturing sectors. Hence, high-tech sectors not only invest more in R&D and are
characterised by a higher productivity performance, but also achieve more in terms of productivity
gains from their own R&D activities.

3) To see whether physical capital emerges as an important second driver of productivity
gains, so confirming the hypothesis that embodied technological change is a crucial determinant
of productivity evolution. This relationship is expected to be particularly strong in the low-tech
sectors, where embodied technological change might be expected the main source of productivity
gains.
Overall, we found consistently with previous studies - a positive and significant impact of
physical capital over labour productivity.
Interestingly enough, the US revealed an advantage similar to the one emerged for the intangible
R&D investments; thus, US firms resulted more efficient in getting productivity gains both from the
R&D and the physical capital investments.

4) To see to what extent the transatlantic differences may be related to the different sectoral
structures and to the peculiar sectoral R&D/productivity relationships detectable in the US and in
the EU.
We differentiated the US/EU comparative empirical exercise by manufacturing vs service sectors: it
came out that both US manufacturing and US service firms were more efficient in translating their
investments (both in R&D and in physical capital) into productivity increases. In addition, the US
efficiency advantage in R&D activities is obvious both in the high-tech manufacturing sectors and
in the rest of the manufacturing sectors. On the whole, US firms are leading in terms of R&D
efficiency regardless of the sectors. Hence, the transatlantic productivity divide can be explained
not only by a lower level of corporate R&D investment, but also by a lower capacity to translate
R&D into productivity gains, and this seems to be obvious both within manufacturing (both high-
and medium/low-tech sectors) and within services.
Finally, productivity growth in the European non-high-tech firms is still heavily dependent on the
investment in physical capital (embodied technological change).

5) To see whether (and how much) the intensity of the R&D/productivity link is affected by
the sectoral composition and the institutional context characterising the different European
countries and regions.

11
On the whole, the EU economy appeared to be divided into two different macro-areas. On the one
hand, there is the Nordic and British world where R&D and productivity are strongly linked, with
exceptionally good results with regard to the high-tech manufacturing sectors. On the other hand,
there is the rest of Europe exhibiting quite lower R&D/productivity impacts. This is particularly
important in terms of European economic and innovation policy, since the European productivity
gains related to R&D activities seem to be largely driven by what is going to happen in the Nordic
countries and in the UK, with the rest of Europe lagging behind, especially with regard to the role of
the high-tech manufacturing sectors.
Turning the attention to the regional level, we grouped together the European regions into two
groups: the innovative regions and the weakly innovative ones, according to their R&D/GDP ratio.
As far as empirical results are concerned, a general conclusion was that those regions that invest
more in R&D are also characterised by a better ability to translate the R&D investment into an
increase in productivity. In particular, the strongest R&D/productivity links were displayed by the
firms belonging to the high-tech sectors and located in the most innovative regions. Symmetrically,
productivity growth in medium and low-tech sectors and in the less innovative regions was found
still heavily dependent on investment in physical capital (embodied technological change), with
corporate R&D playing a secondary role.

6) To see whether the coefficients linking R&D and productivity are stable over time or turn
out to be affected by the business cycle.
On the whole, the analysis of the business cycle revealed that the overall differences in the impacts
along the opposite phases of the cycle were not so significant, with a weak evidence of a larger
productivity impact of both R&D and capital during the recessionary periods. However - with
regard to the US/EU comparison the European gap in terms of lower productivity returns from
both R&D investment and capital formation was fully confirmed and found to be independent of the
business cycle.

12
Conclusions and policy implications
From a general policy point of view, European economies - compared with the US economy
- not only appear to invest less in R&D, but also get less returns from their R&D investment in
terms of productivity gains.
In terms of European industrial policy, high-tech manufacturing sectors appear to enjoy
more benefits in terms of productivity coming out from their company investments on R&D.

In terms of European regional policy, the particular nature of the relationship between R&D
and capital formation on the one hand and productivity evolution on the other hand might heavily
be affected by the industrial structure which characterises a single region.
On the basis of the reports results discussed above, , we can conclude that - at least in
Europe the R&D investment is strongly characterised by increasing returns in terms of its
productivity impact. In fact, this impact is higher in: 1) the high-tech manufacturing sectors; 2) in
the Nordic countries and in the UK; 3) in the most innovative European regions; that is: where more
is invested in R&D, more is achieved in terms of productivity gains. This outcome implies
important policy implications.
Firstly, the obtained results show that the US economy is uniformly more efficient in getting
productivity advantages from investments in R&D activities; while this is obvious for the whole
economy, the efficiency gap is confirmed separately in services and manufacturing and within
manufacturing both in the high-tech sectors and in the other industrial sectors. Hence, the
transatlantic divide is not only a matter either of a lower R&D investment in Europe or of an
European industrial structure specialised in middle and low-tech sectors. With the only exception of
UK and the Nordic Countries, European firms are structurally less able to translate R&D
expenditures into productivity gains. This conclusion has a first important policy implication: just
increasing R&D is a necessary but not sufficient policy if the overall increase in productivity is the
target.
Secondly, this study clearly shows that higher productivity gains from R&D investments can
be achieved in the high-tech manufacturing sectors. Here a second policy implication emerges: the
allocation of R&D efforts is as important as an increase in R&D and high-tech sectors should be
targeted by national and European R&D policies. Indeed, the results coming out from this report
offer a second reason to favour European high-tech sectors: in fact, they not only invest more in

13
R&D, but in these sectors corporate R&D efforts are more effective in achieving productivity gains.
In other words, the findings of this research support a targeted research policy rather than an erga
omnes (horizontal) type of public intervention.
Thirdly, this study shows that R&D investment is not the sole source of productivity gains;
technological change embodied in capital formation is of comparable importance. Also with regard
to the relationship between physical capital and productivity, the US economy exhibits an
advantage, similar to the one detected for the R&D activities. Finally, embodied technological
change appears to be crucial within European non-high-tech firms; hence, an European innovation
policy aiming to increase productivity in the medium/low-tech sectors should support overall capital
formation.
Fourthly, the European aggregate seems to be divided into two different worlds: on the one
hand, the UK and the Nordic countries which exhibit an R&D/productivity pattern similar to the US
one, and on the other hand the rest of the Continent lagging behind. This divide is also obvious
looking at the regional level, where the most R&D-based regions also show the better results in
terms of R&D/productivity elasticities. Overall, Europe is lagging behind the US and within
Europe the R&D/productivity link is clearly characterised by increasing returns both in terms of
countries, regions and sectors: where more is invested in R&D, more is achieved in terms of
productivity gains. An European regional policy targeted to fill the transatlantic productivity gap
should carefully take into account the presence of this strong heterogeneity across European
regions.

14

Corporate R&D and productivity:
Econometric tests based on microdata

15

1. Introduction: background, hypotheses and
objectives of the study

The understanding of the link between R&D, firms performance and competitiveness is one
of the major interests for European policy makers (see European Commission, 2002 and 2008).

A first hypothesis of this study is that the lower European economic performance in
comparison with the US can be explained not only by a lower level of public and corporate R&D
investment, but also by a lower capacity to translate R&D investment into productivity gains, which
in turn foster competitiveness and economic growth. A second hypothesis of this work is that R&D
may be crucial in fostering productivity in the high-tech sectors, while being less important in the
rest of the economy where alternative sources of productivity growth such as embodied
technological change may play a dominant role. The final hypothesis is that the relationship
between R&D and productivity may exhibit important differences across EU countries and regions
and may vary along the business cycle.

If these hypotheses were supported, together with macroeconomic policies based on
aggregate R&D targets, specific European industrial and innovation policies should be designed for
different microeconomic environments - according to the different industrial sectors and the
different national and regional contexts - in order to enforce the link between technological inputs
and productivity gains.

Hence, the aim of this report is to better investigate the link between R&D and labour
productivity, taking into account the time, regional and sectoral dimensions of the phenomenon and
the interactions of R&D with other sources of innovation such as the embodied technological
change incorporated in physical capital (see next section for a theoretical background).

In particular, the objectives of the project can be articulated as follows.

a. Conceptualising - building on the state of the art - the nature of the relationships
between R&D and embodied technological change on the one hand and labour productivity on the
other hand, at the national, sectoral and particularly the firms level (Section 2).

16

b. Building a new unique panel database obtained by elaborations using original IPTS
microdata and applying the methodology discussed in Section 3.

c. Investigating, through econometric analysis, the relationship between R&D and
embodied technological change on the one hand and labour productivity on the other hand; the
analyses will be carried out using panel methodologies (see Section 3), taking into account the
sectoral, country and time dimensions of the available microdata. In particular, the following
research issues will be investigated.

- c.1 To see whether significant differences emerge in the link between R&D and
productivity between the US and the EU, in order to shed some light on the interpretation of the
transatlantic productivity gap (Section 4.1).

- c.2 To see whether further support can be found to the hypothesis that R&D
should be clearly and significantly linked to productivity in the high-tech sectors, while a weaker
impact should emerge in the other sectors of the economy (Section 4.2).

- c.3 To see whether physical capital emerges as an important second driver of
productivity gains, so confirming the hypothesis that embodied technological change is a crucial
determinant of productivity evolution. This relationship is expected to be particularly strong in the
low-tech sectors, where embodied technological change might be expected the main source of
productivity gains (Section 4.2).

- c.4 To see to what extent the transatlantic differences may be related to the
different sectoral structures and to the peculiar sectoral R&D/productivity relationships detectable
in the US and in the EU (Section 4.3).

- c.5 To see whether (and how much) the intensity of the R&D/productivity link is
affected by the sectoral composition and the institutional context characterising the different
European countries and regions (Section 4.4).

17

- c.6 To see whether the coefficients linking R&D and productivity are stable over
time or turn out to be affected by the business cycle (Section 4.5).

d. Discussing the implications of the empirical findings in terms of European industrial,
innovation and regional policies (Section 5).

18

2. Theoretical background and previous empirical
evidence

This section will be devoted to a discussion of the current state of the art in the economic
literature, looking both at theories and previous empirical studies. First, some macroeconomic
trends will be commented on (Section 2.1), then attention will be turned to the microeconomic
level, focusing on the different avenues of research investigating the hypotheses listed in the
previous section (Sections 2.2; 2.3; 2.4).

2.1 The transatlantic productivity gap

With the exclusion of the short-term turbulence associated with the current recession,
starting from the early 90s, the US and the EU show a persistent divide both in terms of economic
growth (see Fig. 1), labour productivity growth (see Fig. 2) and total factor productivity growth (Fig
3).

In fact, the EU15s historical process of catching-up stops around the early 90s (see
OMahony and Van Ark, 2003; Blanchard, 2004): as can be seen in Fig. 1, since 1992 to 2006 the
US GDP growth constantly overcomes the EU15 figure (with only one exception). Moreover,
average annual labour productivity growth (measured as GDP per hour worked), in the US
accelerated from 1.2% in the 1973-95 period to 2.3% in the 1996-06 period (see Van Ark et al.
2008); conversely, in the EU15 labour productivity growth declined from 2.4% in the former period
to 1.5% in the latter one (resulting in the trends shown in Fig. 2). Hence, the labour productivity
slowdown in EU15 since the 90s has reversed what was once thought as a long-term pattern of
convergence.

Most of scholars agree that to explain the transatlantic productivity gap, one has to seriously
take into account the R&D and innovation divide which emerged between the two sides of the
Atlantic with the spread of the ICT technologies (see Daveri, 2002; Crespi and Pianta, 2008). In
particular, in the second half of the 90s there was a burst of higher productivity in ICT producer

19
industries (Jorgenson et al., 2008), while in the 00 there was also a productivity surge in user
industries, including market services such as large-scale retailing and the financial and business
services (see Triplett and Bosworth, 2004; Bosworth and Triplett, 2007; Jorgenson et al., 2008).

Fig. 1: Real GDP growth in the US and the EU15: 1990-2008

Gross Domestic Product, annual growth rates
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
US EU15

Source: OECD (OECD Statistical Extracts: http://stats.oecd.org)

Indeed, these trends linked to the spread of new technologies were more marked and
accelerated in the US than in the EU (see Jorgenson et al., 2005) resulting into a widening gap in
the Total Factor Productivity (TFP) trends (see Fig. 3).

20

Fig. 2: Labour productivity growth in the US and the EU15: 1990-2008

GDP per hour worked, annual growth rates
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
US EU15

Source: OECD (OECD.Statistical Extracts: http://stats.oecd.org)

The TFP (the output growth not attributable to labour and capital growth) indicates the
efficiency with which inputs are used in the production process and so TFP differentials can be due
to intangible inputs such as R&D as well as human capital, scale economies and organizational
change (see Van Ark et al. 2008, McMorrow et al. 2009). Hence, although the underinvestment in
R&D cannot be considered the only culprit of the European delay, it plays a crucial role in the
interpretation of the transatlantic divide in terms of productivity performance and economic growth.

21
Fig. 3: TFP growth in the US and the EU15: 1990-2004

TFP, annual growth rates
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
US EU15

Source: Timmer, M.P. et al., 2003, Appendix Tables, updated June 2005

In fact - as can be seen in Fig. 4 (GERD = Gross Domestic Expenditure on R&D
1
) as far
as the total private and public expenditures in R&D are concerned, the EU has persistently invested
around the 70% of the US economy all over the last two decades.

1
GERD = BERD (Business Enterprise Expenditure on R&D) + HERD (Higher Education Expenditure on R&D) +
GOVERD (Government Expenditure on R&D) + PNPRD (Private Non-profit Expenditure on R&D).

22
Fig. 4: GERD/GDP in the US and in the EU15: 1990-2007
GERD/GDP
1.00
1.20
1.40
1.60
1.80
2.00
2.20
2.40
2.60
2.80
3.00
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
United States EU15

Source: OECD - Main Science and Technology Indicators (2009 edition)

Indeed, R&D expenditures have been demonstrated to play an important role in explaining
the productivity differentials within the industrialised countries (see Jorgenson and Stiroh, 2000;
Oliner and Sichel, 2000; Stiroh, 2002; Turner and Boulhol, 2008). In particular, the role of private
R&D investment by corporate firms (Business Enterprise Expenditure on R&D: BERD) has been
recognised as a fundamental engine for productivity growth both at the macro and microeconomic
level (see Baumol, 2002; Jones, 2002). The EU15 lags considerably and persistently behind the US
in this respect, even more strikingly than in terms of total R&D (see Fig. 5).

Hence, the EU underinvestment in total R&D and particularly in BERD might be considered
one of the main determinants of the growth, productivity and technological transatlantic gaps
discussed above. Not surprisingly, increasing R&D investment is an issue of major concern for the
European long term policy strategy. This is the rationale of the Lisbon agenda 2000 to make
Europe the most dynamic knowledge economy in the world by 2010 and of the more specific
Barcelona target which - two years later - committed the EU to reach the objective of an

23
R&D/GDP level of 3%, two thirds of which accounted for the private sector (European Council,
2002; European Commission 2002).

Fig. 5: Private R&D (BERD)/GDP in the US and in the EU15: 1990-2007

BERD/GDP
0.5
0.7
0.9
1.1
1.3
1.5
1.7
1.9
2.1
2.3
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
United States EU15

Source: OECD Main Science and Technology Indicators (2009 edition)

However, the overall European productivity delay can be explained not only by a lower
level of total and private R&D investment, but also by a lower capacity to translate R&D
investment into productivity gains, in turn fostering competitiveness and economic growth. With
regard to the latter explanation, the European economies may be still affected by a sort of Solow's
(1987) paradox, i.e. by a difficulty to translate their own investments in technology into increases in
productivity.

24
This will be one of the major hypotheses that will be investigated in this study; in fact, it
might be well the case that European economies not only invest less R&D, but also get less from
their R&D investment because of a lower R&D-productivity elasticity in the EU compared with the
US. Indeed, previous literature has shown that the R&D-productivity link is positive and significant
at the microeconomic level, but also that this relationship is stronger in the high-tech sectors. Thus,
it might be the case that the EU industrial structure (disproportionally characterised by traditional,
middle and low-tech sectors) implies a lower capacity to translate R&D efforts in productivity gains
(structural effect). However, it might be also the case that even within the same sectors
European firms reveal a lower capacity of translating R&D investments into productivity gains
(intrinsic effect).

2.2 The link between R&D and productivity at the firm and sectoral
level

Zvi Griliches (1979) started a flourishing literature devoted to investigate the relationship
between R&D and productivity at the firm and sectoral level.

On the whole, previous economic literature has found robust evidence of a positive and
significant impact of R&D on productivity at the firm level. In this literature, the estimated overall
elasticity of productivity in respect to R&D is positive, statistically significant and with a magnitude
- depending on the data and the adopted econometric methodology - ranging from 0.05 to 0.25 (for
comprehensive surveys, see Mairesse and Sassenou, 1991; Griliches 1995 and 2000; Mairesse and
Mohnen, 2001).

It is interesting to notice that the consensus about the existence of a positive and significant
impact of R&D on productivity stands on different studies using different proxies for productivity
according to the data available: labour productivity measured as the ratio between value added and
employment; labour productivity as the ratio between value added and hours worked; total factor
productivity; Solows residual; etc. (see, for instance, Hall and Mairesse 1995; Klette and Kortum,
2004; Janz et al., 2004; Lf and Heshmati, 2006; Rogers, 2006). Hence, the legacy of the previous
microeconometric literature is clear in indicating the role of R&D in enhancing productivity at the
firm level.

25

However, the intensity of the R&D-productivity relationship may widely vary across the
different industrial sectors and this opens the avenue of research that is particularly important for
the aims of this study (see previous section). Indeed, technological opportunities and appropriability
conditions are so different across sectors (see Freeman, 1982; Pavitt, 1984; Winter, 1984; Aghion
and Howitt, 1996; Dosi, 1997; Greenhalgh et al., 2001; Malerba, 2004) as to suggest the possibility
of substantial differences in the specific sectoral R&D-productivity links.

For instance, a controversial hypothesis put forward in the current debate is the alleged
advantage of low-tech compared with high-tech sectors in achieving productivity gains from R&D
investments. The argument here is that catching-up low-tech sectors are investing less in R&D but
benefit from a late-comer advantage, while high-tech sectors should be affected by a sort of
decreasing returns effect (see Marsili, 2001; von Tunzelmann and Acha, 2005; Mairesse and
Mohnen, 2005). If such was the case, we would expect a weaker relationship between R&D and
productivity growth in high-tech sectors in comparison with their low-tech counterparts.

However, this hypothesis is strikingly in contrast with the previously-available empirical
evidence. Indeed, previous sectoral studies clearly suggest a greater impact of R&D investment on
productivity in the high-tech sectors rather than in the low-tech ones.

Examples are Griliches and Mairesse (1982) and Cuneo and Mairesse (1983), who
performed two companion studies using micro-level data and making a distinction between firms
belonging to science-related sectors and firms belonging to other sectors. They found that the
impact of R&D on productivity for scientific firms (elasticity equal to 0.20) was significantly
greater than for other firms (0.10).

By the same token, Verspagen (1995) tested the impact of R&D expenditures using OECD
sectoral-level data on value added, employment, capital expenditures and R&D in a standard
production function framework. The author singled out three macro sectors: high-tech, medium-
tech and low-tech, according to the OECD classification (Hatzichronoglou, 1997). The major
finding of his study was that the impact of R&D was significant and positive only in high-tech
sectors, while for medium and low-tech sectors no significant effects could be found.

Using the methodology set up by Hall and Mairesse (1995), Harhoff (1998) studied the
R&D/productivity link - using a slightly unbalanced panel of 443 German manufacturing firms over
the period 1977-1989 - and found a significant impact ranging from a minimum of 0.068 to a

26
maximum of 0.137, accordingly to the different specifications and the different econometric
estimators adopted. Interestingly, the effect of R&D capital was considerably higher for high-
technology firms rather than for the residual groups of enterprises. In particular, for the high-tech
firms the R&D elasticity always turned out to be highly significant and ranging from 0.125 and
0.176, while for the remaining firms the R&D elasticity resulted either not significant (although
positive) or lower (ranging from 0.090 to 0.096), according to the different estimation techniques.

More recently, Wakelin (2001) applied a Cobb-Douglas production function where
productivity was regressed on R&D expenditures, capital and labour using panel data (170 UK
quoted firms during the period 1988-1992). She found that R&D expenditures had a positive and
significant role in influencing a firms productivity growth; however, in firms belonging to sectors
defined as "net users of innovations" R&D activities turned out to have a significantly larger impact
on productivity.

Rincon and Vecchi (2003) also used a CobbDouglas framework in dealing with panel
micro-data extracted from the Compustat database over the time period 1991-2001. They found that
R&D-reporting firms were more productive than their non-R&D-reporting counterparts throughout
the entire time period. However, the positive impact of R&D expenditures turned out to be
statistically significant both in manufacturing and services in the US, but only in manufacturing in
the main three European countries (Germany, France and the UK). Their estimated significant
elasticities ranged from 0.15 to 0.20.

Kwon and Inui (2003) analysed 3,830 Japanese firms with no less than 50 employees in the
manufacturing sector over the period 1995-1998, also using the methodology set up by Hall and
Mairesse (1995). Using three different estimation techniques (within estimates, first difference and
3-years differences), they found a significant impact of R&D on labour productivity, with high-tech
firms systematically showing higher and more significant coefficients than medium and low-tech
firms.

Dealing with Taiwanese data, Tsai and Wang (2004) investigated the R&D-productivity
relationship using a stratified sample of 156 large firms quoted on the Taiwan Stock Exchange, over
the period 1994-2000. They found that R&D investment had a significant and positive impact on
the growth of a firms productivity (with an average elasticity equal to 0.18). However, this impact
was much greater for high-tech firms (0.3) than for other firms (0.07).

27
Finally, Ortega-Argils et al. (2010) have looked at the top EU R&D investors, using an
unbalanced longitudinal database consisting of 577 large European companies over the period
2000-2005, extracted from the UK-DTI Scoreboards. The authors found that the R&D-productivity
coefficient was significantly different across sectors. In particular, the coefficient increased
monotonically moving from the low-tech to the medium-high and high-tech sectors, ranging from a
minimum of 0.03/0.05 to a maximum of 0.14/0.17. This outcome has been interpreted as evidence
that firms in high-tech sectors are still far ahead in terms of the impact on productivity of their R&D
investments, at least as regards top European R&D investors.

On the whole, previous sectoral empirical studies using different datasets across different
countries - seem to suggest a greater impact of R&D investments on firm productivity in the high-
tech sectors rather than in the low-tech ones. This outcome will be tested again in this study.

2.3 The role of embodied technological change and the peculiarities of
the medium and low-tech sectors

R&D is not the sole determinant of productivity gains: while the R&D input is capturing that
portion of technological change which is related to the disembodied new knowledge, gross
investment is an alternative innovative input capturing the new knowledge embodied in physical
capital, mainly machinery introduced through additional investments or simply through scrapping.
This second input represents the so-called embodied technological change, with his great potential
to positively affect productivity growth.

The embodied nature of technological progress and the effects related to its spread in the
economy were originally discussed by Salter (1960) who underlined that technological progress
might be incorporated in new vintages of capital introduced either through additional investment or
simply by scrapping
2
.

2
On the theoretical side, the embodied nature of technological change was at the core of the controversy between
Robert Solow (1960) and Dale Jorgenson (1966) with Solow arguing that embodied technological change was
dominant, hence investment was the key mechanism of economic growth, while Jorgenson arguing that from the data
available then one could not provide a clear answer. Recent empirical macroeconomic estimates actually conclude

28

Later in the literature, vintage capital models have been used by neo-Schumpeterian
economists to describe an endogenous process of innovation in which the replacement of old
equipment and machinery is the main way through which firms update their own technologies (see
Freeman et al., 1982; Freeman and Soete, 1987). More recently, the role of capital accumulation in
fostering productivity growth and economic development has been also recognised by growth
theorists (see Hulten, 1992; Greenwood et al., 1997; Hercowitz, 1998).

Moving from the macroeconomic scenario to the microeconomic analysis at the level of
sectors and firms, the important role of embodied technological change in fostering innovation and
productivity growth is even more obvious.

For instance, the literature suggests that more complex and radical product innovation
generally relies on formal R&D, while process innovation (which is often incremental rather than
radical) is much more related to embodied technical change achieved by investment in new
machinery and equipment (see Parisi et al., 2005). If such is the case, in traditional low-tech sectors
which are focusing on process innovation productivity gains might be much more related to
capital accumulation rather than to R&D expenditures. This was also one of the main message of
the well-known Pavitt taxonomy (Pavitt, 1984), where firms in traditional sectors (Supplier
Dominated) innovate mainly through embodied technological change acquired from firms in the
Specialised Suppliers sector.

Moreover, not all innovative firms are large corporations in Science Based sectors (see again
Pavitt, 1984). Indeed, economic literature supports the hypothesis that firms in traditional sectors
(most of them SMEs) face a different technological and economic environment (see Acs and
Audretsch, 1988 and 1990; Acs et al., 1994). In particular, in the low and medium-tech sectors,
R&D does not represent the sole input through which firms can achieve innovative outcomes and
productivity gains; for these firms it seems much easier to rely on the market and choose to buy
embodied technical change rather than to make their own technology (see Acs and Audretsch,
1990).

that embodied technological change is the main transmission mechanism of new technologies into economic growth
(see Greenwood et al., 1997).

29
In this framework, Santarelli and Sterlacchini (1990 and 1994) put forward convincing
empirical evidence showing the crucial role of the embodied technological change in determining
process innovation and productivity gains, especially in SMEs and in firms belonging to the
traditional sectors (the supplier dominated sectors, according to the Pavitt,1984, terminology). By
the same token and more recently, Santamara et al. (2009) found that the use of advanced
manufacturing machinery incorporated in capital formation significantly affects the probability of
engaging in both product and process innovation in firms belonging to the low and medium-low
tech sectors.

Thus, the sector to which a firm belongs represents an important analytical level for
understanding the differences in innovative processes and productivity evolution. The two
alternative patterns of technological change originally figured out by Schumpeter (creative
destruction - see Schumpeter 1934 - vs. creative accumulation, see Schumpeter, 1942) are based on
the fact that firms face sector-specific technological opportunities and appropriability conditions
(Nelson and Winter, 1982; Winter, 1984; Dosi, 1988).

In particular, firms in the high-tech sectors are more R&D based, mainly dealing with
product radical innovation and facing a competition based on the performance of products;
differently, firms in traditional middle-low tech sectors are more based on embodied technological
change, mainly dealing with incremental process innovation and facing a competition based on
costs and prices reduction (Malerba and Orsenigo, 1996; Breschi et al., 2000; Malerba, 2002; Conte
and Vivarelli, 2005).

One of the hypotheses that will be tested in this study is therefore whether productivity gains
in sectors other than the high-tech ones depend more on capital formation rather than on formal in-
house R&D.

Unfortunately, previous literature dealing with the R&D-productivity relationship (see
previous section) has generally neglected the investigation of the possible different impacts of
embodied technological change across sectors. Within a production function framework (assumed
by most of previous studies), capital is assumed to have a positive impact on productivity,
independently by the peculiarities of the investigated sectors.

One exception is the already quoted contribution by Ortega-Argils et al. (2010), where the
authors had found that the R&D-productivity coefficient was higher and more significant in the

30
high-tech sectors rather than in the middle and low-tech ones. Interestingly enough, they found that
for capital formation the results were the opposite. The physical capital stock also increased a firm's
productivity, with an overall elasticity which turned out to be around 0.12/0.13; however, this effect
was stronger in the low-tech sectors, lower but still significant in the medium-tech sectors, while it
turned out to be not significant in the high-tech sectors. Consistently with what discussed in this
section, this evidence seems to suggest that embodied technological change is crucial in the low-
tech sectors, while in the high-tech sectors technological progress is mainly introduced through in-
house R&D investments.

2.4 The business cycle and the regional peculiarities

Differently from the issues discussed in the previous sections, there is not an established
literature devoted to investigate possible differences in the firms R&D/productivity elasticity along
the business cycle. The closer strand of literature is that one devoted to study the link between
output and innovation activity, proxied by R&D or other innovative indicators.

In particular, a long standing debate (flourishing in the 80s and 90s) opposed the
proponents of the downturn as the optimal time to introduce innovations with those advocating that
expansionary periods were actually more likely to sustain innovation and diffusion.

The argument put forward by the formers (see Mensch, 1979 and Kleinknecht, 1987) was
that the value of profitability from existing products and processes falls in recessions and so, if this
decline is large enough in comparison with the potential returns to be gained from implementing
new products and processes, firms will implement innovations during cyclical downturns.

The argument put forward by the latter (Freeman et al, 1982 and Freeman and Soete, 1987)
was that the upswing conditions are those that facilitate the large scale introduction and diffusion of
innovation, since firms are confident in terms of sales expectations and less constrained from
financial perspectives. This view is fully consistent with the so called theory of the demand-pull
innovation. Indeed, that rising demand may induce an increase in the innovation effort is a rather
old issue (Schmookler, 1962 and 1966; Scherer, 1982): on the one hand, increasing sales permit the

31
financing of expensive and uncertain R&D activities and relax possible credit constraints, while on
the other hand optimistic sale expectations increase expected profitability from innovation.

Previous empirical studies are massively in favour of the second strand of literature both at
the macroeconomic and the microeconomic level. For instance, Geroski and Walters (1995), using
macroeconomic time series for the UK over the period 1948-83, found significant evidence that
major innovations and patents are pro-cyclical. By the same token, but at the microeconomic level,
Brouwer and Kleinknecht (1996) on Dutch data, Crpon et al. (1998) on French data, and Piva and
Vivarelli (2007) on Italian data found a significant confirmation of the demand-pull hypothesis at
the level of the firm.

However, as far as we know, previous literature has not investigated whether the upswings
are not only favourable to the introduction and diffusion of innovation, but also conducive of a
more effective impact of R&D and innovation on productivity; this will be the hypothesis that will
be tested in this study.

Turning the attention to the regional dynamics, we also find a lack of a previous literature
specifically addressed to the hypothesis that will be investigated in this study.

Our starting point is that regional peculiarities in the relationship between R&D and capital
formation on the one hand and productivity evolution on the other hand should heavily be affected
by the industrial structure which characterises the single region. Thus - according to what discussed
in the previous Section 2.3 - a region characterised by a large presence of high-tech sectors would
probably turn out to be very sensitive to R&D activities in getting productivity gains, while a region
characterised by a disproportionate presence of traditional sectors and SMEs would come out to be
particularly responsive to capital formation. To the best of our knowledge, the literature in regional
economics has not yet investigated possible inter-regional differences in the relationship between
local R&D stock and local labour productivity.

However, recent empirical works has shown how the endogenous growth approach can be
applied at the regional level, underlining the crucial role of knowledge stock (proxied by either
R&D or patents) and human capital in explaining the differences in TFP across regions (see, for
instance, Dettori et al., 2008, studying 199 European regions over the period 1985-2006; Fischer et
al., 2008, analysing 203 European regions over the period 1997-2002; Gumbau-Albert and Maudos,

32
2006, investigating 17 Spanish regions over the period 1986-96; Bronzini and Piselli, 2009,
studying 19 Italian regions over the period 1985-2001).

Going a step further, in this study we will try to see whether the size and significance of the
demonstrated link between knowledge and capital stocks on the one hand and productivity on the
other hand vary across different regions. This will be made possible by the availability of a large
micro-firm database which allows us to test possible significant differences across groups of firms
belonging to different European countries and macro-regions.

33
3. Data and methodology

This section will be organised in three parts; in Section 3.1 the data source will be described
and its characteristics and limitations discussed; in Section 3.2, attention will be devoted to the
process that was developed in order to build a consistent and reliable dataset; finally, Section 3.3
will introduce the econometric specification tested in the empirical part of the study, whose results
will be put forward in Section 4.

3.1 The data source

The microdata used in this study were provided by the JRCIPTS (Joint Research Centre-
Institute for Prospective Technological Studies) of the European Commission, extracted from a
variety of sources, including companies annual reports.

Available data includes:

- Company identification, name and address and industry sector;

- Financial data;

- Fundamental economic data, including the crucial information for this study,
namely: sales, cost of goods (the difference between the former and the latter allows to
obtain value added), capital formation, R&D expenditures, and employment.

Data are filed in current national currencies.

Given the crucial role assumed by the R&D variable in this study, it is worthwhile to discuss
in detail what is intended as R&D in companies' annual reports. This item represents all costs
incurred during the year that relate to the development of new products and services. It is important
to notice that this amount is only the companys contribution and exclude amortization and
depreciation of previous investments, so being a genuine flow of current in-house R&D
expenditures.

34
In particular the figure excludes customer or government-sponsored R&D expenditures, as
well as engineering expenses or market research and testing expenses. On the whole, the adopted
definition of R&D is quite restrictive and refers to the pure flow of current additional resources
coming from internal sources and devoted to the launch and development of entirely new products.

The number of years available for each company depends upon the companys history; thus,
the data source is unbalanced in nature.

3.2 The construction of the dataset

Once acquired the rough data from IPTS (defined and organised as reported in the previous
sub-section), we proceeded in the construction of a longitudinal database that would be adequate to
run panel estimations addressed to test the theoretical hypotheses discussed in the second section of
this report. For sake of simplicity, we will describe the adopted complex procedure by steps. Where
appropriate, some of the data managing analyses were carried out in close cooperation with IPTS
statisticians, whose feedbacks were extremely useful to shape the final panel dataset.

First step: data extraction

In guiding the individuation of R&D performers by IPTS, the following criteria have been
adopted:
- Selecting only those companies with R&D>0 in, at least, one year in a 20 years time
span;
- Selecting only those companies located in the US and in the EU 27 countries;
- Individuating information concerning R&D, sales, cost of goods (the difference between
sales and cost of goods allowed to obtain value added), capital formation, R&D
expenditures, and employment. In more detail, this is the list of the available information
for each firm included in the obtained workable dataset:
o Country (according to the location of the headquarter);

35
o Industry code at 2008;
o R&D expenses (defined as discussed in the previous sub-section);
o Capital expenditures;
o Net turnover;
o Cost of goods sold;
o Employees.

- All the value data were expressed in the current national currency in millions (for
instance: countries which are currently adopting Euro have values in Euro for the entire
examined period).

Second step: deflation of current nominal values

Nominal values were commuted into constant price values trough GDP deflators (source:
IMF) centred in year 2000. For a tiny minority of firms reporting in currencies different from the
national ones, we opted for deflating the nominal values through the national GDP deflator, as well.

Third step: values in PPP dollars

Once obtained constant 2000 prices values, all figures were converted into US dollars using
the PPP exchange rate at year 2000 (source: OECD)
3
. 9 companies from 4 countries (Lithuania,

3
This procedure is consistent with what suggested by the Frascati Manual (OECD, 2002) in order to correctly adjust
R&D expenditures for differences in price levels over time (i.e. intertemporal differences asking for deflation) and
among countries (i.e. interspatial differences asking for a PPP equivalent). In particular ...the Manual recommends the

36
Latvia, Malta and Romania) were excluded, due to the unavailability of PPP exchange rates from
the OECD. The 10 companies reporting in euro but located in non-euro countries (Denmark,
Estonia and the UK) were excluded as well
4
; while the 58 European companies reporting in US
dollars were kept as such.

Fourth step: the format of the final data string

The obtained unbalanced database comprises 2,777 companies, 2 codes (country and sector)
and 5 variables (see the bullet points above) over a period of 19 years (1990-2008).
Since one of the main purposes of this study is to distinguish across high-tech and
medium/low-tech sectors (see Sections 2.2 and 2.3), a third code was added, labelling as High-tech
the following sectors
5
:

- SIC 283: Drugs (ISIC Rev.3, 2423: Pharmaceuticals);

- SIC 357: Computer and office equipments (ISIC Rev. 3, 30: Office, accounting and
computing machinery);

- SIC 36 (excluding 366): Electronic and other electrical equipment and components,
except computer equipment (ISIC Rev. 3, 31: Electrical machinery and apparatus);

- SIC 366: Communication equipment (ISIC Rev. 3, 32: Radio, TV and communications
equipment);

use of the implicit gross domestic product (GDP) deflator and GDP-PPP (purchasing power parity for GDP), which
provide an approximate measure of the average real opportunity cost of carrying out the R&D. (ibidem, p. 217).
4
Given the very small number of firms involved, it was decided not to take the arbitrary choice of using either the
national or the Euro PPP converter.
5
The standard OECD classification was taken (see Hatzichronoglou, 1997) and extended it including the entire
electrical and electronic sector 36 (considered as a medium-high tech sector by the OECD). We opted for this extension
taking into account that we just compare the high-tech sectors with all the other ones and that we need an adequate
number of observations within the sub-group of the high-tech sectors.

37
- SIC 372-376: aircraft and spacecraft (ISIC Rev. 3, 353: Aircraft and spacecraft);

- SIC 38: measuring, analyzing and controlling instruments (ISIC Rev. 3, 33: Medical,
precision and optical instruments)

Finally, since we are interested in investigating possible regional peculiarities within the EU
(see Section 2.4), a regional code was assigned to each firm according to the NUTS 1 classification.

Fifth step: computation of the R&D and capital stocks.

Consistently with the reference literature (see Section 2), the methodology adopted in this
study (see also next Section 3.3) requires to compute the R&D and capital stocks, accordingly with
the perpetual inventory method. In practice, the following two formulas have to be applied:

(1)
) (
&
0
0
o +
=
g
D R
K
t
t
and
t t t
D R K K & ) 1 (
1
+ =

o
where R&D = R&D expenditures

(2)
) (
0
0
o +
=
g
I
C
t
t
and
t t t
I C C + =

) 1 (
1
o
where I = gross investment

where g is generally computed as the ex ante pre-sample compounded average growth rate
of the corresponding flow variable and is a depreciation rate.

However, the reader has to be reminded that our dataset is spanning 19 years and it is
unbalanced in nature. This means that only a minority of firms display continuous information all

38
over the entire period, while many firms have information only for one or more spans over the
1990-2008 period and these spans may be either very short or even isolated data. In addition, many
firms display left-truncated data; for instance, the majority of European firms have data only for the
most recent years.

Given the unbalanced structure of the dataset, to strictly apply the formulas (1) and (2) to
compute initial stocks (using say the first three years to obtain the ex-ante growth rates) would
have implied to lose a huge amount of information. In the best case - say a firm with a complete set
of 19 data over the period - this methodology would have implied to lose 3 observations out of 19;
in the worst case - say a firm characterized by data available only for some spells of three years
each this computation would have implied to lose all the available information for that particular
firm.

In order to avoid this massive dropping of available data, we adopted the following criteria.
First, it was decided to compute a rate of growth using the initial three years of a given spell and
then apply it to the initial flow and not to the fourth year (that is our t0 is the very first year of the
spell and so g is an ex post 3-year compound growth rate). Second, it was iteratively applied this
methodology to all the available spans of data comprising at least three consecutive years
6
. The
combination of these two choices allowed us to keep all the available information, with the only
exceptions of either isolated data or pairs of data.

Although departing from the usual procedure, to rely on ex-post growth rates appears
acceptable in order to save most of the available information in the dataset; however, the impact of
this choice on the values assumed by the stocks is limited, since they are also affected by the flow
values and the depreciation rates. Finally, the chosen growth rate affects only the initial stock and
its impact quickly smoothes out as far as we move away from the starting year
7
.

6
This means that for firms characterised by breaks in the data we computed different initial stocks, one for each
available time span, consistently with what done by Hall (2007); however, differently from Hall (2007), we consider the
different spans as belonging to the same firm and so we will assign in the following econometric estimates a single
fixed or random effect to all the spans belonging to the same company history.
7
Options for the choice of g - different from the standard one - have been implemented by other authors, as well. For
instance, Parisi et al. (2006), assume that the rate of growth in R&D investment at the firm level in the years before the
first positive observation equals the average growth rate of industry R&D between 1980 and 1991 (the time-span
antecedent to the longitudinal micro-data used in their econometric estimates). In general terms, the choice of a feasible
g does not significantly affect the final econometric results of the studies. As clearly stated by Hall and Mairesse (1995,

39

Hence, - in order to be able to compute R&D and capital stocks according to the procedure
described above only R&D and capital expenditure flows data with at least 3 observations in
consecutive years were retained. This implied that 354 companies (mainly European) had to be
dropped because of lacking 3 R&D observations in successive years and 30 additional companies
for lacking 3 capital expenditures observations in successive years. Thus, a total of 2,393 firms were
retained at the end of this stage of the cleaning process.

Turning the attention to the depreciation rates (), we differentiated both between R&D and
capital and between the high-tech sectors vs. the other sectors, taking into account what is common
in the reference literature which assumes = 6% for computing the capital stock and = 15% for
computing the R&D stock (see Nadiri and Prucha, 1996 for the capital stock; Hall and Mairesse,
1995 and Hall, 2007 for the R&D stock).

Indeed, depreciation rates for the R&D stocks have to be assumed to be higher than the
corresponding rates for physical capital, since it is assumed that technological obsolescence is more
rapid than the scrapping of physical capital.

However, depreciation rates for the high-tech sectors have to be assumed to be higher than
the corresponding rates for medium and low-tech sectors under the assumption that technological
obsolescence both related to R&D efforts and to the embodied technologies incorporated in
physical capital - is faster in the high-tech sectors. Specifically, depreciation rates were assumed
equal to 6% and 7% with regard to physical capital in the low-medium and high-tech sectors
respectively, while the corresponding for R&D stocks were assumed equal to 15% and 18%
respectively.

Once computed according to the formulas (1) and (2) and the adopted g and rates, the
resulting stocks were checked and negative ones were dropped
8
. Moreover, we excluded a minority
of unreliable data such as those indicating negative sales and cost of goods equal to zero.

After these further drops of data, we ended up with 1,884 companies (1,210 US and 674 EU,
for a total of 17,064 observations).

p.270, footnote 9): In any case, the precise choice of growth rate affects only the initial stock, and declines in
importance as time passes,....
8
The occurrence of negative stocks happens when g turns out to be negative and larger in absolute value than .

40
Sixth step: outliers.

At this point, in order to check for the presence of outliers (i.e. observations that appear to
deviate markedly in terms of standard deviations from the relevant mean, possibly implying a bias
in the econometric estimates), the Grubbs test (Grubbs, 1969) was run on the two critical variables
in the analysis: the R&D stock (K) and the physical capital stock (C).

Since the outlier test has to be applied to the variables used in the regression analysis, the
test was run on the two normalised stock variables: K/E and C/E (see eq. 5 in Section 3.3).

In detail, the Grubbs test - also known as the maximum normed residual test, (Grubbs, 1969;
Stefansky, 1972) - is used to detect outliers in a dataset, either creating a new variable or dropping
outliers out of the data set. Technically, the Grubbs test detects one outlier at each iteration
9
: the
outlier is expunged from the data set and the test is iterated until no outliers remain.

The Grubbs test is defined under the null hypothesis (H
0
) that there are no outliers in the
dataset; the test statistic is:
(3)
s
Y Y
G
i
N i
=
= ,.., 1
max

with Y and s denoting the sample mean and standard deviation, respectively. Therefore, the
Grubbs test detects the largest absolute deviation from the sample mean in units of the sample
standard deviation
10
.
With a two-sided test, the null hypothesis of no outliers is rejected if:

9
The default number of iterations is 16,000.
10
The Grubbs test can also be defined as one of the following one-sided tests:
- test whether the minimum value is an outlier:
s
Y Y
G
min
= with Ymin denoting the minimum value;

- test whether the maximum value is an outlier:
s
Y Y
G

=
max
with Ymax denoting the maximum value.

41
(4)
( )
) 2 ), 2 /( (
2
) 2 ), 2 /( (
2
2
1
>
N N
N N
t N
t
N
N
G
o
o

with ) 2 ), 2 /( (
2
N N t o denoting the critical value of the t-distribution with (N-2) degrees of
freedom and a significance level of o/(2N).

After running the Grubbs test, 426 observations turned out to be outliers for the K/E variable
and 613 for the C/E variable (54 outliers turned out to be common to both the variables).

Therefore, at the end of the process, we ended up with a final dataset comprising 1,809
companies (1,170 US and 639 EU, for a total of 16,079 observations).
Table 1 reports the distribution of the retained firms and observations across countries.
Tab. 1: Distribution of firms and observations across countries in the final version of the dataset
COUNTRY FIRMS OBSERVATIONS
AUSTRIA 16 51
BELGIUM 20 82
CZECH REPUBLIC 1 4
DENMARK 21 152
ESTONIA 1 3
FINLAND 41 157
FRANCE 54 279
GERMANY 141 749
GREECE 11 41
HUNGARY 3 12
IRELAND 8 55
ITALY 5 19
LUXEMBOURG 3 9
NETHERLANDS 25 165
SLOVENIA 1 4
SPAIN 3 7
SWEDEN 62 386
UNITED KINGDOM 223 1,299
EU 639 3,474
USA 1,170 12,605
Total 1,809 16,079

42
3.3 The econometric specification and descriptive statistics

Consistently with the previous studies discussed in Section 2, we will test the following
augmented production function, obtainable from a standard Cobb-Douglas function in three inputs:
physical capital, labour and knowledge capital (see Hall and Mairesse, 1995, formulas 1-2-3, pp.
268-69)
11
.

(5) c | o + + + + = ) ln( ) / ln( ) / ln( ) / ln( E E C E K E VA

Our proxy for productivity is labour productivity (Value Added, VA, over total employment,
E); our pivotal impact variables are the R&D stock (K) per employee and the physical capital stock
(C) per employee.

As it is common in this type of literature (see Hulten, 1990; Jorgenson, 1990; Hall and
Mairesse, 1995; Parisi et al., 2006), stock indicators rather than flows were considered as impact
variables; indeed, productivity is affected by the cumulated stocks of capital and R&D expenditures
and not only by current or lagged flows.

Moreover, dealing with R&D stocks - rather than flows - has two additional advantages: on
the one hand, since stocks incorporate the cumulated R&D investments in the past, the risks of
endogeneity is minimised; on the other hand, there is no need to deal with the complex (and often
arbitrary) choice of the appropriate structure of lags for the R&D regressor.

In this framework, R&D and physical capital stocks were computed using the perpetual
inventory method, according to the formulas (1) and (2) introduced and discussed in the previous
sub-section (fifth step).

Finally, taking per capita values permits both standardisation of our data and elimination of
possible size effects (see, for example, Crpon et al., 1998, p.123). In this framework, total

11
As clearly stated and demonstrated in Hall and Mairesse (1995), the direct production function approach to measure
returns to R&D capital is preferred on other possible alternative specifications.

43
employment (E) is a control variable: if turns out to be greater than zero, it indicates increasing
returns.

All the variables are taken in natural logarithms.

While K/E (R&D stock per employee) captures that portion of technological change which
is related to the cumulated R&D investment, C/E (physical capital stock per employee) is the result
of the cumulated investment, implementing different vintages of technologies. So, this variable
encompasses the so-called embodied technological change, possibly affecting productivity growth
(see Section 2.3).

The following Table 2 reports the correlation matrix of the variables included in eq. 5. As
can be seen, a preliminary evidence of the expected positive impacts of both K/E and C/E upon
VA/E emerges. Moreover, no evidence of possible serious collinearity problems comes out, since
the three relevant correlation coefficients turn out to be less than 0.285 in absolute values.

Tab. 2: Correlation table: correlation coefficients

Log(Value
added per
employee)
Log(R&D
stock per
employee)
Log(Physical
stock per
employee)
Log(Employment)
Log(Value added
per employee)
1
Log(R&D stock
per employee)
0.451 1
Log(Physical stock
per employee)
0.278 0.252 1
Log(Employment)

-0.040 -0.284 0.209 1

Note: all correlation coefficients are 1% significant.

44

Specification (5) was estimated through different estimation techniques.

Firstly, pooled ordinary least squared (POLS) regressions were run to provide preliminary
reference evidence. Although very basic, these POLS regressions were controlled for
heteroskedasticity (we used the Eicker/Huber/White sandwich estimator to compute robust standard
errors) and for a complete set of three batteries of dummies, namely country (19 countries), time (19
years) and sector (52 two-digit SIC-sectors) dummies.

Secondly, fixed effect (FE) regressions were performed in order to take into account the firm
specific unobservable characteristics such as managerial capabilities. The advantage of the FE
estimates is that different firms are not pooled together but taken into account in their own
singularity. The disadvantage is that country and sector dummies are dropped for computational
reasons, since they are encompassed by the individual dummies.

Thirdly, random effect (RE) regressions were run to provide the more complete results,
where both individual (randomized) effects are taken into account together with the possibility to
retain all the entire batteries of dummies.

The following Tables 3, 4, 5 and 6 report means and standard deviations of the four relevant
variables in eq.5. We will refer to them when appropriate in the following Section 4 that is
devoted to discuss the econometric results.

45
Tab. 3: VA/E (Value Added/Employees)
Mean Standard
deviation
Whole sample (16,079) 102.781 91.008
US (12,605) 108.793 96.475
EU (3,474) 80.965 62.912
Manufacturing (12,876) 99.565 92.914
High-tech manufacturing sectors (7,693) 112.038 108.275
Other manufacturing sectors (5,183) 81.050 58.938
Service sectors (3,203) 115.709 81.648
US Manufacturing sectors (10,214) 104.18 98.355
EU Manufacturing sectors (2,662) 81.324 65.678
US High-tech manufacturing sectors (6,462) 116.125 112.525
EU High-tech manufacturing sectors (1,231) 90.583 79.089
US Other manufacturing sectors (3,752) 83.983 61.733
EU Other manufacturing sectors (1,431) 73.359 50.093
US Service sectors (2,391) 127.907 86.000
EU Service sectors (812) 79.789 52.858
North+UK EU (1,994) 79.252 65.449
Other EU (1,480) 83.274 59.265
Other EU (no south) (1,413) 83.380 60.147
North+UK EU Manufacturing sectors (1,534) 78.744 67.937
Other EU Manufacturing sectors (1,128) 84.833 62.332
Other EU (no south) Manufacturing sectors (1,097) 84.646 62.927
North+UK EU Service sectors (460) 80.947 56.399
Other EU Service sectors (352) 78.276 47.875
Other EU (no south) Service sectors (316) 78.984 49.130
North+UK EU High-tech manufacturing sectors (734) 92.236 80.394
Other EU High-tech manufacturing sectors (497) 88.141 77.136
Other EU (no south) High-tech manufacturing sectors (482) 87.578 78.090
North+UK EU Other manufacturing sectors (800) 66.365 51.044
Other EU Other manufacturing sectors (631) 82.227 47.439
Other EU (no south) Other manufacturing sectors (615) 82.349 47.740
EU Innovative NUTS (1,827) 80.851 61.520
EU Weakly Innovative NUTS (1,604) 80.872 64.339
EU Innovative NUTS High-tech manufacturing sectors (688) 85.962 75.918
EU Weakly Innovative NUTS High-tech manufacturing sectors (529) 95.461 82.379
EU Innovative NUTS Other manufacturing sectors (670) 73.478 45.138
EU Weakly Innovative NUTS Other manufacturing sectors (749) 73.206 54.370
EU Innovative NUTS Manufacturing sectors (1,358) 79.802 62.939
EU Weakly Innovative NUTS Manufacturing sectors (1,278) 82.418 68.247
EU Innovative NUTS Service sectors (469) 83.887 57.169
EU Weakly Innovative NUTS Service sectors (326) 74.814 45.500
Business Cycle Recessions (8,073) 107.640 96.134
Business Cycle Expansions (8,006) 98.462 85.317
Business Cycle US Recessions (6,522) 113.971 101.862
Business Cycle EU Recessions (1,551) 78.018 58.576
Business Cycle US Expansions (6,083) 103.241 90.021
Business Cycle EU Expansions (1,923) 83.342 66.122
Note: the number of observations is reported in brackets.

46
Tab. 4: K/E (R&D Stock/Employees)
Mean Standard
deviation
Whole sample (16,079) 86.076 105.899
US (12,605) 93.467 110.310
EU (3,474) 59.267 82.701
Manufacturing (12,876) 82.470 106.904
Service sectors (3,203) 100.574 100.478
North+UK EU (1,994) 61.618 93.790
Other EU (1,480) 56.083 64.726
Other EU (no south) (1,413) 57.343 65.292

47
Tab. 5: C/E (Physical capital Stock/Employees)
Mean Standard
deviation
Whole sample (16,079) 81.026 80.542
US (12,605) 81.567 79.633
EU (3,474) 79.065 83.742
Manufacturing (12,876) 84.886 81.585
Service sectors (3,203) 65.512 74.222
North+UK EU (1,994) 72.177 85.371
Other EU (1,480) 88.346 80.597
Other EU (no south) (1,413) 88.174 80.048

48
Tab. 6: E (Employees)
Mean Standard
deviation
Whole sample (16,079) 11,204 35,302
US (12,605) 9,124 31,064
EU (3,474) 18,752 46,846
Manufacturing (12,876) 11,951 35,250
High-tech manufacturing sectors (7,693) 8,179 23,264
Other manufacturing sectors (5,183) 17,551 47,237
Service sectors (3,203) 8,199 35,356
US Manufacturing sectors (10,214) 9,714 31,116
EU Manufacturing sectors (2,662) 20,535 46,937
US High-tech manufacturing sectors (6,462) 7,298 21,294
EU High-tech manufacturing sectors (1,231) 12,803 31,259
US Other manufacturing sectors (3,752) 13,876 42,752
EU Other manufacturing sectors (1,431) 27,187 56,244
US Service sectors (2,391) 6,600 30,718
EU Service sectors (812) 12,908 46,096
North+UK EU (1,994) 10,699 29,793
Other EU (1,480) 29,603 61,253
Other EU (no south) (1,413) 29,700 60,858
North+UK EU Manufacturing sectors (1,534) 11,182 23,177
Other EU Manufacturing sectors (1,128) 33,254 64,731
Other EU (no south) Manufacturing sectors (1,097) 33,477 64,831
North+UK EU Service sectors (460) 9,086 45,352
Other EU Service sectors (352) 17,902 46,643
Other EU (no south) Service sectors (316) 16,587 41,907
North+UK EU High-tech manufacturing sectors (734) 9,005 21,601
Other EU High-tech manufacturing sectors (497) 18,412 40,996
Other EU (no south) High-tech manufacturing sectors (482) 18,949 41,514
North+UK EU Other manufacturing sectors (800) 13,180 24,376
Other EU Other manufacturing sectors (631) 44,944 76,563
Other EU (no south) Other manufacturing sectors (615) 44,864 76,529
EU Innovative NUTS (1,827) 17,420 44,445
EU Weakly Innovative NUTS (1,604) 20,725 49,905
EU Innovative NUTS High-tech manufacturing sectors (688) 12,564 24,780
EU Weakly Innovative NUTS High-tech manufacturing sectors (529) 13,442 38,376
EU Innovative NUTS Other manufacturing sectors (670) 24,769 52,522
EU Weakly Innovative NUTS Other manufacturing sectors (749) 29,698 59,672
EU Innovative NUTS Manufacturing sectors (1,358) 18,585 41,329
EU Weakly Innovative NUTS Manufacturing sectors (1,278) 22,969 52,524
EU Innovative NUTS Service sectors (469) 14,046 52,338
EU Weakly Innovative NUTS Service sectors (326) 11,928 36,695
Business Cycle Recessions (8,073) 11,127 35,329
Business Cycle Expansions (8,006) 11,282 35,276
Business Cycle US Recessions (6,522) 9,576 32,054
Business Cycle EU Recessions (1,551) 17,648 46,095
Business Cycle US Expansions (6,083) 8,639 29,961
Business Cycle EU Expansions (1,923) 19,643 47,437

49

4. Econometric analysis

4.1 Overall results; EU vs. US

From Table 3 we get a further confirmation of the US/EU productivity gap that has been
discussed from a macroeconomic point of view in Section 2.1. As can be seen, the US advantage in
labour productivity homogeneously emerges both in aggregate and within the different sectoral
groups: 109 vs. 81 in the whole sample; 104 vs. 81 in manufacturing; 116 vs. 90 in the high-tech
manufacturing sectors; 84 vs. 73 in the other manufacturing sector; 128 vs. 80 in the service sectors.
In this and the following sub-Section 4.2, we will try to provide some explanations of these obvious
differentials.

Table 7
12
provides the overall results concerning the whole sample of 1,809 firms (16,079
observations). As can be seen, consistently with the previous literature (see Section 2.2), we found
robust evidence of a positive and significant impact of R&D on productivity with an elasticity
ranging from 0.089 to 0.205, according to the different adopted estimation techniques. As discussed
in Section 2.2, in the reference literature the estimated overall elasticity of productivity in respect to
R&D is positive, statistically significant and with a magnitude - depending on the data and the
adopted econometric methodology - ranging from 0.05 to 0.25; hence, the obtained estimates are
within the bounds set by the previous empirical studies.

As far as physical capital is concerned, here again we have no surprise in assessing a
positive and significant impact ranging from 0.093 to 0.115.

12
In Section 4 only summary tables will be displayed, reporting the values and the statistical significance of the sole
relevant coefficients (one star means significance at the 90% level of confidence, two stars at the 95% level, three stars
at the 99% level). The full econometric results, including the estimates of the constant and the employment regressor,
the t-statistics values and the outcomes of the diagnostic overall statistical tests are reported in the complete tables
displayed in the Appendix A, where Table A7 corresponds to Table 7 in the text, Table A8 to Table 8 in the text and so
on.

50
The whole sample estimates will be the reference for all the following analyses and the
correspondent results will be reported in the left panel of all the following tables.

Tab. 7: Whole sample, US and EU

Whole sample US EU
POLS FE RE POLS FE RE POLS FE RE
Log(R&D
stock per
employee)
0.205*** 0.089*** 0.107*** 0.228*** 0.098*** 0.119*** 0.144*** 0.058*** 0.074***
Log(Physical
stock per
employee)
0.115*** 0.093*** 0.099*** 0.106*** 0.100*** 0.102*** 0.125*** 0.053*** 0.078***
Obs. 16,079 12,605 3,474
N. of firms 1,809 1,170 639

As discussed in the previous Section 1 and 2.1, a first research hypothesis of this study is
that the lower European economic performance in comparison with the US can be explained not
only by a lower level of public and corporate R&D investment, but also by a lower capacity to
translate R&D investment into productivity gains.

This hypothesis can be tested running specification 5 separately for the US and the EU firms
(1,170 vs. 639 companies). As can be seen in the second and third panel of Table 7, data seems to
fully confirm the proposed hypothesis. Although uniformly positive and statistically significant at
the 99% level of confidence, the R&D coefficients for the US firms turn out to be consistently
larger than the corresponding coefficients for the European firms. Indeed, the three estimation
techniques consistently provide European elasticities equal to about 60% of their US counterparts.

We interpret these unambiguous results as a clear evidence of the better ability of US firms
in translating R&D investments in productivity gains and as a signal of a gap of efficiency that
European firms and European policy have to deal with (see the following sub-sections of this
Section and the conclusive Section 5).

As far as the productivity impact of the physical capital, POLS and FE/RE estimates tell us
different stories in terms of the US-EU comparison. However, if we rely on the more reliable

51
methodologies controlling for the idiosyncratic effects, it appears that the US reveals an advantage
similar to the one emerged for the intangible R&D investments.

Thus, US firms result more efficient in getting productivity gains both from the R&D and
the physical capital investments; these aggregate results will be further analysed in the following
Section 4.3, where we will jointly take into account the geographical and the sectoral dimension of
our available data.

4.2 A sectoral breakdown

The tables 8
13
and 9 report the results split by sectors, according to what discussed in
Section 2.2; so we have manufacturing firms vs. service ones and in turn manufacturing split into
high-tech and other sectors.

Tab. 8: Sectoral decomposition: Manufacturing (High-tech + Other) and Service sectors

Whole sample Manufacturing
sectors
Service
sectors
Log(R&D
stock per
employee)
0.205*** 0.089*** 0.107*** 0.209*** 0.073*** 0.095*** 0.177** 0.113*** 0.127***
Log(Physical
stock per
employee)
0.115*** 0.093*** 0.099*** 0.109*** 0.086*** 0.094*** 0.147*** 0.115*** 0.125***
Obs. 16,079 12,876 3,203
N. of firms 1,809 1,383 426

Somehow surprising, from Table 8 we learn that - at least in the more reliable FE and RE
estimates services firms exhibit higher coefficients than manufacturing firms. Since most of

13
See Table A8 in the appendix for the corresponding full set of econometric results; as explained in the previous note,
this remark also applies to all the following tables discussed in Section 4.

52
company R&D expenditures are performed within the manufacturing sectors, this evidence may
suggest a sort of late comer effect enjoyed by the service sectors where more limited amounts of
R&D stock are able to exhibit a larger impact on firms productivity.

Less surprising is the outcome concerning the role of the embodied technological change,
where the coefficients regarding the service firms are uniformly larger that those related to the
manufacturing firms. Indeed, services implement new technologies through machinery and
equipment coming from the manufacturing sectors (think for instance - to the diffusion of the ICT
technologies).

Turning the attention to the manufacturing firms, the reader would recall that a second
hypothesis of this work (see previous Sections 1, 2.2 and 2.3) was that R&D might be crucial in
fostering productivity in the high-tech sectors, while being less important in the rest of the
manufacturing sectors where alternative sources of productivity growth such as technological
change embodied in physical capital should play a dominant role.

The first statement of this hypothesis seems to be supported by what reported in the
following Table 9. As can be seen, the R&D coefficients related to the high-tech manufacturing
sectors are always larger than the corresponding ones for the other manufacturing sectors. These
outcomes are consistent with the empirical outcomes from the previous literature at the sectoral
level (see Section 2.2). On the whole, high-tech sectors not only invest more in R&D and are
characterised by a higher productivity performance
14
, but also achieve more in terms of productivity
gains from their own research activities. Indeed, our results show that firms in high-tech sectors are
still far ahead in terms of the efficiency through which R&D investments affect productivity
growth.

Differently, the second part of the sectoral hypothesis put forward in Sections 1 and 2.3,
seems not to be supported by the data. In contrast with our expectations, embodied technological
change appears to be more effective in the high-tech sectors rather than in the rest of the
manufacturing sectors. In fact, in all the three estimates, the coefficients turn out to be higher in the
high-tech sectors. While this result deserves further investigation (for instance, in distinguishing the
European from the American firms: see next Section 4.3), at this stage it has to be noticed that high-

14
See previous Tables 3 and 4: high-tech manufacturing sectors exhibit a value of 111 with regard to K/E and 112 with
regard to VA/E vs 40 and 81 in the other manufacturing sectors.

53
tech sectors exhibit a relative efficiency advantage in translating both R&D expenditures and capital
formation into productivity gains.

Tab. 9: Sectoral decomposition: High-tech and Other manufacturing sectors

Whole sample High-tech manufacturing
sectors
Other manufacturing
sectors
Log(R&D
stock per
employee)
0.205*** 0.089*** 0.107*** 0.236*** 0.070*** 0.099*** 0.158*** 0.053*** 0.071***
Log(Physical
stock per
employee)
0.115*** 0.093*** 0.099*** 0.117*** 0.092*** 0.100*** 0.093*** 0.069*** 0.078***
Obs. 16,079 7,693 5,183
N. of firms 1,809 804 579

4.3 The US/EU comparison: crossing the geographical and the sectoral
dimensions

The aim of this sub-section is to deepen the analysis taking jointly into account the
European/American divide and the sectoral dimension of the available data. This perspective will
allow us to better interpret the aggregate differential discussed in the previous sub-sections 4.1 and
4.2. For instance, it may well be the case that the US advantage in terms of R&D efficiency is
totally due to an advantage in the high-tech sectors or in the entire manufacturing sectors or
differently that this advantage is detectable across all the sectors of the economy. By the same
token, it might be the case that the aggregate sectoral differentials discussed in the previous
subsections are confirmed in the US and not in Europe or vice versa or instead, confirmed in both
the geographical areas.

54
Table 10 displays the US/EU comparison with regard to the manufacturing sectors only. As
it is obvious, the aggregate European gap in terms of efficiency (see Table 7) is fully confirmed: as
for the whole economy, in the manufacturing sectors the four relevant US coefficients are uniformly
larger that their European counterparts.

Interestingly enough, Table 11 focusing on the service sectors tells us exactly the same
story, confirming the US advantage across all the coefficients.

Tab. 10: US versus EU: Manufacturing sectors

Whole sample US
Manufacturing sectors
EU
Log(R&D
stock per
employee)
0.205*** 0.089*** 0.107*** 0.228*** 0.078*** 0.103*** 0.147*** 0.052*** 0.070***
Log(Physical
stock per
employee)
0.115*** 0.093*** 0.099*** 0.099*** 0.089*** 0.093*** 0.135*** 0.059*** 0.086***
Obs. 16,079 10,214 2,662
N. of firms 1,809 914 469

Tab. 11: US versus EU: Service sectors

Whole sample US
Service sectors
EU
Service sectors
Log(R&D
stock per
employee)
0.205*** 0.089*** 0.107*** 0.215*** 0.125*** 0.150*** 0.126*** 0.086*** 0.097***
Log(Physical
stock per
employee)
0.115*** 0.093*** 0.099*** 0.149*** 0.140*** 0.143*** 0.108*** 0.045** 0.070***
Obs. 16,079 2,391 812
N. of firms 1,809 256 170

55

Hence, at this stage, we can conclude that both US manufacturing and US service firms are
more efficient in translating their investments (both in R&D and in physical capital) into
productivity increases. Hence, the transatlantic productivity divide can be explained not only by a
lower level of corporate R&D investment
15
, but also by a lower capacity to translate R&D and
capital investment into productivity gains, and this seems to be obvious both within manufacturing
and within services.

Turning the attention to the within area comparisons, we can notice that the general result of
higher R&D coefficients in services rather than in manufacturing is confirmed both in the US and in
the EU. Instead, embodied technological change appears more prominent in the US services rather
than in the US manufacturing (as it was the case for the whole sample), but in Europe the opposite
occurs. Thus, European manufacturing turns out to be more dependent on external technological
acquisition than European services. This transatlantic difference may be due to the role of the
medium and low-tech sectors in European manufacturing, the sectors where embodied
technological change is particularly crucial (these are the supplier dominated sectors, accordingly to
the Pavitts (1984) taxonomy; see Section 2.3).

Tables 12 and 13 display the results concerning only manufacturing firms split both across
the high-tech sectors vs. other sectors and the two main geographical areas. These results can be
commented on along two dimensions: between areas and within areas. Let us start from the between
areas comparison.

As far as the high-tech sectors are concerned, American firms reveal to be more efficient in
translating both the R&D and the capital expenditures into productivity increases. As usual, all the
coefficients are positive, fully significant and within the expected magnitude ranges; however,
looking at the more sophisticated FE and RE estimates, all the four US coefficients are larger than
the corresponding European ones. Hence, at least in the high-tech manufacturing sectors, US firms
are more able to transfer their own investments into productivity gains.

15
Looking at Table 4, the European underinvestment in comparison with the US is obvious and spread across the
sectors: the whole sample K/E is 59 in the EU vs 93 in the US; 59 vs 89 in the manufacturing sectors; 89 vs 115 in the
high-tech manufacturing sectors; 34 vs 70 in the other manufacturing sectors; 60 vs 114 in the service sectors.

56

Tab. 12: US versus EU: High-tech manufacturing sectors

Whole sample US
High-tech manufacturing
sectors
EU
sectors
Log(R&D
stock per
employee)
0.205*** 0.089*** 0.107*** 0.251*** 0.069*** 0.105*** 0.172*** 0.065*** 0.081***
Log(Physical
stock per
employee)
0.115*** 0.093*** 0.099*** 0.112*** 0.101*** 0.105*** 0.127*** 0.029 0.061***
Obs. 16,079 6,462 1,231
N. of firms 1,809 591 213

Tab. 13: US versus EU: Other manufacturing sectors

Whole sample US
Other manufacturing
sectors
EU
Other manufacturing
sectors
Log(R&D
stock per
employee)
0.205*** 0.089*** 0.107*** 0.175*** 0.060*** 0.079*** 0.112*** 0.035** 0.056***
Log(Physical
stock per
employee)
0.115*** 0.093*** 0.099*** 0.074*** 0.063*** 0.066*** 0.146*** 0.093*** 0.118***
Obs. 16,079 3,752 1,431
N. of firms 1,809 323 256

With regard to the rest of the manufacturing sectors, US firms are still more efficient with
regard to the R&D stock, while embodied technological change seems to play a more relevant role
in the European firms. On the whole, US firms are leading in terms of R&D efficiency regardless of
the sectors, while embodied technological change appears the most effective in the US high-tech
sectors and in the EU non-high-tech manufacturing sectors.

57
Turning the attention to the within area comparisons, the following pictures emerge.

Within the US, high-tech sectors display larger productivity elasticities both with regard to
the R&D and the capital investment (all the six coefficients in the high-tech estimates are larger
than their correspondent figures in the other sectors). Hence, in the US manufacturing high-tech
sectors appear to be characterized by a higher efficiency in translating investments into productivity
advantages.

Differently, European firms in the high-tech sectors show higher coefficients concerning the
productivity elasticity of the R&D stock (all the three coefficients), while the reverse happens as far
as physical capital is concerned (all the three coefficients are higher in the non-high-tech
manufacturing sectors, while the FE estimate in the high-tech sectors is even not significant). This
picture largely confirms what has emerged from a previous study based on DTI European microdata
(Ortega-Argils et al., 2010) where the R&D coefficient was found to increase monotonically
moving from the low-tech to the medium and high-tech sectors, while the capital coefficient was
found to be characterised by an opposite pattern. One conclusion - common to both that article and
this study - is that productivity growth in the European non-high-tech firms is still heavily
dependent on the investment in physical capital (embodied technological change).

4.4 A regional breakdown

As discussed in the previous Section 2.4, one of the purposes of this study is to investigate
possible regional peculiarities (within the EU) in the relationships which are the focus of our
empirical investigation. This can be done at two levels of analysis: on the one hand we can
distinguish macro-areas within Europe to see whether more technologically advanced countries are
characterized by better productivity elasticities of the R&D stock (that is an increasing returns
hypothesis) or vice versa (decreasing returns; see Section 4.4.1). On the other hand, we can split
the observations according to the NUTS1 classification into high-tech regions (independently from
the country they belong to) and low-tech regions and to see what happen to our relevant coefficients
(see Section 4.4.2).

58

4.4.1 Macro-regions

Starting with the former perspective, Table 14 reports the results concerning: 1) the overall
European sample (as a benchmark); 2) the sub-sample of firms located in the Nordic EU countries
(Sweden, Denmark and Finland) and the UK
16
; 3) the sub-sample of firms located in the rest of the
EU; 4) the sub-sample of firms located in the rest of the EU without those firms from the South-
European countries (Italy, Spain and Greece)
17
.

Tab. 14: European macroareas: North (Denmark, Finland, Sweden) + UK, Other EU countries,
Other EU countries without South

EU (overall) North + UK
POLS FE RE POLS FE RE
Log(R&D
stock per
employee)
0.144*** 0.058*** 0.074*** 0.155*** 0.091*** 0.102***
Log(Physical
stock per
employee)
0.125*** 0.053*** 0.078*** 0.122*** 0.064*** 0.081***
Obs. 3,474 1,994
N. of firms 639 347
Other EU countries Other EU countries (Italy,
Greece and Spain excluded)
Log(R&D
stock per
employee)
0.146*** 0.029* 0.050*** 0.142*** 0.025* 0.044***
Log(Physical
stock per
employee)
0.126*** 0.039** 0.069*** 0.130** 0.038** 0.070***
Obs. 1,480 1,413
N. of firms 292 273

16
The literature and the data indicate this portion of the EU as the most technologically advanced. In fact, Table 4
assigns a value of K/E equal to 62 in the North+UK macro area and equal to 56 in the rest of the EU. This gap is even
more marked with regard to manufacturing (63 vs 53) and particularly high-tech manufacturing (103 vs 66).
17
Unfortunately, paucity of observations does not allow to investigate South-Europe as a separate sub-sample.

59

Looking at the R&D stock elasticities, it can be seen that the North+UK displays
significantly higher coefficients than those related to the rest of Europe (as usual, we focus on the
more reliable FE and RE estimates). Moreover, in the FE estimate for the rest of the EU, the
relationship is only barely significant. The worse performance of the rest of Europe in comparison
with the North+UK is not due to the presence of 19 firms coming from the South of Europe: in fact,
in the fourth panel the relevant coefficients even decrease.

Looking at the capital stock elasticities, a similar pattern emerges, although to a lesser
extent. The Nordic countries and the UK exhibit a better ability to translate capital formation in
increasing productivity, although the differences in the magnitude and significance of the
coefficients are much less marked than in the case of the R&D expenditures.

On the whole the Scandinavian countries and the UK reveal a higher efficiency than the rest
of the EU; yet, the coefficients displayed in the second panel of Table 14 are still uniformly smaller
than those concerning the US firms, displayed in Table 7 (second panel). In other words both with
regard to R&D and physical capital the US are leading in terms of efficiency, followed by North
of Europe + UK and the rest of Europe lagging behind.

Turning the attention to the following Tables 15-16-17 and 18, the leadership in terms of
R&D efficiency of the North+UK is fully confirmed when we focus - alternatively - on
manufacturing (Table 15); on services (Table 16); on the high-tech manufacturing sectors (Table
17). In the latter case is interesting to note that the North Europe + UK displays coefficients that are
higher than the American ones, displayed in Table 12; on the other hand, the relevant coefficients in
the rest of Europe turn out to be non-significant). Differently, more ambiguous results emerge as far
as the rest of the manufacturing sectors are concerned (Table 18).

60
Tab.15: European macroareas: Manufacturing sectors
Manufacturing sectos
Log(R&D
stock per
employee)
0.144*** 0.058*** 0.074*** 0.154*** 0.086*** 0.102***
Log(Physical
stock per
employee)
0.125*** 0.053*** 0.078*** 0.125*** 0.063*** 0.082***
Obs. 3,474 1,534
N. of firms 639 251
Other EU countries
Other EU countries (Italy,
Log(R&D
stock per
employee)
0.147*** 0.019 0.040*** 0.146*** 0.017 0.037**
Log(Physical
stock per
employee)
0.145*** 0.052*** 0.086*** 0.150*** 0.055*** 0.092***
Obs. 1,128 1,097
N. of firms 218 208

Tab. 16: European macroareas: Service sectors
Service sectors
Log(R&D
stock per
employee)
0.144*** 0.058*** 0.074*** 0.129*** 0.113*** 0.113***
Log(Physical
stock per
employee)
0.125*** 0.053*** 0.078*** 0.104*** 0.069*** 0.089***
Obs. 3,474 460
N. of firms 639 96
Other EU countries
Service sectors

Service sectors
Log(R&D
stock per
employee)
0.106*** 0.074* 0.081*** 0.098*** 0.077* 0.076*
Log(Physical
stock per
employee)
0.143*** 0.025 0.064* 0.136*** 0.014 0.046
Obs. 352 316
N. of firms 74 65

61
Tab. 17: European macroareas: High-tech manufacturing sectors
Log(R&D
stock per
employee)
0.144*** 0.058*** 0.074*** 0.198*** 0.196*** 0.200***
Log(Physical
stock per
employee)
0.125*** 0.053*** 0.078*** 0.094*** 0.003 0.037
Obs. 3,474 734
N. of firms 639 122
Other EU countries
Log(R&D
stock per
employee)
0.161*** 0.013 0.029 0.161*** 0.013 0.029
Log(Physical
stock per
employee)
0.132*** 0.068** 0.089*** 0.135*** 0.071** 0.096***
Obs. 497 482
N. of firms 91 88

Tab. 18: European macroareas: Other manufacturing sectors
Other manufacturing
Log(R&D
stock per
employee)
0.144*** 0.058*** 0.074*** 0.118*** 0.027*** 0.048***
Log(Physical
stock per
employee)
0.125*** 0.053*** 0.078*** 0.132*** 0.110*** 0.120***
Obs. 3,474 800
N. of firms 639 129
Other EU countries
Other manufacturing
Other manufacturing
Log(R&D
stock per
employee)
0.116*** 0.037 0.062*** 0.116*** 0.029 0.057*
Log(Physical
stock per
employee)
0.179*** 0.059* 0.118*** 0.184*** 0.056* 0.119***
Obs. 631 615
N. of firms 127 120

62

On the whole, the EU seems to be divided in between two different worlds. On the one
hand, there is the Nordic and British world where R&D and productivity are strongly linked, with
exceptionally good results with regard to the high-tech manufacturing sectors (the high-tech Nordic
and British firms are the only ones to exhibit larger R&D/productivity elasticities compared with
their American counterparts). On the other hand, there is the rest of Europe exhibiting quite lower
R&D/productivity coefficients, which turn out to be even not significant in the crucial high-tech
manufacturing sectors. In other words, this means that the European productivity gains related to
the R&D investment are largely driven by what is going to happen in the Nordic countries and in
the UK, with the rest of Europe lagging behind, especially with regard to the role of the high-tech
manufacturing sectors.

4.4.2 R&D-intensive regions vs. non R&D-intensive ones

Turning the attention to the NUTS 1 level, we grouped together the European regions into
two groups: the innovative regions and the weakly innovative ones, according to their R&D/GDP
ratio in 2005, as provided by Eurostat. In order to split the European sample into two comparable
subsamples, we assumed an R&D/GDP ratio equal to 1.8% as a feasible threshold, so generating an
innovative group of 328 firms (1,827 observations) vs. a weakly innovative group of 298
companies (1,604 observations).

Looking back at Table 4, we can see that this splitting based on the R&D/GDP intensity is
fully consistent with what emerges comparing K/E in the innovative group vs. the weakly
innovative one
18
.

In the following Table 19, we report the ranking of the regions, their R&D/GDP ratios, the
number of firms and the number of observations.

18
The respective figures of K/E are 66 vs 49 for the whole sample; 67 vs 48 for manufacturing; 94 vs 79 for the high-
tech manufacturing firms; 70 vs 55 for services.

63
Tab. 19: European NUTS R&D intensities (BERD/GDP) (decreasing order)
NUTS Code R&D/GDP Firms Observations
Baden-Wrttemberg DE1 3.4 16 96
Eastern UKH 3.15 30 188
Sdra Sverige SE2 3.08 25 122
stra Sverige SE1 2.89 35 248
Manner-Suomi FI1 2.48 41 157
Zuid-Nederland NL4 2.39 5 42
Sdsterreich AT2 2.36 3 9
Bayern DE2 2.3 41 175
le de France FR1 2.10 44 236
Hessen DE7 2.09 15 112
Berlin DE3 1.87 9 50
Denmark DK0 1.80 21 152
South East UKJ 1.8 43 240
Centre-Est FR7 1.71 8 29
Sud-Ouest FR6 1.68 1 6
Oststerreich AT1 1.64 7 20
North West UKD 1.59 17 100
Weststerreich AT3 1.52 6 22
Niedersachsen DE9 1.49 7 40
Vlaams Gewest BE2 1.44 11 52
Rgion Wallonne BE3 1.36 3 14
Luxembourg (Grand-Duch) LU0 1.35 3 9
East Midlands UKF 1.32 8 51
South West UKK 1.28 17 111
Rheinland-Pfalz DEB 1.22 5 32
Hamburg DE6 1.15 7 31
Nordrhein-Westfalen DEA 1.1 26 133
Sachsen DED 1.07 1 1
Comunidad de Madrid ES3 1.04 2 5
Thringen DEG 0.95 8 51
Nord Ovest ITC 0.93 3 13
Czech Republic CZ0 0.91 1 4
Bremen DE5 0.91 1 3
Est FR4 0.88 1 8
Norra Sverige SE3 0.85 1 12
Slovenia SI0 0.84 1 4
Ireland IE0 0.82 8 55
West Midlands UKG 0.79 9 38
Oost-Nederland NL2 0.76 5 25
Kzp-Magyarorszg HU1 0.69 2 10
West-Nederland NL3 0.62 15 98
Scotland UKM 0.6 8 58
Rgion de Bruxelles-Capitale BE1 0.54 6 16
Schleswig-Holstein DEF 0.52 2 12

64

NUTS Code R&D/GDP Firms Observations
Wales UKL 0.52 2 10
Northern Ireland UKN 0.49 1 3
Nord Est ITD 0.47 1 3
Estonia EE0 0.42 1 3
Centro ITE 0.41 1 3
Yorkshire and The Humber UKE 0.4 15 88
North East UKC 0.39 5 27
Saarland DEC 0.32 2 12
Attiki GR3 0.29 8 32
Sur ES6 0.27 1 2
London UKI 0.26 59 353
Alfld s szak HU3 0.21 1 2
Voreia Ellada GR1 0.08 1 3

Note: intensities are computed based on 2005 values (due to missing values, for FR1, FR4, FR6,
FR7 2004 values were used; for DK0 2007) (Eurostat)

In Table 20, the benchmark European figures
19
are compared with the estimates coming out
from the separate estimates for the group of firms located in innovative regions vs. their
counterparts located in the less innovative ones. As can be seen - as was the case of the high-tech
sectors more is better: those regions that invest more in R&D are also characterised by a better
ability to translate the R&D investment in an increase in productivity. In more detail, all the three
R&D coefficients (uniformly significant) are larger in magnitude when estimated within the group
of the innovative regions. In other words, innovative European regions not only invest more in
R&D, but also achieve more in terms of productivity gains from their own research activities.

If we jointly take into account this result with both the one concerning the high-tech sectors
and that one concerning the European macro-regions, we can conclude that - at least in Europe the
R&D investment is strongly characterised by increasing returns in terms of its productivity
impact. In fact, this impact is higher in: 1) the high-tech manufacturing sectors; 2) in the Nordic
countries and in the UK; 3) in the most innovative European regions; that is: where more is invested
in R&D, more is achieved in terms of productivity gains.

19
For 13 firms (43 observations) it was impossible to assign a NUTS 1 code; this is why the first panel of Table 20
presents some minor differences in comparison with the third panel of Table 7.

65

(Regional BERD/GDP >= 1.8% is the threshold)

EU (overall) Innovative NUTS Weakly innovative NUTS
Log(R&D
stock per
employee)
0.144*** 0.057*** 0.073*** 0.160*** 0.072*** 0.087*** 0.119*** 0.044*** 0.057***
Log(Physical
stock per
employee)
0.122*** 0.053*** 0.079*** 0.091*** -0.01 0.031** 0.145*** 0.093*** 0.111***
Obs. 3,431 1,827 1,604
N. of firms 626 328 298

Note: for 13 firms (43 observations), NUTS classification is not available

As far as the capital stock is concerned, the weakly innovative European regions seem to be
characterised by a dominant role of the embodied technological change, which does not turn out to
be crucial in the R&D-intensive regions. If we consider the latter results together with the evidence
coming out from Tables 12 and 13, we come out with a picture where advanced European regions
characterised by high-tech sectors rely on R&D expenditure as the main lever to increase
productivity, while lagging regions specialised in the non high-tech sectors rely more on the
embodied technological change incorporated in capital formation.

Both these conclusions are reinforced from what emerges from the following Tables 21, 22,
23 and 24 where we replicated the overall estimation reported in the previous Table 20, isolating the
high-tech sectors, the rest of the sectors and manufacturing vs. services. As can be seen focusing
on the more reliable FE and RE estimated coefficients the highest R&D/productivity elasticities
are displayed by the firms belonging to the high-tech sectors and located in the innovative regions
(0.097 and 0.095). Interestingly enough, the innovative regions are also characterised by very high
R&D/productivity elasticities (0.096 and 0.118) in the service sectors. In all sectors (with a partial
exception with regard to the non-high-tech manufacturing sectors) the innovative regions are
characterised by larger R&D coefficients in comparison with the other regions. This is a further
confirmation of the increasing return hypothesis introduced above.

66
Turning the attention to capital formation and embodied technological change, an
unambiguous outcome clearly merges: in all the economic sectors, the weakly innovative European
regions strongly rely on embodied technological change with a capital/productivity elasticity that is
always larger than the one estimated within the firms located in the R&D-intensive regions. Not
surprisingly, the highest capital coefficients come out when the non-high-tech manufacturing
sectors are investigated (0.140 from the FE estimate; 0.176 from the RE estimate). As discussed in
Section 2.3, in the medium-tech and in the traditional low-tech sectors which are focusing on
process innovation productivity gains are much more related to capital accumulation rather than to
R&D expenditures.

Tab. 21: European NUTS: Innovative NUTS versus Weakly innovative NUTS in High-tech
manufacturing sectors

EU (overall) Innovative NUTS
sectors
Weakly innovative NUTS
sectors
Log(R&D
stock per
employee)
0.144*** 0.057*** 0.073*** 0.109*** 0.097*** 0.095*** 0.188*** 0.035 0.054***
Log(Physical
stock per
employee)
0.122*** 0.053*** 0.079*** 0.143*** 0.004 0.067** 0.135*** 0.050** 0.072***
Obs. 3,431 688 529
N. of firms 626 114 96

Table 22: European NUTS: Innovative NUTS versus Weakly innovative NUTS in Other
manufacturing sectors

Other manufacturing
sectors
Other manufacturing
sectors
Log(R&D
stock per
employee)
0.144*** 0.057*** 0.073*** 0.141*** 0.024 0.059*** 0.065*** 0.040* 0.047**
Log(Physical
stock per
employee)
0.122*** 0.053*** 0.079*** 0.085*** 0.006 0.027 0.203*** 0.140*** 0.176***
Obs. 3,431 670 749
N. of firms 626 124 129

67

Hence, we can further confirm and specify what has been already discussed commenting on
the sectoral results reported in the Tables 12 and 13. In the EU, the investment in physical capital is
significantly linked to productivity gains, confirming the hypothesis advanced in this study that
embodied technological change is a crucial driver of productivity evolution. While this
contribution is similar to the one offered by the R&D expenditures in aggregate (see the first panel
in Table 20, columns 2 and 3), when we only consider either the manufacturing non-high-tech
sectors (Table 13, panel 3, columns 2 and 3) or the non-R&D-intensive European regions (Tables
20, 21, 22, 23 and 24; panel 3, columns 2 and 3) the capital coefficient systematically exceeds the
correspondent R&D coefficient.

Tab. 23: European NUTS: Innovative NUTS versus Weakly innovative NUTS in Manufacturing
sectors

Manufacturing
sectors
Manufacturing
sectors
Log(R&D
stock per
employee)
0.144*** 0.057*** 0.073*** 0.129*** 0.068*** 0.084*** 0.128*** 0.038** 0.053***
Log(Physical
stock per
employee)
0.122*** 0.053*** 0.079*** 0.115*** 0.002 0.044** 0.167*** 0.098*** 0.120***
Observations 3,431 1,358 1,278
N. of firms 626 238 225

Tab. 24: European NUTS: Innovative NUTS versus Weakly innovative NUTS in Service sectors

Service
sectors
Service
sectors
Log(R&D
stock per
employee)
0.144*** 0.057*** 0.073*** 0.207*** 0.096*** 0.118*** 0.059** 0.068 0.056*
Log(Physical
stock per
employee)
0.122*** 0.053*** 0.079*** 0.056** -0.007 0.008 0.088*** 0.089** 0.098***
Observations 3,431 469 326
N. of firms 626 90 73

68

On the whole, in Europe, productivity growth in medium and low-tech sectors and in the
less innovative regions is still heavily dependent on investment in physical capital (embodied
technological change), while corporate R&D seems to play a secondary role.

4.5 The business cycle

As discussed in Section 2.4, previous empirical literature has supported the hypothesis that
the business cycle upswings are favourable to the introduction and diffusion of innovation;
however, there are no previous studies to our knowledge - that have investigated whether
upswings are also conducive of a more effective impact of R&D on productivity; this will be the
hypothesis that will be tested in this section.

First of all, we have to identify upswings and downswings both in the US and in the EU.
This has been done on the basis of the evolution of the real GDP growth in the two economic areas,
as reported in Figure 1 in Section 2.1. In particular, we singled out three downswing turning points
(1991, 1995, 2001) and three upswing turning points (1994, 1999, 2004) in the US. On the other
hand, we identified three downswing turning points (1993, 1996, 2002) and three upswing turning
points (1994, 2000, 2006) in the EU
20
. The years of expansion and recession were allocated
accordingly.

Table 25 reports the outcomes of the two separate regressions run for the sample of the
expansionary years and for the sample of the recessionary years (with US and EU pooled together
and benchmarked with the overall results for the whole sample). As can be seen, both the R&D and
the capital elasticities
21
appear to be larger in the recessionary period rather than in the
expansionary ones (although the differences are not so striking). Hence, there is weak evidence that
R&D expenditures and capital formation are more crucial in sustaining productivity during the bad
times rather than during the good ones. A possible interpretation of this result might take into

20
Given the national composition of our final sample, as reported in Table 1, we assumed the evolution of the EU15
real GDP as an adequate benchmark.
21
As usual, we focus our attention upon the more reliable FE and RE estimates.

69
account the role of demand in sustaining productivity growth through the Kaldor-Verdoorn law:
while in the upswing periods increasing demand is fostering productivity growth through
expansionary investments and scale economies, in recessions this virtuous circle is broken and
tangible and intangible investments may turn out to be the key factors affecting productivity
growth.

Tab. 25: Whole sample, Recessions and Expansions

Whole sample Recessions Expansions
Log(R&D
stock per
employee)
0.205*** 0.089*** 0.107*** 0.213*** 0.089*** 0.130*** 0.196*** 0.088*** 0.116***
Log(Physical
stock per
employee)
0.115*** 0.093*** 0.099*** 0.114*** 0.110*** 0.113*** 0.116*** 0.085*** 0.096***
Obs. 16,079 8,073 8,006
N. of firms 1,809 1,728 1,590

Tables 26 and 27 compare overall, American and European coefficients estimated in the
recessionary periods (Table 26) and in the expansionary ones (Table 27).

Tab. 26: US versus EU: Recessions
(US : 1990,1991,1995,2000,2001,2005,2006,2007,2008;
EU : 1991,1992,1993,1995,1996,2001,2002,2007,2008)
Recessions US EU
Log(R&D
stock per
employee)
0.213*** 0.089*** 0.130*** 0.230*** 0.088*** 0.135*** 0.156** 0.082*** 0.116***
Log(Physical
stock per
employee)
0.114*** 0.110*** 0.113*** 0.103*** 0.115*** 0.111*** 0.137*** 0.064*** 0.114***
Obs. 8,073 6,522 1,551
N. of firms 1,728 1,151 577

70

A first consideration is that the above result of higher elasticities in the recessionary period
appears to be fully confirmed only in the European case: in fact, the US R&D coefficients coming
out from the FE and RE estimations turn out to be higher in the expansionary periods. In the US,
expansionary periods seem to offer the opportunity to strength the R&D/productivity relationship.
However, the differences in the estimated coefficients do not allow pushing this argument so far.

Turning the attention to the comparison between the US and the EU, Tables 26 and 27
inform us that the transatlantic gap that was detected in aggregate on the whole sample (see Table
7) is fully confirmed both during the recessionary periods and when upswing years are considered.

Tab. 27: US versus EU: Expansions
(US : 1992,1993,1994,1996,1997,1998,1999,2002,2003,2004;
EU : 1994,1997,1998,1999,2000,2003,2004,2005,2006)
Expansions US EU
Log(R&D
stock per
employee)
0.196*** 0.088*** 0.116*** 0.225*** 0.104*** 0.138*** 0.137*** 0.050*** 0.077***
Log(Physical
stock per
employee)
0.116*** 0.085*** 0.096*** 0.109*** 0.094*** 0.099*** 0.113*** 0.044*** 0.070***
Observations 8,006 6,083 1,923
N. of firms 1,590 1,026 564

As far as the R&D coefficients are concerned, the US estimated figures are always higher
than their EU correspondents (six out of six); hence, the better ability of the US business sector in
translating R&D investments in productivity gains is obvious independently of the business cycle.

With regard to the capital stock coefficients, the outcome are slightly more ambiguous;
however if we restrict our attention to the FE and RE coefficients in three cases out of four the

71
US coefficient is significantly larger than the European one, while in the remaining opposite case
the European coefficient is only marginally higher than its US correspondent.

On the whole, the analysis of the business cycle reveals that the overall differences in the
elasticities along the opposite phases of the cycle are not so significant, with a weak evidence of a
larger productivity impact of both R&D and capital during the recessionary periods. However - with
both R&D investment and capital formation is fully confirmed and seems to be independent of the
business cycle.

72
5. Conclusions and policy implications

According to the discussion and the hypotheses introduced in the first section of this report,
the main focus of the present research project was the understanding of the link between R&D and
firms performance, measured in terms of labour productivity.

From a policy point of view, the idea was that the relationship between R&D and
productivity was at the core of the matter concerning the transatlantic productivity gap. Indeed,
R&D expenditures have been demonstrated to play an important role in explaining the productivity
differentials within the industrialised countries. In particular, the role of private R&D investment by
corporate firms has been recognised as a fundamental engine for productivity growth both at the
macro and microeconomic level. As shown in Section 2.1, the EU15 lags considerably and
persistently behind the US in this respect, even more strikingly than in terms of total R&D.

However, the basic hypothesis introduced in this study was that the transatlantic gap might
be due not only to a lower level of public and corporate R&D investment in the European
economies, but also to a possible lower capacity to translate corporate R&D expenditures into
productivity gains exhibited by the European firms and sectors. Indeed, it might be well the case
that European economies not only invest less in R&D, but also get less from their R&D investment
because of a lower R&D-productivity elasticity in the EU compared with the US.

Yet, R&D is not the sole determinant of productivity gains: while the R&D input is
capturing that portion of technological change which is related to the disembodied new knowledge,
gross investment is an alternative innovative input capturing the new knowledge embodied in
physical capital, mainly machinery introduced through additional investments or simply through
scrapping. This second input represents the so-called embodied technological change, with his great
potential to positively affect productivity growth.

A second hypothesis of this work was that R&D might be crucial in fostering productivity in
the high-tech sectors, while being less important in the low-tech sectors where alternative sources of
productivity growth such as embodied technological change might play a dominant role.

73
In terms of policy perspectives, specific European industrial and innovation policies should
be designed for different microeconomic environments and for different industrial sectors, in order
to maximise the impact of the different sources of innovation onto productivity.

Indeed, while previous economic literature has found robust evidence of a positive and
significant impact of R&D on productivity at the firm level, sectoral empirical studies using
different datasets across different countries have found a greater impact of R&D investments on
firm productivity in the high-tech sectors rather than in the low-tech ones (see Section 2.2).

The third hypothesis tested in this report was that the relationship between R&D and
productivity might exhibit important differences across different EU countries and regions and
might vary along the business cycle. In terms of policy implications, this would mean that policies
should be not only sectorally targeted (see above), but also differentiated across regions and over
time (see Section 2.3).

For instance, with regard to regional policy, the particular nature of the relationship between
R&D and capital formation on the one hand and productivity evolution on the other hand might
heavily be affected by the industrial structure which characterises a single region. Thus - according
to what discussed above - a region characterised by a large presence of high-tech sectors would
probably turn out to be very sensitive to R&D activities in getting productivity gains, while a region
characterised by a disproportionate presence of traditional sectors and SMEs would come out to be
particularly responsive to capital formation.

The three hypotheses discussed so far were tested using firm level data, provided by the
JRCIPTS of the European Commission and extracted from a variety of sources, including
companies annual reports. Once acquired the rough data from IPTS, we proceeded in the
construction of a longitudinal database in order to run panel estimations addressed to test the
theoretical hypotheses discussed above (see Section 3.1). At the end of this process, we ended up
with a final dataset comprising 1,809 companies (1,170 US and 639 EU, over the period 1989-2008,
for a total of 16,079 observations).

Econometrically, we tested an augmented production function in three inputs: physical
capital (capital stock), labour and knowledge capital (R&D stock; for details see Section 3.3).
Firstly, pooled ordinary least squared regressions were run to provide preliminary reference

74
evidence. Secondly, fixed effect (FE) regressions were performed in order to take into account firm
specific unobservable characteristics such as managerial capabilities. Thirdly, random effect (RE)
regressions were run to provide the more comprehensive and reliable results.

The empirical results were reported and discussed in details in Section 4, here we will just
summarize them, using the same list of research question-marks that were proposed in the
introductory Section 1. In particular, the following research issues were investigated.

1) To see whether significant differences emerge in the link between R&D and productivity
between the US and the EU, in order to shed some light on the interpretation of the transatlantic
productivity gap (Section 4.1).
Consistently with the previous literature, we found robust evidence of a positive and
significant impact of R&D on productivity with an elasticity ranging from 0.089 to 0.205, according
to the different adopted estimation techniques.
However, although uniformly positive and statistically significant, the R&D coefficients for
the US firms turn out to be consistently larger than the corresponding coefficients for the European
firms. Indeed, the three estimation techniques consistently provide European elasticities equal to
about 60% of their US counterparts. We interpreted these unambiguous results as a clear evidence
of the better ability of US firms in translating R&D investments into productivity gains and as a
signal of a gap of efficiency that European firms and European policy have to deal with.

2) To see whether further support can be found to the hypothesis that R&D should be clearly
and significantly linked to productivity in the high-tech sectors, while a weaker impact should
emerge in the other sectors of the economy (Section 4.2).
This hypothesis was also confirmed by the microeconometric estimates: the R&D
coefficients related to the high-tech manufacturing sectors always turned out to be larger than the
corresponding ones for the other manufacturing sectors. This outcome is consistent with the
previous literature: on the whole, high-tech sectors not only invest more in R&D and are
characterised by a higher productivity performance, but also achieve more in terms of productivity
gains from their own R&D activities.

75
3) To see whether physical capital emerges as an important second driver of productivity
gains, so confirming the hypothesis that embodied technological change is a crucial determinant
of productivity evolution. This relationship is expected to be particularly strong in the low-tech
sectors, where embodied technological change might be expected the main source of productivity
gains (Section 4.2).
Overall, we found consistently with previous studies - a positive and significant impact of
physical capital over labour productivity, ranging from 0.093 to 0.115.
Interestingly enough, the US revealed an advantage similar to the one emerged for the
intangible R&D investments; thus, US firms resulted more efficient in getting productivity gains
both from the R&D and the physical capital investments.
In contrast with our expectations, embodied technological change appeared to be more
effective in the high-tech sectors rather than in the rest of the manufacturing sectors. In fact,
according to all the three econometric methodologies, the coefficients turned out to be higher in the
high-tech sectors.

4) To see to what extent the transatlantic differences may be related to the different sectoral
structures and to the peculiar sectoral R&D/productivity relationships detectable in the US and in
the EU (Section 4.3).
We differentiated the US/EU comparative empirical exercise by manufacturing vs. service
sectors: it came out that both US manufacturing and US service firms were more efficient in
translating their investments (both in R&D and in physical capital) into productivity increases. In
addition, the US efficiency advantage in R&D activities is obvious both in the high-tech
manufacturing sectors and in the rest of the manufacturing sectors. On the whole, US firms are
leading in terms of R&D efficiency regardless of the sectors. Hence, the transatlantic productivity
divide can be explained not only by a lower level of corporate R&D investment, but also by a lower
capacity to translate R&D into productivity gains, and this seems to be obvious both within
manufacturing (both high- and medium/low-tech sectors) and within services.
Looking inside the American and the European aggregates, within the US, high-tech sectors
displayed larger productivity elasticities both with regard to the R&D and the capital. Hence, in the
US manufacturing, high-tech sectors appear to be characterized by a higher efficiency in translating
investments into productivity advantages.

76
Differently, European firms in the high-tech sectors turned out to be characterised by larger
coefficients concerning the productivity elasticity of the R&D stock, while the reverse happened as
far as physical capital is concerned. Hence, productivity growth in the European non-high-tech
firms is still heavily dependent on the investment in physical capital (embodied technological
change).

5) To see whether (and how much) the intensity of the R&D/productivity link is affected by
the sectoral composition and the institutional context characterising the different European
countries and regions (Section 4.4).
On the whole, the EU economy appeared to be divided into two different macro-areas. On
the one hand, there is the Nordic and British world where R&D and productivity are strongly
linked, with exceptionally good results with regard to the high-tech manufacturing sectors. On the
other hand, there is the rest of Europe exhibiting quite lower R&D/productivity coefficients, which
turn out to be even not significant in the crucial high-tech manufacturing sectors. This is particularly
important in terms of European economic and innovation policy, since the European productivity
gains related to R&D activities seem to be largely driven by what is going to happen in the Nordic
countries and in the UK, with the rest of Europe lagging behind, especially with regard to the role of
the high-tech manufacturing sectors.
Turning the attention to the regional level, we grouped together the European regions into
two groups: the innovative regions and the weakly innovative ones, according to their R&D/GDP
ratio. As far as empirical results are concerned, a general conclusion was that those regions that
invest more in R&D are also characterised by a better ability to translate the R&D investment into
an increase in productivity. In particular, the highest R&D/productivity elasticities were displayed
by the firms belonging to the high-tech sectors and located in the most innovative regions.
Symmetrically, productivity growth in medium and low-tech sectors and in the less innovative
regions was found still heavily dependent on investment in physical capital (embodied
technological change), with corporate R&D playing a secondary role.
On the whole, we can conclude that - at least in Europe the R&D investment is strongly
characterised by increasing returns in terms of its productivity impact. In fact, this impact is
higher in: 1) the high-tech manufacturing sectors; 2) in the Nordic countries and in the UK; 3) in the
most innovative European regions; that is: where more is invested in R&D, more is achieved in
terms of productivity gains. As it will be discussed below, this outcome implies important policy
implications.

77

6) To see whether the coefficients linking R&D and productivity are stable over time or turn
out to be affected by the business cycle (Section 4.5).
On the whole, the analysis of the business cycle revealed that the overall differences in the
elasticities along the opposite phases of the cycle were not so significant, with a weak evidence of a
larger productivity impact of both R&D and capital during the recessionary periods. However - with
both R&D investment and capital formation was fully confirmed and found to be independent of the
business cycle.

Although necessarily tentative, some policy implications can be derived from the empirical
results obtained in this study. These suggestions concern to a different extent research,
innovation, industrial and regional EU policies.
Firstly, the obtained results show that the US economy is uniformly more efficient in getting
productivity advantages from investments in R&D activities; while this is obvious for the whole
economy, the efficiency gap is confirmed separately in services and manufacturing and within
manufacturing both in the high-tech sectors and in the other industrial sectors. Hence, the
transatlantic divide is not only a matter either of a lower R&D investment in Europe or of an
European industrial structure specialised in middle and low-tech sectors. With the only exception of
UK and the Nordic Countries, European firms are structurally less able to translate R&D
expenditures into productivity gains. This can be due to a lower level of human capital or to a lag in
those organizational changes that are indispensable complements of technological change. While
these perspectives are beyond the scope of the present report, this conclusion has a first important
policy implication: just increasing R&D is a necessary but not sufficient policy if the overall
increase in productivity is the target.
Secondly, this study clearly shows that higher productivity gains from R&D investments can
be achieved in the high-tech manufacturing sectors. Here a second policy implication emerges: the
allocation of R&D efforts is as important as an increase in R&D and high-tech sectors should be
targeted by national and European R&D policies. Indeed, the results coming out from this report
offer a second reason to favour European high-tech sectors: in fact, they not only invest more in
R&D, but in these sectors corporate R&D efforts are more effective in achieving productivity gains.

78
In other words, the findings of this research support a targeted research policy rather than an erga
omnes type of public intervention.
Thirdly, this study shows that R&D investment is not the sole source of productivity gains;
technological change embodied in capital formation is of comparable importance. Also with regard
to the relationship between physical capital and productivity, the US economy exhibits an
advantage, similar to the one detected for the R&D activities. Here again, the suspect is that
European firms lack of those complementary factors such as adequate human resources and
updated organizational layouts which fuel the productivity increases resulting from tangible and
intangible investments. Finally, embodied technological change appears to be crucial within
European non-high-tech firms; hence, an European innovation policy aiming to increase
productivity in the medium/low-tech sectors should support overall capital formation.
Fourthly, the European aggregate seems to be divided into two different worlds: on the one
hand, the UK and the Nordic countries which exhibit an R&D/productivity pattern similar to the US
one, and on the other hand the rest of the Continent lagging behind. This divide is also obvious
looking at the regional level, where the most R&D-based regions also show the better results in
terms of R&D/productivity elasticities. Overall, Europe is lagging behind the US and within
Europe the R&D/productivity link is clearly characterised by increasing returns both in terms of
countries, regions and sectors: where more is invested in R&D, more is achieved in terms of
productivity gains. Accordingly, an innovation policy addressed to fill the transatlantic productivity
gap should be targeted to the most R&D-based actors across the different European regions and
sectors.

79

References

80
Acs, Z.J. and D.B. Audretsch. 1988. Innovation in large and small firms: an empirical analysis,
American Economic Review, 78, 678-690.
Acs, Z.J., and D.B. Audretsch. 1990. Innovation and Small Firms, Cambridge: MIT Press.
Acs, Z.J., Audretsch, D.B. and M.P. Feldman. 1994. R&D spillovers and recipient firm size, Review
of Economics and Statistics, 76, 336-340.
Aghion, P. and P. Howitt. 1996. The observational implications of Schumpeterian growth theory,
Empirical Economics, 21, 13-25.
Baumol, W.J. 2002. The Free Market Innovation Machine: Analyzing the Growth Miracle of
Capitalism, Princeton: Princeton University Press.
Blanchard, O. 2004. The economic future of Europe, Journal of Economic Perspectives, 8, 3-26.
Bosworth, B.P. and J.E. Triplett. 2007. The early 21st century productivity expansion is still in
services, International Productivity Monitor, 14, 3-19.
Breschi, S., Malerba, F. and L. Orsenigo. 2000. Technological regimes and Schumpeterian patterns
of innovation, Economic Journal, 110, 388-410.
Bronzini R. and P. Piselli. 2009. Determinants of long-run regional productivity with geographical
spillovers: The role of R&D, human capital and public infrastructure, Regional Science and
Urban Economics, 39, 187-199.
Brouwer, E. and A. Kleinknecht. 1996. Firm size, small business presence and sales in innovative
products: a micro-econometric analysis, Small Business Economics, 8, 189-201.
Conte, A. and M. Vivarelli. 2005. One or many knowledge production functions? Mapping
innovative activity using microdata, IZA Discussion Paper, No. 1878.

81
Crpon, B., E. Duguet and J. Mairesse. 1998. Research, innovation, and productivity: an
econometric analysis at firm level, Economics of Innovation and New Technology, 7, 115-158.
Crespi, F. and M. Pianta. 2008. Diversity in innovation and productivity in Europe, Journal of
Evolutionary Economics, 18, 529-545.
Cuneo, P. and J. Mairesse. 1983. Productivity and R&D at the firm level in French manufacturing,
NBER Working Paper, 1068.
Daveri, F. 2002. The new economy in Europe, 1992-2001, Oxford Review of Economic Policy, 18,
345-362.
Dettori, B., Marrocu, E. and R. Paci. 2008. Total factor productivity, intangible assets and spatial
dependence in the European regions, Working Paper CRENOS, 2008/23.
Dosi, G. 1988. Sources, procedures, and microeconomic effects of innovation, Journal of Economic
Literature, 26, 1120-1171.
Dosi, G. 1997. Opportunities, incentives and the collective patterns of technological change,
Economic Journal, 107, 1530-1547.
European Council. 2002. Presidency conclusions. Barcelona European Council. 15 and 16 March
2002, Brussels.
European Commission. 2002. More research for Europe. Towards 3% of GDP, COM(2002) 499
final, Brussels.
European Commission. 2008. Analysis of the 2007 EU Industrial R&D Investment Scoreboard,
Luxembourg.
Fischer M., T. Scherngell and M. Reismann. 2008. Knowledge spillovers and total factor
productivity, mimeo, ERSA Conference, Liverpool.

82
Freeman, C. 1982. The Economics of Industrial Innovation, London: Pinter.
Freeman, C. and L. Soete. 1987. Technical Change and Full Employment, Oxford: Basil
Blackwell.
Freeman, C., Clark, J. and L. Soete. 1982. Unemployment and Technical Innovation, London:
Pinter.
Geroski, P.A. and C.F. Walters. 1995. Innovative activity over the Business Cycle, Economic
Journal, 105, 916-928.
Greenhalgh, C., Longland, M. and D. Bosworth. 2001. Technology activity and employment in a
panel of UK firms, Scottish Journal of Political Economy, 48, 260-282.
Greenwood, J., Hercowitz, Z. and P. Krusell. 1997. Long-run implications of investment-specific
technological change, American Economic Review, 87, 342-362.
Griliches, Z. 1979. Issues in assessing the contribution of Research and Development to
productivity growth, Bell Journal of Economics, 10, 92-116.
Griliches, Z. 1995. R&D and productivity: econometric results and measurement issues, in P.
Stoneman (Ed.), Handbook of the Economics of Innovation and Technological Change,
Oxford: Blackwell Publishers, 52-89.
Griliches, Z. 2000. R&D, Education, and Productivity, Cambridge, MA: Harvard University Press.
Griliches, Z. and J. Mairesse. 1982. Comparing productivity growth: An exploration of French and
US industrial and firm data, NBER Working Paper 961.
Grubbs, F. 1969. Procedures for detecting outlying observations in samples, Technometrics, 11, 1-
21.

83
Gumbau-Albert, M and J. Maudos. 2006. Technological activity and productivity in the Spanish
regions, Annals of Regional Science, 40, 55-80.
Hall, B.H. and J. Mairesse. 1995. Exploring the relationship between R&D and productivity in
French manufacturing firms, Journal of Econometrics, 65, 263-293.
Hall, B.H. 2007. Measuring the returns to R&D: the depreciation problem, NBER Working Paper
13473.
Harhoff, D. 1998. R&D and productivity in German manufacturing firms, Economics of Innovation
and New Technology, 6, 29-49.
Hatzichronoglou, T. 1997. Revision of the High-technology Sector and Product Classification,
Paris: OECD.
Hercowitz, Z. 1998. The embodiment controversy: A review essay, Journal of Monetary
Economics, 41, 217-224.
Hulten, C.R. 1990. The measurement of capital, in E.R. Berndt and J.E. Triplett (eds.), Fifty Years
of Economic Management, Chicago: University of Chicago Press, 119-152.
Hulten, C.R. 1992. Growth accounting when technical change is embodied in capital, American
Economic Review, 82, 964-980.
Janz, N., H. Lf and B. Peters. 2004. Firm level innovation and productivity Is there a common
story across countries?, Problems and Perspectives in Management, 2, 1-22.
Jones, C.I. 2002. Sources of U.S. economic growth in a world of ideas, American Economic
Review, 92, 220-239.
Jorgenson, D.W. 1966. The embodiment hypothesis, Journal of Political Economy, 74, 1-17.

84
Jorgenson, D.W. 1990. Productivity and economic growth, in E.R. Berndt and J.E. Triplett (eds.),
Fifty Years of Economic Growth, Chicago, Chicago University Press, 19-118.
Jorgenson, D.W. and K.J. Stiroh. 2000. Raising the speed limit: US economic growth in the
information age, Brookings Papers on Economic Activity, 1, 125-235.
Jorgenson, D.W., Ho, M.S. and K.J. Stiroh. 2005. Information Technology and the American
Growth Resurgence, Cambridge: MIT Press.
Jorgenson, D.W., Ho, M.S. and K.J. Stiroh. 2008. A retrospective look at the U.S. productivity
growth resurgence, Journal of Economic Perspectives, 22, 3-24.
Kleinknecht, A. 1987. Innovation Patterns in Crisis and Prosperity, London: Macmillan.
Klette, J. and S. Kortum. 2004. Innovating firms and aggregate innovation, Journal of Political
Economy, 112, 986-1018.
Kwon, H.U. and T. Inui. 2003. R&D and productivity growth in Japanese manufacturing firms,
ESRI Discussion Paper Series No.44, Tokyo.
Lf, H. and A. Heshmati. 2006. On the relation between innovation and performance: A sensitivity
analysis, Economics of Innovation and New Technology, 15, 317-344.
Mairesse, J. and P. Mohnen. 2001. To be or not to be innovative: an exercise in measurement,
NBER Working Paper, 8644.
Mairesse, J. and P. Mohnen. 2005. The importance of R&D for innovation: A reassessment using
French survey data, Journal of Technology Transfer, 30, 183-197.
Mairesse, J. and M. Sassenou. 1991. R&D and productivity: a survey of econometric studies at the
firm level, NBER Working Paper, 3666.
Malerba, F. 2002. Sectoral systems of innovation and production, Research Policy, 31, 247-264.

85
Malerba, F. (Ed.). 2004. Sectoral Systems of Innovation, Cambridge: Cambridge University Press.
Malerba, F., and L. Orsenigo. 1996. Schumpeterian patterns of innovation, Cambridge Journal of
Economics, 19, 47-65.
Marsili, O. 2001. The Anatomy and Evolution of Industries, Northampton, MA: Edward Elgar.
McMorrow K., Rger, W. and A.A. Turrini. 2009. The EU-US total factor productivity gap: An
industry-level perspective, CEPR Discussion Paper, No.7237.
Mensch, G. I979. Stalemate in Technology: Innovations Overcome the Recession, New York:
Bollinger.
Nadiri, M.I. and I. Prucha. 1996. Estimation of the depreciation rate of physical and R&D capital in
the US total manufacturing sector, Economic Inquiry, 34, 43-56.
Nelson, R.R. and S.G. Winter. 1982. An Evolutionary Theory of Economic Change, Harvard:
Harvard University Press.
OECD. 2002. Frascati Manual - Proposed Standard Practice for Surveys on Research and
Experimental Development, Paris: OECD.
OECD. 2009. Main Science and Technology Indicators. Paris: OECD (2009 edition).
Oliner, S. and D. Sichel. 2000. The resurgence of growth in the late 1990s: Is information
technology the story?, Journal of Economic Perspectives, 14, 3-22.
OMahony, M. and B. Van Ark (eds.) 2003. EU Productivity and Competitiveness: An Industry
Perspective. Can Europe Resume the Catching-up Process?, Luxembourg: European
Commission.
Ortega-Argils R., Piva, M., Potters, L. and M. Vivarelli. 2010. Is corporate R&D investment in
high-tech sectors more effective?, Contemporary Economic Policy, forthcoming.

86
Parisi, M., F. Schiantarelli and A. Sembenelli. 2006. Productivity, innovation creation and
absorption, and R&D. Microevidence for Italy, European Economic Review, 50, 2037-2061.
Pavitt, K. 1984. Sectoral patterns of technical change: Towards a taxonomy and a theory, Research
Policy, 13, 343-373.
Piva, M. and M. Vivarelli (2007) Is demand-pulled innovation equally important in different groups
of firms?, Cambridge Journal of Economics, 31, 691-710.
Rincon, A. and M. Vecchi. 2003. Productivity performance at the company level, in M. OMahony
and B. Van Ark (eds.), EU Productivity and Competitiveness: An Industry Perspective. Can
Europe Resume the Catching-up Process?, Luxembourg: European Commission, 169-208.
Rogers, M. 2006. R&D and productivity in the UK: evidence from firm-level data in the 1990s,
Economics Series Working Papers, 255, University of Oxford.
Salter, W.E.G. 1960. Productivity and Technical Change, Cambridge: Cambridge University Press.
Santamara, L., Nieto, M.J., and A. Barge-Gil. 2009. Beyond formal R&D: taking advantage of
other sources of innovation in low-and medium-technology industries, Research Policy, 38,
507-517.
Santarelli, E. and A. Sterlacchini. 1990. Innovation, formal vs. informal R&D, and firm size: some
evidence from Italian manufacturing firms, Small Business Economics, 2, 223-228.
Santarelli, E. and A. Sterlacchini. 1994. Embodied technological change in supplier dominated
firms, Empirica, 21, 313-27.
Scherer, F.M. 1982. Demand-pull and technological invention: Schmookler revisited, Journal of
Industrial Economics, 30, 225-237.
Schmookler, J. 1962. Economic sources of inventive activity, Journal of Economic History, 22, 1-
10.

87
Schmookler, J. 1966. Invention and Economic Growth, Cambridge: Harvard University Press.
Schumpeter, J.A. 1934. The Theory of Economic Development, Cambridge: Harvard University
Press.
Schumpeter, J.A. 1942. Capitalism, Socialism and Democracy, New York: Harper and Brothers.
Solow, R.M. 1960, Investment and technical progress, in K.J. Arrow, S. Karlin and P. Suppes
(eds.), Mathematical Methods in the Social Sciences, Stanford: Stanford University Press, 89-
104.
Solow, R.M. 1987. Wed Better Watch Out, New York Times, book review, July 12.
Stefansky, W. 1972. Rejecting outliers in factorial designs, Technometrics, 14, 469-479.
Stiroh, K.J. 2002. Information technology and the US productivity revival: what do the industry
data say?, American Economic Review, 92, 1559-1576.
Timmer, M.P., Ypma, G. and B. Van Ark. 2003. IT in the European Union: Driving productivity
divergence?, Groningen Growth and Development Centre Research Memorandum GD-67,
Groningen: University of Groningen. http://www.ggdc.net/databases
Tsai, K.H. and J.C. Wang. 2004. R&D productivity and the spillover effects of high-tech industry
on the traditional manufacturing sector: The case of Taiwan, World Economy, 27, 1555-1570.
Triplett, J.E. and B.P. Bosworth. 2004. Productivity in the U.S. Service Sector: New Sources of
Economic Growth, Washington, DC: Brookings Institution Press.
Turner, L. and H. Boulhol. 2008. Recent trends and structural breaks in US and EU15 Labour
Productivity Growth, OECD Economics Department Working Papers, n. 628.
Van Ark, B., OMahony M. and M.P. Timmer. 2008. The productivity gap between Europe and the
United States: trends and causes, Journal of Economic Perspectives, 22, 25-44.

88
Verspagen, B. 1995. R&D and productivity: A broad cross-section cross-country look, Journal of
Productivity Analysis, 6, 117-135.
von Tunzelmann, N. and V. Acha. 2005. Innovation in low-tech industries, in J. Fagerberg, D.C.
Mowery and R.R. Nelson (eds.), The Oxford Handbook of Innovation, New York: Oxford
University Press, 407-432.
Wakelin, K. 2001. Productivity growth and R&D expenditure in UK manufacturing firms, Research
Policy, 30, 1079-1090.
Winter, S.G. 1984. Schumpeterian competition in alternative technological regimes, Journal of
Economic Behaviour and Organisation, 5, 287-320.

89

Appendix

90

Tab. A7: Whole sample, US and EU (complete)

Whole sample US EU
Log(R&D stock per
employee)

0.205***
(0.006)
0.089***
(0.007)
0.107***
(0.007)
0.228***
(0.007)
0.098***
(0.008)
0.119***
(0.008)
0.144***
(0.013)
0.058***
(0.011)
0.074***
(0.010)
Log(Physical stock per
employee)

0.115***
(0.006)
0.093***
(0.006)
0.099***
(0.006)
0.106***
(0.007)
0.100***
(0.007)

0.102***
(0.007)
0.125***
(0.012)
0.053***
(0.011)
0.078***
(0.009)
Log(Employees)

0.031***
(0.003)
-0.049***
(0.007)
-0.012**
(0.007)
0.035***
(0.004)
-0.034*
(0.008)
-0.006
(0.006)
0.015**
(0.007)
-0.162***
(0.017)

-0.059***
(0.011)
Constant

0.860
(0.493)
3.529***
(0.038)
1.115
(0.984)
3.906***
(0.103)
3.523***
(0.036)
1.124
(0.697)
2.330***
(0.135)
3.744***
(0.079)
2.256**
(1.016)

Wald time-dummies
(p-value)

4.51***
(0.000)
11.41***
(0.000)
165.44***
(0.000)
5.41***
(0.000)
9.58***
(0.000)
199.22***
(0.000)
1.98***
(0.009)
2.39***
(0.001)
19.27
(0.313)
Wald country-dummies
(p-value)

52.46***
(0.000)
- 67.22***
(0.000)
- - - 18.62***
(0.000)
- 25.77*
(0.078)
Wald sectoral-dummies
(p-value)

174.22***
(0.000)
- 233.08***
(0.000)
86.37***
(0.000)
- 154.02***
(0.000)
99.85***
(0.000)
- 83.00***
(0.000)

R
2
(overall) 0.32 0.18 0.29 0.34 0.21 0.31 0.27 0.01 0.17
Obs. 16,079 12,605 3,474
N. of firms 1,809 1,170 639
Notes: - (Robust in POLS) standard-errors in parentheses; * significance at 10%, ** 5%, *** 1%.
- For Time-dummies, Country-dummies and Sectoral-dummies Wald test of joint significance are reported.

91
Tab. A8: Sectoral decomposition: Manufacturing (High-tech + Other) and Service sectors (complete)

Whole sample Manufacturing
sectors
Service
sectors
Log(R&D stock per
employee)

0.205***
(0.006)
0.089***
(0.007)
0.107***
(0.007)
0.209***
(0.007)
0.073***
(0.008)
0.095***
(0.007)
0.177***
(0.014)
0.113***
(0.013)
0.127***
(0.011)
employee)

0.115***
(0.006)
0.093***
(0.006)
0.099***
(0.006)
0.109***
(0.007)
0.086***
(0.007)

0.094***
(0.006)
0.147***
(0.014)
0.115***
(0.014)
0.125***
(0.012)
Log(Employees)

0.031***
(0.003)
-0.049***
(0.007)
-0.012**
(0.007)
0.025***
(0.004)
-0.078***
(0.009)
-0.029
(0.007)
0.062**
(0.009)
0.021
(0.014)

0.034***
(0.011)
Constant

0.860
(0.493)
3.529***
(0.038)
1.115
(0.984)
2.957***
(0.481)
3.559***
(0.036)
1.118
(1.018)
3.513***
(0.164)
3.744***
(0.084)
-0.163
(0.975)

Wald time-dummies
(p-value)

4.51***
(0.000)
11.41***
(0.000)
165.44***
(0.000)
4.28***
(0.000)
15.16***
(0.000)
216.51***
(0.000)
2.74***
(0.000)
3.13***
(0.000)
61.00***
(0.002)
(p-value)

52.46***
(0.000)
- 67.22***
(0.000)
72.87***
(0.000)
- 49.30***
(0.000)
13.47***
(0.000)
- 30.63***
(0.000)
(p-value)

174.22***
(0.000)
- 233.08***
(0.000)
69.91***
(0.000)
- 142.36***
(0.000)
94.79***
(0.000)
- 88.64***
(0.000)

R
2
(overall) 0.32 0.18 0.29 0.31 0.13 0.26 0.39 0.28 0.37
Obs. 16,079 12,876 3,203
N. of firms 1,809 1,383 426

92
Tab. A9: Sectoral decomposition: High-tech and Other manufacturing sectors (complete)

Whole sample High-tech manufacturing
sectors
Other manufacturing
sectors
Log(R&D stock per
employee)

0.205***
(0.006)
0.089***
(0.007)
0.107***
(0.007)
0.236***
(0.011)
0.070***
(0.012)
0.099***
(0.009)
0.158***
(0.007)
0.053***
(0.009)
0.071***
(0.008)
employee)

0.115***
(0.006)
0.093***
(0.006)
0.099***
(0.006)
0.117***
(0.010)
0.092***
(0.010)

0.100***
(0.011)
0.093***
(0.008)
0.069***
(0.008)
0.078***
(0.007)
Log(Employees)

0.031***
(0.003)
-0.049***
(0.007)
-0.012**
(0.007)
0.041***
(0.005)
-0.086***
(0.012)
-0.028***
(0.009)
-0.010*
(0.006)
-0.107***
(0.013)

-0.053***
(0.009)
Constant

0.860
(0.493)
3.529***
(0.038)
1.115
(0.984)
2.839***
(0.084)
3.514***
(0.051)
2.465***
(0.831)
3.144***
(0.457)
3.757***
(0.049)
1.286*
(0.761)

Wald time-dummies
(p-value)

4.51***
(0.000)
11.41***
(0.000)
165.44***
(0.000)
2.25***
(0.001)
10.22***
(0.000)
133.96***
(0.000)
2.53***
(0.000)
9.01***
(0.000)
138.51***
(0.000)
(p-value)

52.46***
(0.000)
- 67.22***
(0.000)
11.33***
(0.000)
- 26.36**
(0.034)
74.74***
(0.000)
- 45.67***
(0.000)
(p-value)

174.22***
(0.000)
- 233.08***
(0.000)
92.31***
(0.000)
- 28.03***
(0.000)
67.07***
(0.000)
- 196.77***
(0.000)

R
2
(overall) 0.32 0.18 0.29 0.28 0.09 0.22 0.40 0.08 0.36
Obs. 16,079 7,693 5,183
N. of firms 1,809 1,383 579

93
Tab. A10: US versus EU: Manufacturing sectors (complete)

Whole sample US
EU
Log(R&D stock per
employee)

0.205***
(0.006)
0.089***
(0.007)
0.107***
(0.007)
0.228***
(0.008)
0.078***
(0.009)
0.103***
(0.009)
0.147***
(0.014)
0.052***
(0.013)
0.070***
(0.012)
employee)

0.115***
(0.006)
0.093***
(0.006)
0.099***
(0.006)
0.099***
(0.008)
0.089***
(0.008)

0.093***
(0.008)
0.135***
(0.014)
0.059***
(0.013)
0.086***
(0.012)
Log(Employees)

0.031***
(0.003)
-0.049***
(0.007)
-0.012**
(0.007)
0.027***
(0.004)
-0.069***
(0.010)
-0.029
(0.008)
0.023***
(0.008)
-0.166***
(0.022)

-0.045***
(0.014)
Constant

0.860
(0.493)
3.529***
(0.038)
1.115
(0.984)
2.155***
(0.468)
3.560***
(0.040)
2.309
(1.158)
2.607***
(0.146)
3.769***
(0.093)
3.016***
(0.921)

Wald time-dummies
(p-value)

4.51***
(0.000)
11.41***
(0.000)
165.44***
(0.000)
4.89***
(0.000)
13.25***
(0.000)
204.38***
(0.000)
1.75*
(0.029)
2.14**
(0.004)
16.13
(0.514)
(p-value)

52.46***
(0.000)
- 67.22***
(0.000)
- - -

26.80***
(0.000)
- 20.60
(0.244)
(p-value)

174.22***
(0.000)
- 233.08***
(0.000)
73.72***
(0.000)
- 128.29***
(0.000)
14.41***
(0.000)
- 33.84
(0.139)

R
2
(overall) 0.32 0.18 0.29 0.32 0.15 0.28 0.28 0.01 0.17
Obs. 16,079 10,214 2,662
N. of firms 1,809 914 469

94
Tab. A11: US versus EU: Service sectors (complete)

Whole sample US
Service sectors
EU
Service sectors
Log(R&D stock per
employee)

0.205***
(0.006)
0.089***
(0.007)
0.107***
(0.007)
0.215***
(0.017)
0.125***
(0.017)
0.150***
(0.014)
0.126***
(0.023)
0.086***
(0.024)
0.097***
(0.018)
employee)

0.115***
(0.006)
0.093***
(0.006)
0.099***
(0.006)
0.149***
(0.017)
0.140***
(0.016)

0.143***
(0.015)
0.108***
(0.024)
0.045**
(0.022)
0.070***
(0.020)
Log(Employees)

0.031***
(0.003)
-0.049***
(0.007)
-0.012**
(0.007)
0.086***
(0.010)
0.051***
(0.016)
0.058
(0.013)
-0.017
(0.015)
-0.141***
(0.030)

-0.076**
(0.021)
Constant

0.860
(0.493)
3.529***
(0.038)
1.115
(0.984)
3.268***
(0.195)
3.755***
(0.093)
3.242***
(0.520)
-0.457***
(0.301)
3.623***
(0.185)
3.138***
(0.854)

Wald time-dummies
(p-value)

4.51***
(0.000)
11.41***
(0.000)
165.44***
(0.000)
3.36***
(0.000)
4.93***
(0.000)
92.64***
(0.000)
1.10
(0.348)
0.59
(0.902)
13.11
(0.728)
(p-value)

52.46***
(0.000)
- 67.22***
(0.000)
- - -

4.09***
(0.000)
- 15.53
(0.213)
(p-value)

174.22***
(0.000)
- 233.08***
(0.000)
84.71***
(0.000)
- 70.41***
(0.000)
73.50***
(0.000)
- 64.79***
(0.000)

R
2
(overall) 0.32 0.18 0.29 0.40 0.30 0.39 0.31 0.03 0.27
Obs. 16,079 2,391 812
N. of firms 1,809 256 170

95
Tab. A12: US versus EU: High-tech manufacturing sectors (complete)

Whole sample US
sectors
EU
sectors
Log(R&D stock per
employee)

0.205***
(0.006)
0.089***
(0.007)
0.107***
(0.007)
0.251***
(0.010)
0.069***
(0.013)
0.105***
(0.012)
0.172***
(0.032)
0.065***
(0.020)
0.081***
(0.019)
employee)

0.115***
(0.006)
0.093***
(0.006)
0.099***
(0.006)
0.112***
(0.011)
0.101***
(0.011)

0.105***
(0.010)
0.127***
(0.025)
0.029
(0.022)
0.061***
(0.020)
Log(Employees)

0.031***
(0.003)
-0.049***
(0.007)
-0.012**
(0.007)
0.041***
(0.005)
-0.080***
(0.013)
-0.028***
(0.010)
0.054***
(0.013)
-0.155**
(0.033)

-0.026
(0.023)
Constant

0.860
(0.493)
3.529***
(0.038)
1.115
(0.984)
3.147***
(0.074)
3.525***
(0.055)
3.060***
(0.205)
2.691***
(0.226)
3.499***
(0.166)
3.579***
(1.022)

Wald time-dummies
(p-value)

4.51***
(0.000)
11.41***
(0.000)
165.44***
(0.000)
2.23***
(0.002)
8.35***
(0.000)
112.82***
(0.000)
2.00***
(0.009)
2.44***
(0.000)
25.65*
(0.081)
(p-value)

52.46***
(0.000)
- 67.22***
(0.000)
- - -

11.55***
(0.000)
- 10.69
(0.710)
(p-value)

38.24***
(0.000)
- 233.08***
(0.000)
78.82***
(0.000)
- 31.22***
(0.000)
14.39***
(0.000)
- 2.41
(0.790)

R
2
(overall) 0.32 0.18 0.29 0.28 0.12 0.23 0.26 0.01 0.13
Obs. 16,079 6,462 1,231
N. of firms 1,809 591 213

96
Tab. A13: US versus EU: Other manufacturing sectors (complete)

Whole sample US
Other manufacturing
sectors
EU
Other manufacturing
sectors
Log(R&D stock per
employee)

0.205***
(0.006)
0.089***
(0.007)
0.107***
(0.007)
0.175***
(0.008)
0.060***
(0.010)
0.079***
(0.010)
0.112***
(0.013)
0.035**
(0.015)
0.056***
(0.014)
employee)

0.115***
(0.006)
0.093***
(0.006)
0.099***
(0.006)
0.074***
(0.009)
0.063***
(0.012)

0.066***
(0.009)
0.146***
(0.016)
0.093***
(0.016)
0.118***
(0.016)
Log(Employees)

0.031***
(0.003)
-0.049***
(0.007)
-0.012**
(0.007)
-0.007
(0.007)
-0.087**
(0.015)
-0.049***
(0.011)
-0.014
(0.009)
-0.209***
(0.028)

-0.075***
(0.017)
Constant

0.860
(0.493)
3.529***
(0.038)
1.115
(0.984)
2.351***
(0.446)
3.763***
(0.054)
2.501**
(0.487)
3.815***
(0.172)
4.006***
(0.159)
-0.245
(0.362)

Wald time-dummies
(p-value)

4.51***
(0.000)
11.41***
(0.000)
165.44***
(0.000)
4.20***
(0.000)
9.93***
(0.000)
171.01***
(0.000)
0.78
(0.718)
1.05
(0.397)
30.79**
(0.030)
(p-value)

52.46***
(0.000)
- 67.22***
(0.000)
- - -

28.72***
(0.000)
- 18.50
(0.237)
(p-value)

38.24***
(0.000)
- 233.08***
(0.000)
79.51***
(0.000)
- 207.28***
(0.000)
12.62***
(0.000)
- 44.26***
(0.001)

R
2
(overall) 0.32 0.18 0.29 0.44 0.12 0.41 0.35 0.01 0.27
Obs. 16,079 3,752 1,431
N. of firms 1,809 323 256

97
Tab. A14: European macroareas: North (Denmark, Finland, Sweden) + UK, Other EU countries, Other EU countries without South
(complete)
EU (overall) North + UK Other EU countries Other EU countries
(Italy, Greece and Spain excluded)
POLS FE RE POLS FE RE POLS FE RE POLS FE RE
Log(R&D stock per
employee)

0.144***
(0.013)
0.058***
(0.011)
0.074***
(0.010)
0.155***
(0.015)
0.091***
(0.016)
0.102***
(0.014)
0.146***
(0.024)
0.029*
(0.015)
0.050***
(0.013)
0.142***
(0.025)
0.025*
(0.014)
0.044***
(0.014)
employee)

0.125***
(0.012)
0.053***
(0.011)
0.078***
(0.009)
0.122***
(0.016)
0.064***
(0.013)
0.081***
(0.013)
0.126***
(0.019)
0.039**
(0.017)

0.069***
(0.015)
0.130***
(0.021)
0.038**
(0.018)
0.070***
(0.017)
Log(Employees)

0.015**
(0.007)
-0.162***
(0.017)

-0.059***
(0.011)
0.041
(0.010)
-0.128***
(0.023)

-0.040**
(0.017)
-0.030***
(0.010)
-0.193***
(0.024)
-0.089***
(0.017)
-0.027
(0.011)
-0.194
(0.025)
-0.085
(0.017)
Constant

2.330***
(0.135)
3.744***
(0.079)
2.256**
(1.016)
3.721***
(0.154)
3.496***
(0.139)
-1.382
(0.862)
2.351***
(0.446)
4.007***
(0.233)
3.026***
(1.061)
4.408***
(0.396)
4.020***
(0.236)
-1.171
(0.969)

Wald time-dummies
(p-value)

1.98***
(0.009)
2.39***
(0.001)
19.27
(0.313)
1.04
(0.409)
1.12
(0.323)
35.57***
(0.004)
2.51***
(0.000)
2.21***
(0.004)
30.51**
(0.015)
2.55***
(0.000)
2.23***
(0.003)
31.05**
(0.013)
(p-value)

18.62***
(0.000)
- 25.77*
(0.078)
16.47***
(0.000)
- 5.72
(0.126)
21.18***
(0.000)
- 20.32*
(0.087)

25.26***
(0.000)
- 17.51*
(0.063)
(p-value)

99.85***
(0.000)
- 83.00***
(0.000)
34.75***
(0.000)
- 37.53
(0.352)
97.98***
(0.000)
- 77.12***
(0.000)
158.19***
(0.000)
- 79.84***
(0.000)

R
2
(overall) 0.27 0.02 0.17 0.27 0.03 0.17 0.32 0.02 0.22 0.32 0.02 0.22
Obs. 3,474 1,994 1,480 1,413
N. of firms 639 347 292 273

98
Tab. A15: European macroareas: Manufacturing sectors (complete)

Other EU countries
Other EU countries
Log(R&D stock per
employee)

0.144***
(0.013)
0.058***
(0.011)
0.074***
(0.010)
0.154***
(0.015)
0.086***
(0.019)
0.102***
(0.018)
0.147***
(0.030)
0.019
(0.016)
0.040***
(0.014)
0.146***
(0.030)
0.017
(0.014)
0.037**
(0.015)
employee)

0.125***
(0.012)
0.053***
(0.011)
0.078***
(0.009)
0.125***
(0.019)
0.063***
(0.018)
0.082***
(0.017)
0.145***
(0.024)
0.052***
(0.020)

0.086***
(0.018)
0.150***
(0.024)
0.055***
(0.020)
0.092***
(0.019)
Log(Employees)

0.015**
(0.007)
-0.162***
(0.017)

-0.059***
(0.011)
0.067***
(0.011)
-0.139***
(0.030)

-0.022
(0.021)
-0.044***
(0.011)
-0.177**
(0.030)
-0.079***
(0.020)
-0.043***
(0.012)
-0.178
(0.031)
-0.078***
(0.020)
Constant

2.330***
(0.135)
3.744***
(0.079)
2.256**
(1.016)
3.763***
(0.177)
3.507***
(0.159)
4.611***
(0.909)
3.088***
(0.336)
4.065***
(0.296)
2.937***
(0.975)
1.909***
(0.164)
4.060***
(0.299)
3.014***
(1.123)

Wald time-dummies
(p-value)

1.98***
(0.009)
2.39***
(0.001)
19.27
(0.313)
1.43
(0.113)
1.18
(0.277)
9.88
(0.908)
2.20***
(0.004)
2.01***
(0.010)
26.88**
(0.042)
2.19***
(0.004)
1.96**
(0.013)
26.88**
(0.042)
(p-value)

18.62***
(0.000)
- 25.77*
(0.078)
16.33***
(0.000)
- 2.61
(0.455)
30.16***
(0.000)
- 15.67
(0.267)

25.99***
(0.000)
- 12.65
(0.244)
(p-value)

99.85***
(0.000)
- 83.00***
(0.000)
24.00***
(0.000)
- 22.32
(0.617)
12.60***
(0.000)
- 26.38***
(0.334)
16.11***
(0.000)
- 25.72
(0.367)

R
2
(overall) 0.27 0.02 0.17 0.28 0.02 0.17 0.32 0.02 0.23 0.33 0.02 0.23
Obs. 3,474 1,534 1,128 1,097
N. of firms 639 251 218 208

99
Tab. A16: European macroareas: Service sectors (complete)

Service sectors
Other EU countries
Service sectors
Other EU countries
Service sectors
Log(R&D stock per
employee)

0.144***
(0.013)
0.058***
(0.011)
0.074***
(0.010)
0.129***
(0.027)
0.113***
(0.030)
0.113***
(0.023)
0.106***
(0.028)
0.074*
(0.039)
0.081***
(0.031)
0.098***
(0.032)
0.077*
(0.044)
0.076*
(0.044)
employee)

0.125***
(0.012)
0.053***
(0.011)
0.078***
(0.009)
0.104***
(0.031)
0.069***
(0.026)
0.089***
(0.024)
0.143***
(0.034)
0.025
(0.040)

0.064*
(0.033)
0.136***
(0.036)
0.014
(0.044)
0.046
(0.036)
Log(Employees)

0.015**
(0.007)
-0.162***
(0.017)

-0.059***
(0.011)
-0.054***
(0.019)
-0.093**
(0.038)

-0.075***
(0.028)
0.056**
(0.023)
-0.194***
(0.049)
-0.024
(0.030)
0.075***
(0.023)
-0.202***
(0.053)
-0.023
(0.033)
Constant

2.330***
(0.135)
3.744***
(0.079)
2.256**
(1.016)
4.330***
(0.252)
3.554***
(0.327)
3.702***
(0.810)
-0.488
(0.317)
3.624***
(0.255)
3.175***
(0.781)
3.296***
(0.353)
3.553***
(0.2267)
3.005***
(0.669)

Wald time-dummies
(p-value)

1.98***
(0.009)
2.39***
(0.001)
19.27
(0.313)
0.79
(0.707)
0.29
(0.997)
5.73
(0.994)
1.94**
(0.021)
0.84
(0.625)
17.56
(0.286)
2.12**
(0.011)
0.84
(0.628)
15.48
(0.346)
(p-value)

18.62***
(0.000)
- 25.77*
(0.078)
5.90***
(0.000)
- 6.37*
(0.095)
5.57***
(0.000)
- 13.76
(0.088)

8.89***
(0.000)
- 11.47**
(0.042)
(p-value)

99.85***
(0.000)
- 83.00***
(0.000)
34.15***
(0.000)
- 16.30***
(0.060)
63.13***
(0.000)
- 66.38***
(0.000)
85.75***
(0.000)
- 63.08***
(0.000)

R
2
(overall) 0.27 0.02 0.17 0.31 0.11 0.28 0.42 0.01 0.35 0.43 0.01 0.35
Obs. 3,474 460 352 316
N. of firms 639 96 74 65

100
Tab. A17: European macroareas: High-tech manufacturing sectors (complete)
sectors
Other EU countries
sectors
Other EU countries
sectors
Log(R&D stock per
employee)

0.144***
(0.013)
0.058***
(0.011)
0.074***
(0.010)
0.198***
(0.027)
0.196***
(0.037)
0.200***
(0.033)
0.106***
(0.028)
0.013
(0.022)
0.029
(0.021)
0.161***
(0.055)
0.013
(0.023)
0.029
(0.021)
employee)

0.125***
(0.012)
0.053***
(0.011)
0.078***
(0.009)
0.094***
(0.035)
0.003
(0.031)
0.037
(0.029)
0.161***
(0.056)
0.068**
(0.028)

0.089***
(0.027)
0.135***
(0.040)
0.071**
(0.030)
0.096***
(0.028)
Log(Employees)

0.015**
(0.007)
-0.162***
(0.017)

-0.059***
(0.011)
0.139***
(0.016)
-0.100**
(0.047)

0.041
(0.033)
0.132**
(0.039)
-0.129***
(0.047)
-0.067**
(0.032)
-0.073***
(0.019)
-0.132***
(0.048)
-0.067**
(0.033)
Constant

2.330***
(0.135)
3.744***
(0.079)
2.256**
(1.016)
3.346***
(0.358)
3.050***
(0.294)
3.190***
(0.550)
-0.073***
(0.019)
4.088***
(0.173)
-0.111
(0.571)
2.725***
(0.206)
3.553***
(0.226)
3.001***
(0.887)

Wald time-dummies
(p-value)

1.98***
(0.009)
2.39***
(0.001)
19.27
(0.313)
1.99***
(0.010)
1.95***
(0.010)
21.76
(0.194)
1.04
(0.407)
0.85
(0.619)
36.46***
(0.001)
0.96
(0.489)
0.82
(0.650)
16.15
(0.304)
(p-value)

18.62***
(0.000)
- 25.77*
(0.078)
16.29***
(0.000)
- 26.90***
(0.000)
8.09***
(0.000)
- 9.96
(0.444)

6.87***
(0.000)
- 9.07
(0.336)
(p-value)

99.85***
(0.000)
- 83.00***
(0.000)
17.84***
(0.000)
- 1.86
(0.868)
3.98***
(0.001)
- 1.55
(0.906)
3.98***
(0.001)
- 1.44
(0.920)

R
2
(overall) 0.27 0.02 0.17 0.31 0.01 0.30 0.35 0.06 0.27 0.34 0.06 0.26
Obs. 3,474 734 497 482
N. of firms 639 122 91 88

101
Tab. A18 : European macroareas: Other manufacturing sectors (complete)
Other manufacturing
sectors
Other EU countries
Other manufacturing
sectors
Other EU countries
Other manufacturing
sectors
Log(R&D stock per
employee)

0.144***
(0.013)
0.058***
(0.011)
0.074***
(0.010)
0.118***
(0.019)
0.027***
(0.012)
0.048***
(0.018)
0.116***
(0.021)
0.037
(0.024)
0.062***
(0.020)
0.116***
(0.021)
0.029
(0.025)
0.057*
(0.031)
employee)

0.125***
(0.012)
0.053***
(0.011)
0.078***
(0.009)
0.132***
(0.020)
0.110***
(0.020)
0.120***
(0.019)
0.179***
(0.026)
0.059*
(0.031)

0.118***
(0.026)
0.184***
(0.027)
0.056*
(0.031)
0.119***
(0.026)
Log(Employees)

0.015**
(0.007)
-0.162***
(0.017)

-0.059***
(0.011)
-0.026*
(0.013)
-0.218***
(0.037)

-0.101***
(0.025)
-0.024*
(0.013)
-0.022***
(0.042)
-0.089***
(0.025)
-0.022*
(0.013)
-0.227***
(0.042)
-0.086***
(0.026)
Constant

2.330***
(0.135)
3.744***
(0.079)
2.256**
(1.016)
4.491***
(0.184)
3.835***
(0.174)
4.656***
(0.789)
3.437***
(0.217)
4.181***
(0.282)
2.612
(0.891)
3.056***
(0.362)
4.224***
(0.282)
2.001
(1.287)

Wald time-dummies
(p-value)

1.98***
(0.009)
2.39***
(0.001)
19.27
(0.313)
0.28
(0.998)
0.50
(0.953)
11.61
(0.823)
1.62*
(0.058)
2.90***
(0.000)
30.59**
(0.015)
1.63*
(0.056)
3.04
(0.000)
41.80***
(0.000)
(p-value)

18.62***
(0.000)
- 25.77*
(0.078)
1.91
(0.125)
- 0.41
(0.937)
31.74***
(0.000)
- 14.23
(0.220)

26.54***
(0.000)
- 11.58
(0.171)
(p-value)

99.85***
(0.000)
- 83.00***
(0.000)
14.38***
(0.000)
- 33.05**
(0.02)
14.49***
(0.001)
- 34.64***
(0.010)
23.30***
(0.000)
- 34.81***
(0.010)

R
2
(overall) 0.27 0.02 0.17 0.32 0.03 0.22 0.42 0.01 0.31 0.43 0.01 0.31
Obs. 3,474 800 631 615
N. of firms 639 129 127 120

102
Tab. A20: European NUTS: Innovative NUTS versus Weakly innovative NUTS (Regional BERD/GDP >= 1.8% is the threshold)
(complete)

EU (overall) Innovative NUTS Weakly innovative NUTS
Log(R&D stock per
employee)

0.144***
(0.013)
0.057***
(0.011)
0.073***
(0.009)
0.160***
(0.020)
0.072***
(0.016)
0.087***
(0.014)
0.119***
(0.019)
0.044***
(0.015)
0.057***
(0.013)
employee)

0.122***
(0.012)
0.053***
(0.011)
0.079***
(0.010)
0.091***
(0.018)
-0.010
(0.017)

0.031**
(0.015)
0.145***
(0.017)
0.093***
(0.014)
0.111***
(0.014)
Log(Employees)

0.014**
(0.007)
-0.162***
(0.017)
-0.056***
(0.011)
0.037
(0.011)
-0.174***
(0.024)
-0.054***
(0.017)
-0.002
(0.010)
-0.166***
(0.023)

-0.064***
(0.016)
Constant

-1.642***
(0.165)
3.751***
(0.079)
-1.151
(1.016)
3.835***
(0.134)
3.637***
(0.202)
3.210**
(0.487)
-1.116***
(0.187)
3.739***
(0.169)
3.777***
(0.752)

Wald time-dummies
(p-value)

1.89**
(0.014)
2.35***
(0.001)
17.34
(0.431)
1.80**
(0.022)
3.04***
(0.000)
25.50**
(0.084)
1.08
(0.365)
0.58
(0.910)
9.37
(0.950)
(p-value)

17.38***
(0.000)
- 27.02*
(0.057)
6.52***
(0.000)
- 16.76**
(0.019)

15.36***
(0.000)
- 15.51
(0.415)
(p-value)

100.42***
(0.000)
- 88.02***
(0.000)
38.35***
(0.000)
- 40.62
(0.274)
82.94***
(0.000)
- 87.00***
(0.000)

R
2
(overall) 0.28 0.01 0.18 0.26 0.01 0.15 0.38 0.01 0.30
Obs. 3,431 1,827 1,604
N. of firms 626 328 298

103
Tab. A21: European NUTS: Innovative NUTS versus Weakly innovative NUTS in High-tech manufacturing sectors (complete)

sectors
sectors
Log(R&D stock per
employee)

0.144***
(0.013)
0.057***
(0.011)
0.073***
(0.009)
0.109***
(0.034)
0.097***
(0.033)
0.095***
(0.030)
0.188***
(0.051)
0.035
(0.023)
0.054***
(0.022)
employee)

0.122***
(0.012)
0.053***
(0.011)
0.079***
(0.010)
0.143***
(0.036)
0.004
(0.037)

0.067**
(0.032)
0.135***
(0.036)
0.050**
(0.024)
0.072***
(0.024)
Log(Employees)

0.014**
(0.007)
-0.162***
(0.017)
-0.056***
(0.011)
0.085
(0.016)
-0.131***
(0.050)
0.018
(0.033)
0.002
(0.020)
-0.190***
(0.042)

-0.065**
(0.030)
Constant

-1.642***
(0.165)
3.751***
(0.079)
-1.151
(1.016)
1.802***
(0.201)
3.436***
(0.431)
2.633***
(1.012)
3.780***
(0.150)
3.530***
(0.163)
0.057
(0.440)

Wald time-dummies
(p-value)

1.89**
(0.014)
2.35***
(0.001)
17.34
(0.431)
6.05***
(0.000)
1.03
(0.423)
10.92
(0.860)
2.35***
(0.001)
2.38***
(0.001)
47.71***
(0.000)
(p-value)

17.38***
(0.000)
- 27.02*
(0.057)
9.01***
(0.000)
- 4.52
(0.718)

5.37***
(0.000)
- 11.48
(0.404)
(p-value)

100.42***
(0.000)
- 88.02***
(0.000)
9.19***
(0.000)
- 9.19
(0.163)
10.17***
(0.000)
- 6.33
(0.275)

R
2
(overall) 0.28 0.01 0.18 0.25 0.01 0.16 0.40 0.01 0.24
Obs. 3,431 688 529
N. of firms 626 114 96

104
Tab. A22: European NUTS: Innovative NUTS versus Weakly innovative NUTS in Other manufacturing sectors (complete)

Other manufacturing
sectors
Other manufacturing
sectors
Log(R&D stock per
employee)

0.144***
(0.013)
0.057***
(0.011)
0.073***
(0.009)
0.141***
(0.019)
0.024
(0.021)
0.059***
(0.019)
0.065***
(0.020)
0.040*
(0.022)
0.047**
(0.020)
employee)

0.122***
(0.012)
0.053***
(0.011)
0.079***
(0.010)
0.085***
(0.020)
0.006
(0.021)

0.027
(0.020)
0.203***
(0.025)
0.140***
(0.025)
0.176***
(0.022)
Log(Employees)

0.014**
(0.007)
-0.162***
(0.017)
-0.056***
(0.011)
0.001
(0.013)
-0.166***
(0.042)
-0.043
(0.027)
-0.023
(0.015)
-0.262***
(0.039)

-0.094***
(0.024)
Constant

-1.642***
(0.165)
3.751***
(0.079)
-1.151
(1.016)
4.311***
(0.221)
3.789***
(0.283)
3.919***
(0.737)
2.784***
(0.246)
3.906***
(0.150)
1.976
(0.746)

Wald time-dummies
(p-value)

1.89**
(0.014)
2.35***
(0.001)
17.34
(0.431)
1.76**
(0.029)
2.60***
(0.000)
28.31**
(0.041)
0.92
(0.552)
0.66
(0.844)
19.08
(0.387)
(p-value)

17.38***
(0.000)
- 27.02*
(0.057)
4.78***
(0.000)
- 6.38
(0.496)

27.22***
(0.000)
- 16.67
(0.214)
(p-value)

100.42***
(0.000)
- 88.02***
(0.000)
49.73***
(0.000)
- 32.26**
(0.040)
12.85***
(0.000)
- 37.98***
(0.006)

R
2
(overall) 0.28 0.01 0.18 0.39 0.01 0.17 0.45 0.01 0.38
Obs. 3,431 670 749
N. of firms 626 124 129

105
Tab. A23: European NUTS: Innovative NUTS versus Weakly innovative NUTS in Manufacturing sectors (complete)

Log(R&D stock per
employee)

0.144***
(0.013)
0.057***
(0.011)
0.073***
(0.009)
0.129***
(0.021)
0.068***
(0.020)
0.084***
(0.018)
0.128***
(0.023)
0.038**
(0.016)
0.053***
(0.015)
employee)

0.122***
(0.012)
0.053***
(0.011)
0.079***
(0.010)
0.115***
(0.022)
0.002
(0.021)

0.044**
(0.020)
0.167***
(0.020)
0.098***
(0.017)
0.120***
(0.016)
Log(Employees)

0.014**
(0.007)
-0.162***
(0.017)
-0.056***
(0.011)
0.056***
(0.012)
-0.149***
(0.033)
-0.008
(0.021)
-0.004
(0.011)
-0.193***
(0.027)

-0.072***
(0.018)
Constant

-1.642***
(0.165)
3.751***
(0.079)
-1.151
(1.016)
3.685***
(0.220)
3.744***
(0.223)
0.730
(0.776)
2.837***
(0.251)
3.740***
(0.104)
1.208
(1.053)

Wald time-dummies
(p-value)

1.89**
(0.014)
2.35***
(0.001)
17.34
(0.431)
1.37
(0.143)
2.15***
(0.000)
24.27
(0.146)
1.17
(0.284)
0.77
(0.735)
8.88
(0.944)
(p-value)

17.38***
(0.000)
- 27.02*
(0.057)
34.02***
(0.000)
- 7.69
(0.361)

21.80***
(0.000)
- 18.34
(0.245)
(p-value)

100.42***
(0.000)
- 88.02***
(0.000)
8.21***
(0.000)
- 25.53
(0.489)
13.22***
(0.000)
- 36.43***
(0.065)

R
2
(overall) 0.28 0.01 0.18 0.27 0.01 0.17 0.39 0.02 0.29
Obs. 3,431 1,358 1,278
N. of firms 626 238 225

106
Tab. A24: European NUTS: Innovative NUTS vs. Weakly innovative NUTS in Service sectors (complete)

Service sectors
Service sectors
Log(R&D stock per
employee)

0.144***
(0.013)
0.057***
(0.011)
0.073***
(0.009)
0.207***
(0.032)
0.096***
(0.029)
0.118***
(0.024)
0.059**
(0.027)
0.068
(0.043)
0.056*
(0.029)
employee)

0.122***
(0.012)
0.053***
(0.011)
0.079***
(0.010)
0.056**
(0.028)
-0.007
(0.033)

0.008
(0.030)
0.088***
(0.033)
0.089**
(0.035)
0.098***
(0.030)
Log(Employees)

0.014**
(0.007)
-0.162***
(0.017)
-0.056***
(0.011)
-0.006***
(0.023)
-0.199***
(0.040)
-0.123***
(0.029)
-0.008
(0.022)
-0.081***
(0.051)

-0.024
(0.031)
Constant

-1.642***
(0.165)
3.751***
(0.079)
-1.151
(1.016)
4.601***
(0.336)
3.706***
(0.204)
3.439***
(0.700)
-0.025
(0.346)
3.469***
(0.426)
2.987***
(0.859)

Wald time-dummies
(p-value)

1.89**
(0.014)
2.35***
(0.001)
17.34
(0.431)
1.40
(0.132)
1.20
(0.258)
17.76
(0.404)
1.94**
(0.016)
0.24
(0.991)
5.28
(0.980)
(p-value)

17.38***
(0.000)
- 27.02*
(0.057)
14.42***
(0.000)
- 14.02**
(0.029)

4.43***
(0.000)
- 61.55***
(0.000)
(p-value)

100.42***
(0.000)
- 88.02***
(0.000)
10.81***
(0.000)
- 23.22***
(0.002)
67.34***
(0.000)
- 10.86
(0.285)

R
2
(overall) 0.28 0.01 0.18 0.36 0.05 0.26 0.44 0.03 0.41
Obs. 3,431 469 326
N. of firms 626 90 73

107
Tab. A25: Whole sample, Recessions and Expansions (complete)

Whole sample Recessions Expansions
Log(R&D stock per
employee)

0.205***
(0.006)
0.089***
(0.007)
0.107***
(0.007)
0.213***
(0.009)
0.089***
(0.010)
0.130***
(0.008)
0.196***
(0.009)
0.088***
(0.009)
0.116***
(0.008)
employee)

0.115***
(0.006)
0.093***
(0.006)
0.099***
(0.006)
0.114***
(0.008)
0.110***
(0.009)
0.113***
(0.008)
0.116***
(0.008)
0.085***
(0.008)
0.096***
(0.007)
Log(Employees)

0.031***
(0.003)
-0.049***
(0.007)
-0.012**
(0.007)
0.032***
(0.004)
-0.046***
(0.011)
0.004
(0.007)
0.030***
(0.004)
-0.050***
(0.011)
0.002
(0.007)
Constant

0.860
(0.493)
3.529***
(0.038)
1.115
(0.984)
3.575***
(0.159)
3.572***
(0.094)
0.301
(0.990)
3.587***
(0.040)
3.529***
(0.038)
2.046**
(1.024)

Wald time-dummies
(p-value)

4.51***
(0.000)
11.41***
(0.000)
165.44***
(0.000)
3.30***
(0.000)
7.82***
(0.000)
65.93***
(0.000)
7.70***
(0.000)
11.41***
(0.000)
75.14***
(0.000)
(p-value)

52.46***
(0.000)
- 67.22***
(0.000)
23.08***
(0.000)
- 65.36***
(0.010)
17.84***
(0.000)
- 34.76***
(0.010)
(p-value)

174.22***
(0.000)
- 233.08***
(0.000)
139.39***
(0.000)
- 228.96***
(0.000)
94.79***
(0.000)
- 195.37***
(0.000)

R
2
(overall) 0.32 0.18 0.29 0.34 0.21 0.31 0.31 0.16 0.29
Obs. 16,079 8,073 8,006
N. of firms 1,809 1,728 1,590

108
Tab. A26: US versus EU: Recessions (complete)

Recessions US EU
Log(R&D stock per
employee)

0.213***
(0.009)
0.089***
(0.010)
0.130***
(0.008)
0.230***
(0.009)
0.008***
(0.012)
0.135***
(0.010)
0.156***
(0.018)
0.082***
(0.022)
0.116***
(0.016)
employee)

0.114***
(0.008)
0.110***
(0.009)
0.113***
(0.008)
0.103***
(0.010)
0.115***
(0.010)

0.111***
(0.009)
0.137***
(0.019)
0.064***
(0.022)
0.114***
(0.017)
Log(Employees)

0.032***
(0.004)
-0.046***
(0.011)
0.004
(0.007)
0.033***
(0.006)
-0.038***
(0.011)
0.003
(0.035)
0.023
(0.010)
-0.173***
(0.032)

-0.012
(0.014)
Constant

3.575***
(0.159)
3.572***
(0.094)
0.301
(0.990)
3.889***
(0.110)
3.517***
(0.045)
1.813
(0.729)
1.913***
(0.196)
3.577***
(0.147)
-0.667
(0.698)

Wald time-dummies
(p-value)

3.30***
(0.000)
7.82***
(0.000)
65.93***
(0.000)
5.71***
(0.000)
10.62***
(0.000)
68.88***
(0.000)
0.56
(0.811)
1.51
(0.150)
11.0460
(0.273)
(p-value)

23.08***
(0.000)
- 65.36***
(0.010)
- - 4.84**
(0.027)
10.54***
(0.000)
- 16.40
(0.495)
(p-value)

139.39***
(0.000)
- 228.96***
(0.000)
137.52***
(0.000)
- 211.31***
(0.000)
12.37***
(0.000)
- 47.39***
(0.167)

R
2
(overall) 0.34 0.20 0.33 0.35 0.210 0.33 0.29 0.02 0.27
Obs. 8,073 6,522 1,551
N. of firms 1,728 1,151 577

109
Tab. A27: US versus EU: Expansions (complete)

Expansions US EU
Log(R&D stock per
employee)

0.196***
(0.009)
0.088***
(0.009)
0.116***
(0.008)
0.225***
(0.009)
0.104***
(0.012)
0.138***
(0.010)
0.137***
(0.017)
0.050***
(0.014)
0.077***
(0.011)
employee)

0.116***
(0.008)
0.085***
(0.008)
0.096***
(0.007)
0.109***
(0.010)
0.094***
(0.010)

0.099***
(0.009)
0.113***
(0.015)
0.044***
(0.015)
0.070***
(0.013)
Log(Employees)

0.030***
(0.004)
-0.050***
(0.011)
0.002
(0.007)
0.038***
(0.005)
-0.035***
(0.012)
0.009
(0.008)
0.009
(0.010)
-0.124***
(0.022)

-0.029***
(0.013)
Constant

3.587***
(0.040)
3.529***
(0.038)
2.046**
(1.024)
3.494***
(0.138)
3.676***
(0.058)
0.349
(0.743)
2.459***
(0.127)
3.962***
(0.081)
3.746***
(0.787)

Wald time-dummies
(p-value)

7.70***
(0.000)
11.41***
(0.000)
75.14***
(0.000)
4.27***
(0.000)
7.93***
(0.000)
79.52***
(0.000)
2.25**
(0.021)
2.22**
(0.023)
12.60
(0.181)
(p-value)

17.84***
(0.000)
- 34.76***
(0.010)
- - - 12.13***
(0.000)
- 19.18
(0.318)
(p-value)

94.79***
(0.000)
- 195.37***
(0.000)
62.03***
(0.000)
- 162.89***
(0.000)
84.76***
(0.000)
- 70.86***
(0.001)

R
2
(overall) 0.31 0.16 0.29 0.34 0.21 0.32 0.25 0.01 0.19
Obs. 8,006 6,083 1,923
N. of firms 1,590 1,026 564

Corporate R&D and Productivity

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Corporate R&D and Productivity

Uploaded by

Copyright:

Available Formats

1

= with Ymin denoting the minimum value;

You might also like

Corporate R&amp;D and Productivity

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Corporate R&amp;D and Productivity

Uploaded by

Copyright:

Available Formats

1

= with Ymin denoting the minimum value;

You might also like

Corporate R&D and Productivity

Corporate R&D and Productivity