UNIVERSIDADE TÉCNICA DE LISBOA
INSTITUTO SUPERIOR TÉCNICO
Innovation and Productivity: What can we
learn from the CIS III Results for Portugal?
Pedro Morais Martins de Faria
(Licenciado)
Dissertação para a obtenção do Grau de Mestre
em Engenharia e Gestão de Tecnologia
Doutor Pedro Filipe Teixeira da Conceição
Orientador:
Co-Orientadora: Doutora Elsa Beatriz Padilla
Presidente:
Doutor Manuel Frederico Tojal de Valsassina Heitor
Vogais:
Doutor Rui Miguel Loureiro Nobre Baptista
Doutor Francisco Miguel Rogado Salvador Pinheiro Veloso
Doutor Pedro Filipe Teixeira da Conceição
Dezembro 2004
Título: Inovação e Produtividade: O que podemos aprender com os resultados do CIS III
para Portugal?
Nome: Pedro Morais Martins de Faria
Curso de Mestrado em: Engenharia e Gestão de Tecnologia
Orientador: Pedro Filipe Teixeira da Conceição, Professor Auxiliar, IST-UTL
Co-Orientador: Elsa Beatriz Padilla
Provas Concluídas em:
Resumo
É comum assumir-se que a inovação contribui para o crescimento da produtividade. Embora a
relação positiva entre inovação e produtividade seja inquestionável a longo prazo e ao nível
macro, há mais ambiguidade no curto prazo e ao nível micro. É também neste domínio
(análises micro e de curto prazo) que há menos trabalho empírico, especialmente dadas as
dificuldades na obtenção de “matching data” ao nível da empresa sobre inovação e
produtividade.
O contributo desta tese insere-se neste domínio. Utilizando dados do terceiro inquérito
comunitário à inovação (CIS III) para Portugal, e na sequência da abordagem proposta por
Conceição et al. (2003), os resultados suportam a hipótese de que empresas inovadoras têm
um menor crescimento de produtividade quando comparadas com a média no curto prazo
(período de dois anos). O problema econométrico de endogeneidade entre produtividade e
inovação é resolvido através do uso de variáveis instrumentais. Esta tese apresenta, pela
primeira vez na literatura, a inclusão de uma variável que aproxima determinantes da
produtividade ligados à estratégia e gestão da empresa. Esta contribuição original da tese
mostra que a relação negativa entre inovação e crescimento de produtividade no curto prazo e
ao nível micro é robusta à inclusão desses determinantes.
Palavras-Chave: Inovação, Produtividade, Inquérito Comunitário à Inovação, Portugal;
Murphy-Topel, Endogeneidade
I
Title: Innovation and Productivity: What can we learn from the CIS III Results for
Portugal?
Abstract
Often, it is assumed that innovation is a determinant of productivity growth. Despite the fact
that the positive nature of the relationship between innovation and productivity is
unquestionable in the long run and at the macro level, there exists an ambiguity at the short
run and at the micro level. It is also in this domain (micro and short run analysis) that exists a
fewer number of empirical works, a consequence of the difficulty to obtain matching data
about innovation and productivity at the firm level.
The contribution of this thesis is included on this domain. Using data from the third
community innovation survey (CIS III) for Portugal, and in line with Conceição et al. (2003),
the results support the hypothesis that innovative firms have a lower degree of productivity
growth when compared with non-innovative firms in the short run (two years period). The
econometric problem of endogeneity between productivity and innovation is solved through
the use of instrumental variables. This thesis presents, for the first time in the literature, the
inclusion of a variable that measures the management and strategy of the firm. This unique
contribution shows that the negative relationship between innovation and productivity growth
in the short run and at the micro level is robust to the inclusion of these determinants.
Keywords: Innovation, Productivity, Community Innovation Survey, Portugal, MurphyTopel, Endogeneity
II
Acknowledgments
First, I would like to thank my thesis supervisors, Pedro Conceição and Beatriz Padilla, for
the motivation, the teaching and, specially, their friendship.
The research reported in this thesis was partially supported by Observatório da Ciência e do
Ensino Superior (OCES) [Observatory of Science and Higher Education, Ministry for Science
and Higher Education, Portugal], to whom I would like to thank.
I would also like to thank Francisco Veloso and Pedro Ferreira for their crucial help in the
econometric analysis and in the results interpretation.
To Professor Manuel Heitor, the Coordinator of the Master in Engineering Policy and
Management of Technology, and Director of the IN+ research center, I would like to thank
for the research environment he created in the master program and at the IN+.
Finally, I special thank to my IN+ colleagues for their opinions about my research and for all
the support, in particular to Manuel João Bóia, Miguel Amaral and Miguel Torres Preto.
III
Table of Contents
Chapter 1 - Introduction.......................................................................................1
2.1 - Productivity and Innovation: What we Know, What we Need to Know: A Literature
Review.................................................................................................................................. 5
2.1.1 - Productivity Determinants .................................................................................... 5
2.1.1.1 - Productivity Definition and Macro Studies ................................................... 5
2.1.1.2 - Firm Level Studies......................................................................................... 8
2.1.2 - What is the relationship between innovation and productivity?......................... 11
2.1.2.1 - Innovation .................................................................................................... 11
2.1.2.2 - Innovation vs. Productivity.......................................................................... 13
2.1.2.3 - ICT and Productivity.................................................................................... 16
2.1.2.4 - Final Remarks on the Literature Review ..................................................... 23
2.2 - Productivity and Innovation: Hypotheses and Observable Implications................... 24
2.2.1 - Learning .............................................................................................................. 25
2.2.2 - Technology and Organizational Rigidities ......................................................... 27
2.2.3 - Adjustments Costs .............................................................................................. 28
Chapter 3 - Data ..................................................................................................32
3.1 - The Third Community Innovation Survey................................................................. 32
3.1.1 - The CIS Methodology......................................................................................... 32
3.1.2 - Studies using the CIS Data ................................................................................. 35
3.1.3 - CIS Methodology Limitations ............................................................................ 38
Chapter 4 – Model and Methods .......................................................................40
4.1 – The Model ................................................................................................................. 40
4.1.1 - Productivity Equation ......................................................................................... 40
4.1.2 - Endogeneity ........................................................................................................ 42
4.1.3 - Innovation Equation............................................................................................ 44
4.2 - Methods...................................................................................................................... 45
4.2.1 - Murphy Topel Application ................................................................................. 46
Chapter 5 – Results and Conclusions ................................................................50
5.1 - Results and Discussion .............................................................................................. 50
5.1.1 - Descriptive Statistics........................................................................................... 50
IV
5.1.2 Regression results.................................................................................................. 53
5.1.2.1 - Complete Sample ......................................................................................... 53
5.1.2.2 - Complete Sample (process and product innovations).................................. 54
5.1.2.3 - Model with the level of productivity in 2000 .............................................. 55
5.2 - Conclusions................................................................................................................ 56
References ............................................................................................................60
Annex I – The Community Innovation Survey ................................................71
Annex II - Endogeneity and the Hausman Test ...............................................91
V
List of Tables
Table 1- Theoretical arguments that explain the negative relationship between innovation and
productivity ........................................................................................................... 30
Table 2 - Surveyed Sub-sectors................................................................................................ 41
Table 3 – Descriptive Statistics................................................................................................ 50
Table 4 – Descriptive Statistics (continuation) ........................................................................ 51
Table 5 – Regression results for the complete sample stage.................................................... 53
Table 6 – Regression results for the complete sample (process and product innovations)...... 55
Table 7 - Regression results for the model with the level of productivity in 2000 as dependent
variable .................................................................................................................. 55
Table 8 - Hausman Test Stata Output ...................................................................................... 92
VI
Abbreviations
2SLS – Two Stages Least Squares
CIS – Community Innovation Survey
DEA – Data Envelopment Analysis
GDP - Gross Domestic Product
ICBT – Information, Communication and Biotechnologies
ICT – Information and Communication Technologies
IT – Information Technologies
IV – Instrumental Variable
LIML – Limited Information Maximum Likelihood
MFP – Multifactor Productivity
OLS – Ordinary Least Squares
SIEPI - Structure of Innovation and Economic Performance Indicators
TFP – Total Factor Productivity
TFPG - Total Factor Productivity Growth
VII
Chapter 1 - Introduction
With global changes and economic re-structuring, technological change and productivity are
fundamental issues in which firms have to make the right decisions in order to be competitive.
Many decisions relate to innovation, that is, whether or not to introduce new products or
processes, when and how. Several studies have focused on deepening the knowledge about
the role of technological change in relation to output and productivity growth. Stoneman
(2002) stated that, in this area, there are large theoretical and empirical works of which the
most developed part explores the determinants of productivity growth. However, there are
some edges and aspects that have been overlooked, especially studies that consider the
consequences of innovation in the short run.
To argue that productivity is a source of economic growth, and that productivity is a function
of technological change, is equivalent to asserting that these characteristics of society are the
decisive factors underlying economic growth. In this context, economic history researchers
have revealed the crucial role played by technology in economic growth, through productivity
increase, throughout history and especially in the industrial era.
Despite this important role of technology, for new technological innovations to be able to
diffuse throughout the entire economy and enhancing productivity growth at a considerable
rate, culture and social institutions, business firms, and the factors intervening in the
production process need to experience a significant change. So, it is important not to reduce
the time lag between the introduction of a technology and its effects in productivity to a black
box. Thus, some questions can be raised: why and how new technologies had to wait to
enhance productivity? Which is the ideal environment for such enhancement? How does it
differ depending on the characteristics of technology? How different is the rate of diffusion of
technology and impact on productivity, in different industries? (Castells 1996)
In other words, the relationship between innovation and productivity is the subject of a vast
range of studies but, if this relationship is expected to be positive in the long run and at the
macro level (countries and regions), in the short run and at the micro level (firms) there exist
some work that suggest the existence of a negative relation linking these two variables. In this
thesis, we will describe three theoretical arguments that justify the negative relationship
1
between productivity and innovation in the short run, in order to contextualize the results
obtained.
In addition, the direction of causation between productivity and innovation is not clear: are
firms more productive because they engage in innovation activities or do firms innovate
because they are more productive? This relationship is another version of the often cited
question: Which came first, the chicken or the egg? Therefore, these two variables are
considered to be endogenous in this study and, in consequence, are jointly determined by
using a set of instruments to predict the correct model.
The database used in this study is the Third Community Innovation Survey (CIS III) database
that contains information about several firms’ characteristics in a two year period (19982000). The analysis used in the present thesis corrects for the fact that innovation and
productivity change are simultaneously determined. In other words, this database made
possible the employment of instruments that helped to deal with the problem of endogeneity
between variables.
The literature about the relationship between productivity and innovation has mainly focused
on the comprehension of productivity determinants and, in recent years, on the effect of
Information and Communication Technologies (ICT) investments on productivity.
Crépon et al. (1998) found a correlation between innovation and high levels of productivity
acknowledging the problem of endogeneity. Still, our results are different from those obtained
by Crépon et al. (1998) because their work took measures of innovation over a longer period
of time (five years). Therefore, they measured the effect of innovation on productivity in the
medium run and not in the short run, which is the aim of the present study. The short run
analysis of this relationship is an important issue since the understanding of the first steps
effects of innovation on the firms’ functioning can give explanations to momentary
productivity and efficiency losses. So, this kind of analysis can prevent precipitated
judgments about the value of a specific innovation, since it supports the idea that innovation
efficiency is only correctly evaluated after a time period. In order to contextualize the analysis
done on this work, a more complete analysis of the state of the art will be done in the
following chapter of this thesis.
Within this framework, the present thesis presents support to the hypothesis that innovating
firms grow less in productivity than non-innovating firms, when the effects are measured in
the short run (in this case a two year period).
It also shows that firms with higher productivity levels tend to innovate more than the
average. This finding shows, following the work of Conceição et al. (2003), that the
2
adjustment cost/liquidity constrains explanation of a negative relationship between innovation
and productivity growth has a larger role than the other two theoretical explanations in
explaining this correlation between innovation and productivity in the short run.
The results described were obtained using a model that tested the relationship between the
two referred variables. In this kind of model, some controlling variables are included in order
to identify the real relation between productivity and innovation. A novelty in this study was
the inclusion of a new independent variable, gross investments in tangible goods, which can
be interpreted as giving an indication of the firm's strategy. Although the literature on
productivity states that management and strategy influence the level and dynamics of
productivity, these aspects of firm behavior are exceptionally difficult factors to quantify and
to measure. The inclusion of this variable can bring some new light to the comprehension of
the productivity/innovation relationship, since investment in tangible goods is expected to
increase the level of productivity because this kind of investments are done in order to benefit
the production condition of the firm.
To sum up, the objective of the present thesis is to contribute to the empirical analysis of the
relationship between innovation and productivity using recent data (from the period 19982000) that measures generic technological innovation directly at the firm level. In other
words, we aim at answering the following research question: What is the short run
relationship between productivity and innovation, in the context of CIS III Portuguese
responding firms?
From the results of this thesis some policies implications can be drawn. The most relevant is
that the evaluation of innovation efficiency and impact cannot be done immediately:
technology adoption is a complex process that is not instantaneous. In order to understand the
real costs of adoption, it has to be considered that the adopter will take time to fully take
advantage of the technology. Therefore, when evaluating a new technology, decision makers
at the firm and state level have to consider this time lag between adoption and productivity
impact: a technology that is inefficient in the short run can raise productivity in the long run.
The remaining of this thesis is organized as follows. In Chapter 2 we describe the results from
recent research in the area of productivity (productivity definition and productivity
determinants) with a special focus on the relationship between this variable and innovation.
Also in Chapter 2 we review three theories that try to explain the relationship between
innovation and productivity in the short run: learning, technology and organizational rigidities
and adjustment costs. Chapter 3 discusses the CIS data characteristics, reviewing work done
by other authors using this database. Chapter 4 describes the model and some of the
3
econometric methods used. Chapter 5 describes the data, reports our main findings, and
concludes.
4
Chapter 2 – Literature Review
As described in the introductory chapter, the main objective of this thesis is to study the short
run relationship between productivity and innovation. In order to do so, we will build an
econometric model that must be founded in theoretical background and whose results must be
interpreted in light of previous work done by other researchers.
In this context, it is essential to build a literature review that answers these to demands. This
chapter was divided in two independent sections. The first one, broader, contextualizes this
thesis in the recent research on productivity and innovation, giving theoretical background to
the applied model. The second fraction, narrower, reviews the literature that tries to explain
the negative relationship between productivity and innovation and is essential to explain the
results obtained in this thesis.
2.1 - Productivity and Innovation: What we Know, What we Need to Know:
A Literature Review
This section aims at doing a survey of the literature that permits building a theoretical
framework to this thesis. The objective is to contextualize this thesis in face of the current
research trend that study productivity and in particular the relationship between this variable
and innovation.
2.1.1 - Productivity Determinants
2.1.1.1 - Productivity Definition and Macro Studies
Even though the aim of this work is to study the relationship between productivity and
innovation, it is essential to consider other variables that influence productivity to construct a
model that correctly captures the effect of innovation on productivity. The objective of this
section is to try to identify the main variables that may influence a firm’s productivity.
5
Economic growth is measured by the increase in the quantity of goods and services produced
by a country in a given period, and, consequently the economic growth will come from two
sources: increases in the factors of production (labor and capital), and / or efficiency gains
(Steindel and Stiroh 2001).
Productivity growth is clearly an essential measure of economic health and all of the major
measures of aggregate labor and total factor productivity have been good indicators of
economic performance of countries and firms. So, the concept of productivity is closely
linked to economic growth, since it can be defined as the ratio of output to the factors of
production - real output per hour of work or real output per unit of all inputs - total factor
productivity (TFP). It measures the improvement in the efficiency with which a business,
industry or country produces goods and services. Nevertheless, the concept of efficiency or
productivity is difficult to quantify when compared with other economic indicators, as
unemployment, because no direct method exists for doing so. For example, the TFP residual
captures changes in the amount of output that can be produced by a given quantity of inputs,
so, it measures the shift in the production function.
Following this idea, Rogers (1998) introduced various methods that have been used to analyze
productivity (defined by the author as the ratio of output to input for a specific production
situation). Productivity measures can be made at the process, plant, firm, industry or economy
level; each of which involves some specific issues, concepts and variables. For this author, all
of these concepts of productivity make this variable harder to analyze and comprehend. In this
context, he stated that productivity changes can be caused by movements in the “best
practice” production technology, or by changes in the level of efficiency. He also discussed
the problems found in measuring productivity when there are multiple outputs and inputs and
the problems concerning the measurement of inputs and outputs.
Islam (1999) compared methodologies from several different approaches (time-series, panel
data and cross-section) to make an international comparison between TFP estimates. To
conduct this comparison, the author used two samples: the G7 countries and a large sample
that includes developing nations. The results obtained showed that the different
methodologies produce, in some aspects, similarities but in other aspects some dissimilarities.
The author also showed how these different approaches to international TFP comparison can
play complementary role in enhancing our understanding of such important phenomena as
technological diffusion and TFP-convergence.
The authors cited above defend that productivity, despite being a simple idea; it has to be
analyzed very carefully because it is a concept that, in order to be correctly examined has to
6
be very well contextualized. Illustrating this point, Singh and Trieu (1996) explained total
factor productivity growth (TFPG) residuals for Japan, South Korea and Taiwan, by
expenditures on basic, applied and experimental research. They found that there is evidence
that R&D expenditures in these countries had a positive impact on TFPG and so growth in
these countries is not simply explained by input accumulation.
Galarneau and Maynard (1995) defined productivity growth at the national level as the
difference between the increase in the quantity of goods and services produced by all
businesses and the additional quantities of all inputs used, representing, in the long term, the
improvement in the efficiency of the entire production process. In this context, they concluded
that increased productivity is an important element of economic growth because, without it,
output would increase only with the addition of larger quantities of the factors of production.
In order to understand how labor productivity evolved at the national level in the USA,
Steindel and Stiroh (2001) identified the factors that are believed to determine this variable:
long-run growth, living standards, and inflation. They began this analysis by using the
traditional sources of growth analysis that decomposes labor productivity growth into three
primary components – capital deepening, labor quality, and total factor productivity, to
conclude that, after examination of the sources of productivity growth, the major source of the
better aggregate performance has been the remarkable surge of the high-technology sector.
Beyer and Vergara (2002) argued that the economic performance is not only limited by the
level of resources and technology but also by the structure of incentives embodied in
institutions and economic policies. They analyzed Chile’s economic success in the last years
and linked it to the application of sensible economic policies and the existence of a sound
institutional environment. So, Beyer and Vergara (2002) concluded that, if the country is able
to keep and improve these policies and institutions, an additional period of high growth may
be assured. A country like Chile has to invest in the improvement of its educational system to
master its economic growth. Since Chile is a country with an economic profile similar to
Portugal, the work of Beyer and Vergara (2002) can be a good starting to a country level
analysis of the Portuguese economy.
Nicoletti and Scarpetta (2003) addressed the differences in multifactor productivity (MFP)
growth outcomes exploring the links between productivity performance and privatization and
regulatory reform policies. In this analysis, they did not attempt to measure directly the degree
of entrepreneurial incentives and competition at the industry or nation-wide levels but they
focused on policy determinants. In order to do so, they described policy patterns by means of
an original set of indicators of product market regulation and public ownership, checking
7
whether convergence in policies has occurred in recent years. Within this framework, they
examined the impact of privatization and liberalizations policies on growth focusing on MFP,
which is an important determinant of GDP growth, and concluded that it accounts for a high
share of its cross-country variance. Then they estimated the effects of these policies on the
level of MFP and the speed of its convergence to best practice. In this work Nicoletti and
Scarpetta (2003) explored some policies issues and the effects that government policy can
have on productivity indicators, concluding that these policies affect directly these indicators
at the national level. At firm level a similar statement can be made: government and firm
policies and strategy affect directly the firm productivity.
The studies at national level and international comparisons are a complement to the study
developed in the present thesis since they can give an international framework to firm level
studies and can fill some of their gaps.
2.1.1.2 - Firm Level Studies
At firm level, Dedrick et al. (2001) considered that there are three ways of increasing labor
productivity: by increasing the level of capital applied per unit of labor (capital deepening);
improving the quality of inputs, and labor in particular, as a result of education and training;
and contributing to the MFP growth, which is the residue of growth that cannot be accounted
for by the first two factors. An MFP increase means that production methods have improved
or the quality of products have improved, allowing higher output levels to be achieved from
the same inputs.
Also with the objective of identifying the productivity determinants, but to the Swedish
context, Josephson and Schön (2002) analyzed TFP behavior in manufacturing in a long term
perspective 1950-1994. They aimed at two objectives: to develop a framework of explanatory
variables of TFP-behavior and of structural change in industry and to relate the analysis of
productivity development to different explanations of Swedish economic behavior since the
1970s.
Using a sample of 21 OECD-countries and analyzing a different context (the academic field
and not the firm productivity measurement), Kocher et al. (2004) measured productivity in
top-edge economic research institutions by using data envelopment analysis (DEA), with the
following input variables: R&D expenditures, number of universities with economics
departments and (as uncontrolled variable) total population. DEA was defined by these
8
authors as a tool for evaluating relative efficiency and is widely used when there are multiple
inputs and outputs and one lacks a specific functional form of a production function.
After these overall considerations about the productivity evolution and before identifying the
variables that justify the differences between different levels of productivity and productivity
growth, we have to stress that there is a wide diversity of productivity performance of firms.
Even firms with a very similar profile (industrial sector, geographic location, products and
services commercialized, processes utilized to produce those goods and services) have
notably different productivity levels, as well as growth rates. As noted by Syverson (2004),
recent empirical work emphasizes the enormous magnitude of plant-level productivity
variation and the fact that much of the variation cannot be explained by differences between
industries.
Entering and exiting of firms also influence the aggregate level of productivity, nevertheless
the productivity dynamics of new entrants does not have a relevant impact in the long run
trend of productivity because only the productive entrants reach maturity (Ahn 2001). In other
words, creation and closing of firms is an indicator of the dynamics of the economy, a
characteristic that have to be considered when studying productivity.
Another characteristic of productivity that is highlighted by the literature is the non-existence
of a direct link between productivity and downsizing (Conceição et al. 2003)
Bartelsman and Doms (2000) reviewed the research that used longitudinal microdata (in
particular the Longitudinal Research Database which is a large panel data set of U.S.
manufacturing plants developed by the U.S. Census Bureau) to study productivity flows and
to inspect the factors that influence productivity growth. They found that these works explore
a wide range of subjects: the dispersion of productivity across firms and establishments, the
persistence of productivity differentials, the consequences of entry and exit, and the
contribution of resource reallocation across firms to aggregate productivity growth. In
addition, they found that recent research on productivity begun to address the more difficult
questions. These studies have tested the importance of factors such as managerial ability,
technology use, human capital, and regulation towards productivity growth.
Bartelsman and Doms (2000) summarized conclusions drawn by different studies in order to
construct a framework of the productivity research. The first feature to be highlighted is the
extremely high rate of dispersion of the values of productivity. So as stated by Womark et al.
(1989), there are firms with much higher values of productivity than others and these highly
productive firms today are more than likely to be highly productive firms tomorrow, even
though there is a reasonable amount of change in the productivity distribution. In third place
9
and associated to the fact that the manufacturing sector is typified by large shifts in
employment and output across establishments every year, a large portion of aggregate
productivity growth is attributable to resource reallocation; a fact that is coherent with the
concept of creative destruction. Concerning the effect of technology on productivity, it is now
recognized that looking at the correlation between a factor of production, such as computers,
and productivity is not enough to comprehend causal mechanisms because the technology use
is highly correlated to other variables, like human capital and managerial structure. Finally,
another aspect highlighted by these studies, is the difficulty to study the effect of some
variables in productivity, as the case of regulatory environment. Nonetheless it is clear that
any regulations that inhibit resource reallocation can have negative effects on productivity
growth (Bartlesman and Doms 2000).
Tybout (2000) developed a study similar to the Bartlesman and Doms (2000) but with a
special focus on developing countries. He reviewed the literature that centered the analysis on
how conditions, such as protected markets and imperfect competition, affect productivity
dispersion and productivity growth in developing countries.
Stressing the important role of machinery and capital investment as a source of economic
growth, De Long and Summers (1991) presented quantitative results that support the view
that the accumulation of machinery, more than the other components of investment, is a major
determinant of productivity growth. Jorgenson (1988) found considerable complementarity
between equipment investment and total factor productivity. More recently, Caselli and
Wilson (2003) accounted for large part of the observed differences of TFP across countries
when finding the composition of capital investment.
In this context, after the conduction of a more specific literature review in order to understand
which the variables that influence productivity differences across firms are and considering
the limited information available on the CIS III, we identified the following important
variables that affect productivity:
1) youth of the firm, since the entry of new actors influences the level and the rate of change
of productivity (new firms have greater productivity growth rates but lower absolute levels),
in other words, since new firms are not completely focused in production activities, the level
of productivity is expected to be lower when compared to established firms (Haltiwanger
2000);
2) the average human capital of the firm, given that a more skillful workforce is expected to
produce higher productivity levels and higher growth rates because is often associated to
10
acquisition of more sophisticated technologies and to new competencies (Acemoglu and
Zilibotti 2001);
3) innovation, despite the fact that the direction of the relationship is not clear and that it has
to be joint determined with productivity (Crépon et al. 1998) – this variable will analyzed
more deeply in the next section;
4) exports, because a high level of international exposition of the firms is expected to be
closely linked to high degree and growth rates of productivity. Only firms that have high rates
of productivity can compete in the international market (Bernard et al. 2000) and;
5) management and strategy of the firm, since the strategy of the firm have influence in the
level and dynamics of productivity (De Long and Summers 1991; Jorgenson 1988; Kuznetsov
and Muraviev 2001; Caselli and Wilson 2003);
6) being part of a group, because it can create positive externalities, like cooperation, that
contribute to productivity growth.
This choice followed closely the work of Conceição et al. (2003), since the present thesis used
a similar methodology and a similar database.
The main objective of this section was to make a reflection about the different factors that
may influence productivity besides innovation. In order to complement this objective, the
concept of productivity was characterized, some works that discussed productivity were
presented and the importance of productivity measurement was discussed.
2.1.2 - What is the relationship between innovation and productivity?
2.1.2.1 - Innovation
In relation to other crucial variable in this study, innovation, academic work has demonstrated
that innovation propensity differs greatly across firms and that it is influenced by a large
number of factors, internal and external to the enterprise. Despite the fact that similar firms
could have very dissimilar innovating profiles, an analysis of the characteristics that are
responsible for the innovative character of a firm could be done. This categorization can be
done using typical firm characteristics and classifications in order to verify if exists a
correlation between these features and differences in innovation capability. This analysis
11
could be performed considering two major axes, dimension and sector affiliation, but other
enterprise characteristics cannot be ignored to fully comprehend the innovation process: being
part of a group, year of entrance in the market, location, preferential market, product life cycle
characteristics, the occurrence of a merger or sale of the enterprise or partial closure, and
education level of the personnel, only to refer a short list (Archibugi et al. 2000; Sirilli 2002;
Bóia 2003).
In this context, Criscuolo and Haskel (2003) tried to identify the link between innovation and
productivity growth and the factors that are linked with high innovation. They found two
main empirical approaches to study the link between innovation and productivity: the
correlation between inputs to the innovation process (R&D) to productivity growth; and the
relation between productivity growth and a measure of output of the innovation process,
patents. Both these approaches have problems since they are not completely representative of
all innovation activities (R&D is an input to the innovation process and not an output and not
all innovations are patented specially in services and traditional sectors). The impact of
intellectual property protection depends on the stage of technology development and on the
cultural framework of country / region. For example, in Europe exists an inability to
commercialize and patent the results of research when comparing with the USA, mostly
because of regulatory and cultural factors (Conceição et al. 1998). In the Portuguese context,
the very low values of patents per million of people (2.9 patents request in the EPO against
the EU average of 107.7) indicate that the study of innovation using patents data is inadequate
because, presumably, most of innovation is not patented.
Within this framework, Criscuolo and Haskel (2003) described the effort done by the OECD
to overcome these methodological problems. This institution has developed guidelines for
company surveys that measure innovations directly. The most important of these surveys is
the Community Innovation Survey (CIS), in which it is set out a definition of innovation and
where it is asked companies to report the output of the innovation process (introduction of
innovative products, new processes, percentage of sales arising from new and improved
products and “soft” innovations, such as organizational change), the inputs to innovations
(R&D, scientists, sources of knowledge) and the obstacles to innovation (finance, bad luck,
etc.). The Oslo manual OECD (1992, 1996) codifies such survey models and the CIS applied
the concepts by enquiring enterprises in the UE countries.
Using this data, Criscuolo and Haskel (2003) achieved some important findings: detection of
a statistically robust association between (process) innovations and TFP growth; confirmation
that process innovations are more likely in firms who devote resources internally to the
12
innovation process; evidence that external sources of information flows affect innovations;
and exposing that process innovations are more probable in firms co-operating with other
firms and using patent disclosure.
2.1.2.2 - Innovation vs. Productivity
After this brief introduction about innovation determinants, we will review the literature that
analyzes the relationship between the two most important variables considered in this study:
innovation and productivity.
Licandro et al. (2004) analyzed the role of replacement and innovation activity in shaping
investment behavior and labor productivity in a panel of Spanish manufacturing firms from
1990 to 2001. They found evidence of the role of replacement activity as a determinant of
investment for firms that are not involved in process innovation or in plant expansion. In
addition, they explored how large investment events affect the evolution of labor productivity
under different innovative strategies and found that innovative firms increase their
productivity after an investment.
In other words, Licandro et al. (2004) explored the occurrence and implications of different
types of investment: expansionary investment and replacement investment. Expansionary
investment could have positive effects on labor productivity at the firm level if new
equipment is more productive than the existing one. On the other hand and despite the fact
that replacement does not necessarily mean the adoption of better technologies and long
learning curves might be associated with new technologies, replacement investment could
imply productivity growth when the new equipment is more productive than the old
equipment. Doing this analysis, these authors argued that a distinction between these two
kinds of investment is helpful to identify the nature of the observed episodic behavior of
firm's investment.
After clarifying this distinction, they checked the existence of productivity effects associated
with investment episodes based on the premise that large investment episodes should raise
productivity. Licandro et al. (2004) stressed that different innovative strategies should have
different effects on productivity and so they were concerned with the relationship between
innovation activity and investment activity and its effect on labor productivity.
13
In the same context, Power (1998) emphasized that there is limited empirical evidence on the
relation between investment and productivity. This author focused on investment episodes
and examined the relationship between investment and labor productivity at the plant-level.
With the same objective, Sakellaris (2001) conducted a study using manufacturing data to
describe the patterns of employment and capital adjustment and the response of total factor
productivity during those adjustment events.
Parisi et al. (2002) presented empirical evidence on the effect of process and product
innovations on productivity and on the role played by R&D and fixed capital investment in
enhancing the introduction of innovations at the firm level. This exploitation was realized
using a firm-level database. They used panel data that permitted the characterization of the
pattern of innovation activity for a large sample of Italian firms and the relationship between
this variable and productivity. They found two interesting characteristics of this relation: the
introduction of process innovation has a sizeable effect on productivity and that the
productivity effect of a process innovation is larger than the one of a product innovation.
Their results imply that process innovation has a large impact on productivity and that R&D
spending is strongly positively associated with product innovation, whereas fixed capital
spending increases the likelihood of introducing an innovation of process. These results
reflect that new technologies are commonly embodied in new capital goods but that, in the
other hand, the effect of fixed investment on the probability of introducing a process
innovation is overstated by R&D spending internal to the firm. R&D affects productivity
growth by easing the absorption of new technologies. They also stated that R&D spending is
highly positively associated with the probability of introducing a new product, but not with
the probability of introducing a new process (associated to spending on new fixed capital). In
conclusion, Parisi et al. (2002) have provided evidence that, at the firm level, the ability to
introduce new technologies generated outside the firm increases with internal R&D spending.
So, these results emphasize the theories that defend the role of R&D in increasing “absorptive
capacity” as crucial.
Vivero (2002) investigated the effect that a measure of the process innovation performance of
a firm has on its labor productivity growth. The analysis is based on two reflections: the
differentiation between the effect of product and process innovations have on firm’s
performance and the assumption that the knowledge capital of a firm is mainly composed by
its successful research.
14
In this study, Vivero (2002) demonstrated that process innovation has a positive and
significant effect on firm’s productivity growth under a wide range of alternative
specifications.
Comín (2004) measured the contribution of R&D investments to productivity growth, taking
in consideration the relationship between the resources devoted to R&D and the growth rate
of technology. He showed that the resulting contribution of R&D to productivity growth in
the US is smaller than three to five tenths of one percentage point and defended that, if the
innovation technology takes the form assumed in the literature, the actual US R&D intensity
may be the socially optimal. In conclusion, he showed that R&D is not responsible for a large
share of productivity growth (defined as the learning to use factors more efficiently) in the
US.
To complement this analysis, Comín (2004) highlighted two aspects of the relationship
between R&D activity and productivity: R&D affects labor productivity through the
development of new final, intermediate or capital goods and the existence of intentional nonR&D innovations that lead to improvements in productivity without the need to adopt any
new capital or intermediate good.
Doms et al. (1997) documented how plant-level wages, occupational mix, workforce
education, and productivity differ with the adoption of new factory automation technologies
such as programmable controllers, computer-automated design, and numerically controlled
machines. Their cross-sectional results showed that firms that use a large number of new
technologies employed more educated workers, employed relatively more managers,
professionals, and precision-craft workers, and paid higher wages. Nevertheless, this
longitudinal analysis showed little correlation between skill upgrading and the adoption of
new technologies. It appears that plants that adopt new factory automation technologies have
more skilled workforces both pre- and post- adoption.
Based on previous studies, Crépon et al. (1998) proposed a novel empirical approach to the
problem of assessing both the innovation impacts of research and the productivity impacts of
innovation and research. Based on new sources of data available for French manufacturing
firms, they built a model that summarized the process that goes from the firm decision to
engage in research activities to the use of innovations in its production activities.
This approach followed three steps:
1) accounting for the fact that it is not innovation input (R&D) but innovation output that
increases productivity;
15
2) using new data on innovation output in French manufacturing, in addition to the more
common information on the firm current accounts and R&D expenditures;
3) the estimation of the model using econometric methods that avoid the possible
selectivity biases usual on innovation studies.
These econometric problems associated to innovation studies exist when the focus of analysis
is on patenting firms and on firms that engaged R&D activities, which are a minority. In this
context, there is the concern of the endogeneity between innovative input and output that
implies an issue of simultaneity in the model (R&D is endogenous in the innovation equation
and patents or innovative sales are endogenous in the productivity equation). In addition, the
disturbances in the equations of Crépon et al. (1998) model, which reflects, in part,
unobserved variables and firms effects, are probably correlated.
2.1.2.3 - ICT and Productivity
Recent productivity and innovation relationship research have focused on the effects of
adoption and diffusion of ICT on productivity. Thus we will devote a part of this literature
review to the relation between these recent innovations and productivity.
At the country level, two of the most used factors that have been considered to affect the
relation between productivity and innovation are institutional environment and product and
labor market policies. These factors may have played a particularly important role in the past
decade when growth was associated with the spread of ICT, and the latter has arguably been
boosted by the entry of new, innovative firms in most markets as well as by technology
adoption by incumbent firms.
The relationship between information and communication technology and productivity has
been debated in the last decades but early empirical research usually did not found significant
productivity improvements associated with ICT investment (Bender 1986; Loveman 1994;
Roach 1989; Strassmann 1990). Various reasons were recommended to enlighten this
paradox: the problems of using simple bivariate correlations between aggregate productivity
and ICT indicators (Lehr and Licthenberg 1999), the potential negative effect of augmented
variety on productivity (Barua et al. 1995), the deferred effect of ICT investment on
productivity gains and its dependence on network externalities and on changes in the
complementary infrastructure (David 1990).
16
In recent years, with new data and new methodologies, empirical investigations have found
evidence that ICT is associated with improvements on productivity, in intermediate measures
and in economic growth (Oliner and Sichel 1994; Brynjolfsson and Hitt 1996; Sichel 1997;
Lehr and Licthemberg 1999; Paganetto et al. 2003).
In this context, many studies have examined the contemporaneous effects of investment on
output and productivity growth but not many scholars have investigated the effects of
investment in new capital on productivity growth over a long period of time. This issue is
relevant since investment raises the stock of capital and output, but adjustment or adoption
costs may initially obscure these gains. If the analysis of the effect of innovation on
productivity is done in the short run may lead to erroneous conclusions about the real
effectiveness of a particular innovation since most of the technologies do not have immediate
favorable effects on firms.
Timmer et al. (2003) analyzed the contributions of IT (Information Technology), capital
deepening and total factor productivity growth (TFP) in IT-production on aggregate labor
productivity growth patterns within the European Union. Doing a comparison with the USA,
these authors tried to contribute to the analysis of the impact of IT on growth paths of
European countries and the US through two channels: IT investment and the production of IT
goods. They found that differences in the direct effects of IT almost completely explain the
US lead in labor productivity growth over the EU aggregate over the period 1995-2001.
Zagler (2002) tried to distinguish stylized facts of the new economy, focusing on the
productivity debate and on the recent evidence suggesting that total factor productivity gains
were low in the early phase of the new economy, but increased in more recent years. He found
that productivity is created, not within each firm, but through the mixture of varied products
provided by several firms. This fact makes it hard to capture productivity gains in the
provision products of the New Economy.
Zagler (2002) found that a simple model of production, which includes old technology and a
new technology in a single production function, as good fit to reality. With this model, Zagler
(2002) found that, if information, communication and biotechnologies (ICBT) innovations
increase productivity only through influencing other innovations, it is expected that the
productivity gains may be only visible in the capital deepening term of a growth accounting
equation. In this context, Zagler (2002) found that inflation rates are biased downwards for
two reasons: the declining share of the old technology in production and the fact that actual
prices of ICBT applications have to be adjusted downward with an increase in the number of
ICBT applications.
17
Regardless of the fact that all firms have access to the new technology, not all firms are
adopters, since, in some contexts, there are political restrictions to use the Internet, technical
limits for mobile phones, or legal restrictions to experiment with genetic modifications. This
fact has as consequence that firms located in particular regions will continue to use the old
technology, even if in other parts of the world a shift to new technologies has already taken
place.
As a conclusion, the author discussed several other features of the new economy, such as an
inflation rate below its fundamentals, increased stock market volatility, and a threshold for the
adoption of new technologies that explains the regional divide between countries.
To take advantage of the productive capacity embodied in the new capital, firms must apply
resources to integrate the new technology into their production processes and thus costs
appear (direct, in the form of installation and training costs). In addition, there are more
invisible costs, associated to the development of ways of using the new technology, or costs
associated with implementing organizational change that complement the installation of new
technologies.
Leung (2004) used aggregated annual Canadian data from 1961 to 2001, to investigate the
magnitude of the outcome that investment in new technology (new computer hardware), can
have on output growth. He found that this kind of investment has a positive effect on output
growth that cannot be explained by growth in inputs. Additionally, the author’s findings
suggest that the effect of computer hardware investment has grown over time by presenting
evidence that investment in computer hardware leads to growth in output and productivity
higher than the accumulation of computer capital alone. In addition, Leung (2004) also
stressed that the full impact of computer investment is not entirely realized until three years
after the initial investment.
In this context, Leung (2004) stated that his results do not suggest that computer investment
does not raise output immediately. Instead, he stated that his results imply that computer
investment raises output levels more than the amount usually attributed by traditional growth
accounting methods, but with a temporal lag. The basic neoclassical model of production is
often used to assess the contribution of investment to output growth. In this model, investment
raises the capital stock and output growth increases in proportion to the growth in capital.
However computers when considered as a “general purpose technology” have a more
inclusive outcome; stimulating the process innovations and facilitating organizational change.
Brynjolfsson and Hitt (2000) stated that the link between IT and increased productivity
emerged well before the recent surge in the aggregate productivity statistics. To be aware of
18
the economic value of computers, one must widen the traditional definition of both the
technology and its effects. To bring some light to this issue, these authors defend that the
realization of econometric analysis and of case studies is an important instrument. Results
already obtained suggest that organizational “investments” have a large influence on the value
of IT investments; and that the benefits of IT investment are often intangible and
disproportionately difficult to measure.
Brynjolfsson and Hitt (2000), based on the analysis of firm-level studies (firm-level analysis
has significant measurement advantages for examining intangible organizational investments
and product and service innovation associated with computers), reviewed the evidence on
how investments in IT are connected to higher productivity and organizational issues and
other measures of economic performance.
They sustained the proposition with two arguments supported by the existing literature: the
important component of the value of IT that is related to the capacity of computers to enable
complementary organizational investments and the productivity increase that is a result from
the reducing costs associated to IT investments. In addition, these investments also enable
firms to increase output quality in the form of new products or in improvements in existing
products.
Brynjolfsson and Hitt (2000) also argued that these factors are not well captured by traditional
macroeconomic measurement approaches, and so, the economic aspect of computers is
expected to be understated or mistaken by aggregate level analyses.
Nevertheless, empirical evidence has shown that, in last decades, IT has created substantial
value for firms that have invested in it. The explanation to the existent difference between
theory and practice is that traditional growth accounting methods are centered on the
observable aspects of investment, such as the price and quantity of computer hardware in the
economy. Intangible investments in developing complementary new products, services,
markets, business processes, and worker skills are forgotten. However, standard growth
accounting techniques provide a useful benchmark for the contribution of IT to economic
growth.
To illustrate this opinion, Brynjolfsson and Hitt (2000) stated that the firm level studies
suggest that computers have had an impact on economic growth that is disproportionately
large compared to their share of capital stock or investment. This impact is likely to grow
further in coming years, not confirming the paradoxically unproductive image reflected in the
fact that computer technology is largely uncounted in national accounts. Case studies and
econometric work identify new business processes, new skills and new organizational and
19
industry structures as a major driver of the contribution of IT. These complementary
investments, and the resulting assets, compensate the investments in the computer technology
itself. So, the use of firm-level data has brought a new light on the black box of production in
the increasingly IT-based economy and, in consequence, contributes to a better
comprehension of the key inputs and of the key outputs (Brynjolfsson and Hitt 2000).
In a later study, Brynjolfsson and Hitt (2003) also looked at the effect of computerization on
productivity and output growth using data from 527 large US firms over 1987-1994. They
found that computerization can contribute to measure productivity and output growth in the
short term coherently with normal returns to computer investments. They also stated that, in
the long run, the productivity and output contributions associated with computerization are up
to five times greater that in the short run. These results imply that the observed contribution of
computerization is accompanied by relatively large and time-consuming investments in
complementary inputs, such as organizational capital, that are often omitted in conventional
calculations of productivity (a problem that is the focus of the present study).
In addition, Brynjolfsson and Hitt (2003) presented evidence that computers contribute to
productivity and output growth as conventionally measured in a broad cross-section of large
firms. Computers are part of the system of technological and organizational change that
increases firm-level productivity over time, a fact that is consistent with the conception of
computers as a general-purpose technology.
These authors stated that computers contribute to output growth but not to productivity
growth in the short run but that, in the long run, computerization is associated with an output
contribution that is substantially greater than the factor share of computers alone and thus
contributes to long-run productivity growth. In conclusion, the impact of IT investment on
labor productivity is significant and positive, despite the fact that the productivity impacts of
IT investments vary broadly among different firms.
Brynjolfsson and Hitt (1995) estimated that “firm effects” may account for as much as half of
the productivity benefits attributed to IT investment: market position, rigidities in cost
structures (e.g., labor contracts), brand recognition, the vision and leadership abilities of key
executives, organizational structure, strategy, and management practices that can be compared
systematically across companies.
Van Leeuwen and Van der Wiel (2003) also questioned the effect of investments in ICT in the
growth of productivity by presenting evidence on the relation between ICT use, innovation
and productivity based on firm-level panel data. Their results indicated that the potential of
many firms to catch up their ICT investment may increase labor productivity growth and that
20
ICT can contribute to labor productivity growth directly through capital deepening and
indirectly by enhancing innovation. More precisely, Van Leeuwen and Van der Wiel (2003)
presented further evidence on the role of ICT at the firm level for the Netherlands, particularly
in market services. They exploited an extensive balanced panel of firm-level data to
investigate the contribution of ICT to productivity growth and to analyze the link between
ICT and innovation. ICT and innovation appear to be closely related.
They presented detailed evidence on the direct and indirect productivity impacts of ICT by
comparing the performance of firms that showed diverse levels of innovation strategies and
ICT use. They used a labor database to the two waves of the innovation survey (CIS II,
covering 1994–1996, and CIS II.5, covering 1996–1998) to determine which firms were
innovative in the period considered.
Using this firm-level data, their work showed that ICT has the potential to remain an
important source of productivity growth since the estimates obtained pointed to a sizable
direct contribution of ICT to labor productivity growth. Van Leeuwen and Van der Wiel
(2003) found support for the assumption that ICT enhances the innovation performance of
firms, thereby contributing to labor productivity growth in a more indirect way.
Inklaar et al. (2003) used an industry-level database to analyze sources of growth in four
major European countries: France, Germany, Netherlands and United Kingdom (EU-4), in
comparison with the United States for the period 1979-2000. They decomposed the aggregate
labor productivity growth into industry-level contributions of labor quality, ICT and non-ICT
capital deepening and TFP. The conclusions were that a small group of service industries is
mainly responsible for the acceleration in ICT capital deepening, but their contribution to
growth is lower in the EU-4 than in the U.S. (TFP in these industries accelerated in the U.S in
the 1990s, but not in Europe). Additionally, widespread deceleration in non-ICT capital
deepening in the EU-4 has led to a European productivity deceleration.
They defended that the explanation to the acceleration in US labor productivity growth and
the lack of it in the EU countries is the difference in performance of industries that use with
intensity ICT and those that do not. In addition, they stated that, even though the differences
in ICT investment are quite important to explain the aggregate labor productivity growth
differential between the US and these European countries, TFP growth also has a substantial
role to play.
Paganetto et al. (2003) analyzed the determinants of ICT investment and the impact of
information technology on productivity and efficiency on a representative sample of small
and medium sized Italian firms. In order to test the most relevant theoretical predictions from
21
the ICT literature they estimated the impact of investment in software, hardware and
telecommunications of these firms on a series of intermediate variables and on productivity.
As controlling variables, they considered the demand for skilled workers, the introduction of
new products and processes and the rate of capacity utilization. As productivity measures they
used total factor productivity and the productivity of labor.
With their model, they showed that the effect of ICT investment on firm efficiency can be
more evidently perceived at firm level data by decomposing it into software and
telecommunications investment. This differentiation was done since these two kinds of
investment have different characteristics. Software investment increases the demand for
skilled workers, average labor productivity and proximity to the optimal production frontier
and telecommunications investment positively affects the creation of new products and
processes but negatively affects average labor productivity.
Crafts (2002) used a growth accounting methodology to compare the contributions to growth
in terms of capital-deepening and total factor productivity growth of three general-purpose
technologies: steam in Britain during 1780-1860, electricity and information and
communications technology in the US during 1899-1929 and 1974-2000, respectively.
With this analysis the author compared these technological eras with the results for ICT
diffusion obtained by Oliner and Sichel (2000) and tried to fill the gap in research that
compares the recent experience of ICT with other historical technological breakthroughs
(adoption of general purpose technologies). The results obtained by Crafts (2002) with the
gather of the available evidence into formats which readily permit comparison between these
episodes of technological change, suggested that the growth contribution of ICT in the past 25
years has surpassed that of steam and at least matched that of electricity over comparable
periods.
Recent estimates of the contribution of ICT to US economic growth have employed growth
accounting methodologies in which the new technology potentially has impacts through use
of new capital goods, TFP growth in making the new capital goods and TFP spillovers. Crafts
(2002) has used a similar approach to generate estimates of the impacts of earlier general
purpose technologies, electricity and steam, that can be compared with those of ICT.
Using pooled cross-section, time-series data for 44 industries over the decades of the 1960s,
1970s, and 1980s in the US, Wolff (2002) did not found econometric evidence that computer
investment is positively linked to TFP growth. Nevertheless, the author defended that
computerization is positively associated with occupational restructuring and changes in the
composition of intermediate inputs and capital coefficients.
22
Wolff (2002) concentrated his analysis on the relation of skills, education, and
computerization to productivity growth and other indicators of technological change on the
industry level. He found no evidence that the growth of educational attainment has any
statistically measured effect on industry productivity growth. The growth in cognitive skills,
on the other hand, is significantly related to industry productivity growth, though the effect is
very modest and the degree of computerization is not significant. In contrast, computerization
has had a statistically significant effect on changes in industry input coefficients.
Gera et al. (1999) examined the effect of information technology investments on labor
productivity in Canada and the US, using OECD databases. These authors reached the
following results: IT investments are an important source of labor productivity growth across
Canadian industries; in terms of the impact of international R&D spillovers on productivity
growth, those emanating from IT imports are much more important than those from non- IT
imports; and IT investments and international R&D spillovers embodied in IT imports have
positive and significant impact on labor productivity growth across US industries.
2.1.2.4 - Final Remarks on the Literature Review
From this review of literature on the relationship between productivity and innovation some
ideas can be highlighted.
The first one is that the conclusions of the numerous analyses done at the firm level are far
from being unanimous. In other words, some authors question the effect of innovation on
productivity. This situation is a consequence of the fact that the intensity of the effect of
innovation (an in particular of ICT investment) on productivity rates vary in a considerable
way, depending on the context of the study, the time period considered and the indicators
chosen to measure the variables innovation and productivity. In this context, a considerable
amount of studies compare the effect of a particular innovation in the US and UE economies.
The conclusions of these analyses confirm the importance of contextual factors on the
relationship between productivity and innovation: the US can take a bigger advantage from
innovation activity, and that is reflected on the productivity indicators.
The selection of the best productivity measure is raised in several studies. Various authors
inquire the real representativeness of the more frequently used indicators, questioning if this
kind of indicators really reflect the real productivity of firms. This issue gains an augmented
importance in studies like the one realized in this thesis where different sectors are compared
23
and global conclusions are drawn, since different sectors of activities have different
production schemes and so also have different productivity profiles.
At the core of the analysis done on this thesis and as discussed in some of the productivity
literature, it is the time period that innovation takes to influence productivity figures. Different
innovations and different sectors of activity influence the time of absorption of a new product
or process in a firm and so the return of the investment done in innovations has different
periods to become profitable.
Another aspect discussed in the literature is the different effect that process innovations and
product innovations have on productivity. As expected, the results obtained by different
authors confirm the hypothesis that process innovations have a bigger and more direct effect
on productivity.
To sum up, the study of the relationship between productivity and innovation is a very active
research area and in particular the discussion about the real impact of technological
breakthroughs in productivity and the time interval that must be considered to study the real
effect of innovation on productivity. The present thesis, by using a database on firm level
innovation activity in the Portuguese context, try to contribute to this discussion and in
particular to the comprehension of the difference between long and short term effects of
innovation on productivity.
2.2 - Productivity and Innovation: Hypotheses and Observable Implications
While innovation should be expected to enhance productivity and, in consequence, to
maximize profits (from the firm’s point of view), there are some theoretical arguments that
suggest that, in the short run, innovating activities can lead to productivity losses. Since this
thesis aims at studying the relationship between innovation and productivity growth in a two
year period data, this thesis should be included in this literature trend. So, and after the
broader literature review done in the previous sub-section, it is essential to review these
theoretical arguments.
We highlight three important strands of theoretical arguments that explain the negative
relationship between innovation and productivity. While all three strands are somewhat
related to each other, it is important to discuss their theoretical frameworks to understand
what sort of empirical results ratify each theory. This discussion follows Conceição et al.
(2003).
24
The first theory focuses on the learning lags that result from introducing a new technology.
The second is centered in the fact that new technologies, in the first steps of the diffusion
process in the economy, are not as perfected and developed as older ones. The last is
concerned with the adjustments costs associated with the introduction of innovation.
2.2.1 - Learning
Firms have to modify their production schemes to fully work with the new technology. The
first theory suggests that when firms do so, they need new skills and thus will be less
productive than if they stayed with the incumbent technology (Jovanovic and Nyarko 1996;
Ahn 1999).
This theory justifies the relationship between productivity and innovation with the
arduousness that very productive firms may have in changing their production scheme to
incorporate new technologies.
As Ahn (1999) refers, the learning process and the full use of a new technology is not an
instantaneous procedure. The adoption of new technology tends to reduce productivity
temporarily, even though potential productivity gains in the long run outweigh the short run
losses. In other words, the adoption of new technology is like investing in physical goods
because it requires short term expenditures that offer long term returns when the technology is
appropriately implemented.
In the process of adoption of a new technology, firms must obtain the necessary skills and
know-how to take the full advantages of the new equipment and to realize its maximum
potential productivity gain. These competencies are acquired by gaining an advanced
understanding of the technology and experience with the application of the technology in a
particular business / industry context. As a result, since no technology can be implemented
instantaneously, it takes time and resources for the potential productivity gain associated to
the new technology to be fully reached. The time and effort spent to adopt the new technology
should be considered as part of the cost for technology adoption, and is called by Ahn (1999)
the “learning cost”.
There are two kinds of costs associated to the adoption of a new technology: tangible costs
(purchasing the technology itself) and intangible costs (learning and adjustment costs
associated with implementing a new technology). To illustrate this distinction, Ahn (1999)
described the example of Compaq Computer Corporation, the world’s largest personal
25
computer manufacturer. This company estimated that the initial acquisition price of a personal
computer represents only about 20% of the total cost of operating that PC in a corporate
network environment and that, in consequence, 80% of the costs are associated to learning or
other intangible costs. Upgrading this technology will imply a short run loss and a long run
gain, because only when the long learning process is completed the firm can take complete
advantage of the innovation.
Ahn (1999) focused his attention on the learning cost that occurs in the process of technology
upgrading, an explanation to the bivalent productivity behavior reported in the information
technology literature. Two main facts illustrate this situation: the little productivity gains
observed in the 1980s when US companies made huge IT investments and the irrelevant Total
Factor Productivity (TFP) growth of the “Newly Industrializing Countries” in East Asia in
spite of the important investment done in the technology sector in these countries.
Jovanovic and Nyarko (1996) developed a one-agent Bayesian model of learning by doing
and technology choice based on the assumption that the more the agent employs a technology,
the better he/she becomes skilled at its features (a form of augmenting the human capital
level), and, as a consequence, the level of productivity is increased.
This model is also based on the fact that all technologies have bounded productivity and, a
firm to keep increasing its performance and productivity growth, has to change its
technological framework. Doing so, the firm reduces technological expertise temporarily,
delaying the expected productivity increase. In an extreme scenario, a high level of the firm
skills for the incumbent technology allied with the perspective of short run losses may result
in a decision not to adopt new technology. In other words, sometimes the less skilled and less
productive actors are the first to perform technological leapfrogs in order to try to be more
competitive (Jovanovic and Nyarko 1996).
The empirical works that follow this theory stress the prediction that innovating firms will
have a drop in their short-run productivity performance. This productivity loss is associated to
the need to engage with non-productive activities in order to gain expertise to work with the
new technology. The theory predicts that innovating firms will exhibit productivity declines
in the short-run, and that the level of productivity is inversely related to innovation. More
precisely, the theory forecasts that the probability of a firm to adopt a new technology is
inversely proportional to the level of productivity of the firm and to the level of radicalness of
the technology. In the limit, the very productive firm, when facing significantly new
technological opportunities, may fail to innovate.
26
2.2.2 - Technology and Organizational Rigidities
The second hypothesis to explain the negative relationship between innovation and
productivity growth in the short run is grounded on the fact that when new technologies
emerge, they perform less effectively than the technologies already diffused. Thus, leading
firms and more productive firms may be reluctant to switch to new technologies that would
imply significant productivity losses (Leonard-Barton 1988, 1992; Young 1991, 1993;
Utterback 1994; Christensen and Bower 1996; Christensen 1997).
Leonard-Barton (1992) examined the nature of the core capabilities of a firm, taking special
focus on the links between these capabilities, and the projects of product and process
development. She analyzed twenty cases of new products and process development projects in
five firms. Leonard-Barton said that managers of new products and processes development
projects face a strategic decision: how to maximize the core capabilities in the environment of
change associated to innovation.
In a previous work, Leonard-Barton (1988) analyzed the relationship between implementation
and innovation. Taking as a starting point the conceptual separation, done by theorists and
practitioners, between the implementation of new technologies and their creation, LeonardBarton (1988) defended that this two concepts are one, and only one, reality. The author
defended that the technology transfer process requires a long term commitment to the process
of change and management techniques aiming at minimizing the inevitable non-perfect
synchronization between the firm characteristics and the innovation. Besides the inherent
characteristics of the technology, the accomplishment of technology transfer depends on the
way that different actors of this process (developers and users) engage in the transfer process
(Leonard-Barton 1988).
Young (1993) built a model to study the key factors behind the firms’ decision to innovate.
This model has two different levels: high or small cost of invention. In the first scenario, the
profitability of the invention is small, the innovation engagement becomes a factor that
constrains growth, and thus the learning parameter is irrelevant (an invention-constrained
equilibrium).
Benner and Tushman (2002) made a contribution to the understanding of how pervasive
practices can affect technological innovation, based on the fact that activities aimed at
refining and stabilizing processes may be in conflict with exploratory innovation required for
adaptation the change process.
27
Tripsas and Gavetti (2000) explored the relationship among capabilities, cognition, and
inertia, by trying to understand the role of managerial cognition in driving the dynamics of
capabilities. They demonstrated, through a case study developed in the Polaroid enterprise,
that search and innovation processes in a new learning environment are intensely related to
the way managers model the new problem and develop strategic and organizational strategies.
Contrasting with the previous theory, this second conjecture suggests that the decision not to
innovate is not associated only to the level of productivity but also to the level of
organizational rigidity (the more productive firms, are those that stick more closely to existing
routines). Then, the theory predictions’ are similar to those from the learning theory and thus
is expected that a negative relationship between innovation and levels of productivity exists.
2.2.3 - Adjustments Costs
If firms fail to innovate due to adjustments costs and liquidity constrains, then the more
productive firms are those that are more capable to deal with these constrains and, in
consequence, unlike the previous two sets of theoretical justifications, the third strand predicts
a positive relationship between levels of productivity and innovation.
Bessen (2002) studied this potential positive relationship between innovation and
productivity. He began by analyzing the problem posed by technology transitions to
productivity measurement, because the new and earlier technologies are never perfect
substitutes of the incumbent technological structure. New technologies may come with
adoption costs associated with learning new skills, implementing new forms of organization,
and developing complementary investments. The incorporation of these investments in the
productivity measurement is sometimes not correct because these expenses are only seen as
costs and not has investments.
Bernstein et al. (1999) extended previous empirical dynamic production models to account for
adjustment costs, and quality improvements associated with all factors of production. With
this analysis they reached results regarding speeds of adjustment, rates of quality change, and
productivity growth rates. In what concerns adjustment speeds, they found that none of the
factors of production analyzed completely adjusted within one year, confirming the
hypothesis that there are adjustment costs associated to innovation.
Hall (2002) stated that adjustment costs determine the dynamics of the response of an
industry’s output to a shift in demand. In order to contextualize his model, Hall (2002)
28
described the effect of adjustment costs in the functioning of an enterprise. If adjustment costs
do not exist, an increase in demand not escorted by any change in factor prices raises output,
labor, capital, and materials in the same proportion. In the presence of adjustment costs, the
elasticity of the response of factors with higher costs is less than one while the elasticity of
those without adjustment costs exceeds one.
In this context, he developed a model of industry dynamics to capture these properties and a
related econometric framework to infer adjustment costs from the observed ratios of factor
responses to output responses. Hall (2002) found fairly precise evidence of reasonable
adjustment costs.
Leung (2004) defended that, since there may be a learning phase before firms grasp the
complete potential of the new technology and begin to employ new processes, there may be a
lag between the growth in investment and its benefits. In reality, during periods of rapid
adoption of new technologies and equipment, firms may incur in adjustment costs and
struggle to maintain previous levels of output. Leung (2004) defined adjustment costs as the
costs related to the direct setting up of new equipment, the training of employees, resources
used to fully utilize the capital, and the reorganization costs. The extent of adjustment costs
found in empirical studies depends on the methods and data used to calculate the estimates.
Unlike the previous two sets of theoretical justifications, this strand predicts a positive
relationship between levels of productivity and innovation. According to this theory, it is
expected that during the introduction of the innovation stage, firms that introduce new
technologies will have a lower rate of productivity growth than those firms that did not try to
innovate (firms that have centered their activities on production). If this context is confirmed
and the reasons not to innovate are adjustment costs and liquidity constrains, it is expected
that more productive firms constitute the group of first innovators since they have more
capabilities to deal with these constraints.
In order to summarize the information exposed in this chapter, Table 1 presents the main
arguments and references of each of the theories referred above:
29
Hyphotesis
Observable
Implications
Arguments
Main References
Innovation - new skills - productivity decrease
negative relationship between innovation and levels of productivity
Time and costs of the adoption process not
neglegetable - learning cost
Jovanovic and Nyarko (1996)
Ahn (1999, 2001)
Innovation implies the execution of non-productivity
activities - drop in productivity in the short run
More productive firms have difficulties to change
technology
When technologies appear perform less effectively
than the technologies already diffused
Technology transfer imply a change on management
techniques in order to synchronize the firm
Leonard-Barton (1988, 1992)
characteristics with the innovation
Utterback (1994)
More productive firms may be reluctant to switch to Christensen and Bower (1996)
Christensen (1997)
new technologies that would imply significant
Young (1991, 1993)
productivity losses
Benner and Tushman (2002)
More productive firms are those that stick more
Tripsas and Gavetti (2000)
closely to existing routines
Decision not to innovate level of productivity and level of organizational rigidity
positive relationship between levels of productivity
and innovation
Adjustment Costs
Technology and Organizational Rigidities
Learning
New skills necessary to adopt correctly new
technologies
Periods of adoption of new technologies adjustment costs and decrease of levels of output
May be a lag between the growth in investment and
its benefits
Adjustment costs costs related to setting up new equipment, training of
employees (resources used to fully utilize the capital)
During the introduction of the innovation stage,
innovative firms will have a lower rate of productivity
growth than non-inovative firms
More productive firms are those that are more
capable to deal with adjustments costs and liquidity
constrains
Bessen (2002)
Bernstein et al. (1999)
Hall (2002)
Leung (2004)
Table 1- Theoretical arguments that explain the negative relationship between innovation and
productivity
These theories offer explanations for a short-run negative influence of innovation on
productivity, but so far, they lack empirical testing. More precisely, it has not been, with the
exception of Conceição et al. (2003) and to the best of our knowledge, possible to use micro30
level data matching information on the innovating behavior of firms with productivity levels
and dynamics and thus a direct test of the hypotheses suggested by the theories has been
hampered.
31
Chapter 3 - Data
3.1 - The Third Community Innovation Survey
3.1.1 - The CIS Methodology
In this section we describe the database used and its advantages and limitations in the context
of an innovation / productivity study.
The theories described above provide a framework to comprehend the relation between
innovation and productivity. However, when analyzing the empirical work done in light of
these strands (that supports the negative relationship between innovation and productivity in
the short run), we identified a shortcoming: there are few studies done with micro-level data
that make possible studying the relationship between the innovating behavior of firms and
productivity levels and dynamics (Ahn 1999; Basu et al. 1998; Bessen 2002).
Some empirical tests were performed at the macro level, considering the impact of aggregate
indicators of technological change (Basu et al. 1998). Other works considered patents as
proxies of innovation, (Crépon et al. 1998; Jaffe et al. 2000; Lanjow and Schankerman 1999),
even though the deficiencies of patents as proxies for innovation have long been documented
(Pavitt 1982; Bessen and Maskin 2000), along with the fact that they only analyze radical
innovations and, in consequence, do not capture the bulk of small firms innovation activities
and the innovation behavior in countries where there is no patenting tradition. In other cases,
micro-level analyses have been performed, but using indicators of research and development
(an input to innovation) and not indicators that may correctly proxy the output of the
innovating process (Mulkay et al. 2000). Yet another kind of work, analyzed the diffusion of a
single technological innovation, as the computer and related information technologies
(Hubbard 1998). In order to avoid some of these methodological problems, we used data from
innovation output at the firm level, in other words, a variable that capture the innovation that
occur at the firm level and not at the market level.
In 1992, the statistical agency of the European Union (Eurostat) coordinated an effort to
collect firm-level data on innovation in the EU member countries - the Community
32
Innovation Survey (CIS I). Using the CIS methodology the data is collected by the National
Offices for Statistics of each country using a similar questionnaire and comparable sampling
procedures. Following this first survey the CIS II and the CIS III, collected information on the
innovation activities of European firms in the reference periods of 1994 to 1996 and 1998 to
2000 respectively.
As referred above, one of the main advantages of the present study is that it uses innovation
variables directly measured at the firm level by the CIS III. The CIS III is a survey executed
under the supervision of the European Community (EU) centered on the observation and
collection of quantitative data about technological innovation (Bóia 2003). Details about the
way in which the survey was conducted in Portugal, which followed the general Eurostat
rules, can be found in Bóia (2003) that followed the work done by Conceição and Ávila
(2001) for the CIS II (the Portuguese version of the CIS III is presented in Annex I).
Frenz (2002) stated that comparing CIS II and CIS III is a very complicated task since the
surveys are dissimilar in the underlying questionnaires and in the industry sectors surveyed.
Despite this fact these two surveys are useful instruments to assess the dynamics between the
two survey periods, 1994 to 1996 and 1998 to 2000. These surveys permit the assessment of
the evolution of innovation in European countries given that the innovation output questions
in the two surveys focus mainly on whether or not an enterprise engaged in a certain type of
innovation activity, such as product (goods or services) and process innovation.
The CIS surveys are the most complete surveys into innovation activities in the European
context and give a unique opportunity to shed light on these activities. The questions asked
relate to input and output measures of innovation, aims and effects of innovation, factors
hindering innovation, co-operations and governmental support. General information about the
enterprise is also presented, in particular enterprise size, in terms of turnover and
employment, and nationality of ownership.
The CIS, developed in the context of the Oslo Manual OECD (1992, 1996), was created in
order to capture several aspects of the firm level innovation process:
(a) A larger variety of innovation activities besides the R&D expenditures, such as the
acquisition of patents and licenses, product design, personnel training, trial production, and
market analysis;
(b) Indicators of innovation output other than patents, such as the introduction of new
products, processes and organizational changes, the percentage of sales arising from new
products, the percentage of sales arising from products new to the industry, and the share of
products at various stages of the product life-cycle; and
33
(c) Information about the way innovation proceeds, such as the sources of knowledge, the
reasons for innovating, the perceived obstacles to innovation, the perceived strength of
various appropriability mechanisms, and the resource to research cooperation (Mohnen and
Dagenais 2002).
The CIS methodology was developed to analyze the innovation process from the enterprise
perspective and within its boundaries, as an alternative method to the ones that examine this
process from the invention point of view. In this context, data is assembled in order to
characterize the organization of activities oriented towards innovation, the motivations
underlying the introduction of innovations, the difficulties hampering the innovation process,
and the network of liaisons and cooperation’s with other entities as other enterprises (parent,
suppliers, clients, competitors), Universities and R&D Laboratories (Bóia 2003).
Along with other questions on the resources devoted to innovation activities, the impact of
innovation in the firms’ turnover, and other more qualitative questions on the nature of the
process of innovation, this European-wide survey employs a harmonized questionnaire to
inquire if firms have introduced at least one innovation within two years (the survey looks at
innovation, in general, not only at the adoption of a specific technological innovation, such as
computers). Since this survey looks at innovation in a different way than the survey that are
centered in technology and not in the firm, this questionnaire’s structure may help to
understand the relationship between innovation and productivity. It provides, in addition to
the innovation inquiries, matching data on productivity, changes in productivity in the shortrun and other variables considered in the literature to influence productivity levels and
dynamics (exports as a share of total revenue, firm size, and expenditure on R&D, among
others).
Also important to make a complete interpretation of the results, is to know the way in which
the questions are raised. In the CIS the innovation question is asked as a binary query: have
your firm incorporated any innovation in the last two years? So, when analyzing the output of
this question we have to understand that firms that introduced only one innovation with little
impact and firms that introduced several crucial technologies in their production scheme are
treated as equals (the Portuguese version of this question was complemented by a query,
serving as a validation question, that asked firms to describe their innovations). After this
question and if the answer was no, it was asked if the firm tried to innovate. To the firm that
assumed to have any innovation activity (broad definition), a number of questions associated
with the resources devoted to innovation, the objectives, barriers and information sources are
asked.
34
In brief, the advantages of the survey data are:
1) Separation between firms that do not innovate, those that have attempted to innovate and
innovative firms;
2) Gathering information, not only about radical innovations linked to patents applications,
but also about not radical innovations in the context of the market but new to the firm. By
crowding this kind of information, the questionnaire contributes decisively to understand the
innovation process in countries where patents are not common, such as Portugal that is far
from being in the technological frontier;
3) Inquiring firms, not only from the manufacturing sector, but also from the service sector,
making possible a more complete analysis from the Portuguese economic reality;
4) Existence of information that permits the creation of instruments to correct endogeneity
between innovation and productivity;
5) Differentiation between product and process innovation.
3.1.2 - Studies using the CIS Data
Therefore, this survey is an important instrument to measure innovation activities, because it
clarifies an important set of innovation-related activities of firms. So, we are capable to follow
the more micro approach of Kline and Rosenberg (1986), Bresnahan and Trajtenberg (1995)
and Conceição et al. (2003), centered on “what firms do”, rather than looking at a
technological breakthrough such as the computer.
Given the vast number of questions in the CIS surveys, several studies and conclusions can be
drawn using a variety of variables and permutations. In this context, Frenz (2002) proposed
some principles that should guide the outline of a report using the CIS data: (1) comparability
of data between the different CIS; (2) use of as many observations as possible; (3) analysis of
categories according to the main sector (production versus distribution and services); (4)
analysis according to level of knowledge and technology (low and high-tech sectors); and (5)
making comparisons in the most relevant variables in the survey, that is outputs, inputs of
innovation and constraints.
Illustrating the important role that the CIS data can take to understand the European
innovation process, some works were made using this data in order to study different
innovation process characteristics. The results obtained by Cassiman and Veugelers (2002,
2003); Mohnen and Dagenais (2002); Leiponen and Helfat (2003) and Hipp et al. (2003)
35
show the reality of the firms in several countries and thus constitute proof that the data
collected using this methodology may be adequate to study innovation at the micro level. To
stress this statement, we describe briefly some other works that used the CIS data to analyze
the innovation process in several countries.
Frenz (2002) considered that the CIS are the most comprehensive surveys into firms’
innovation activities in the UK because it provides data on direct innovation outputs and
factors influencing innovation activities across most industry sectors. As referred above, his
work focused on a comparison between various forms of innovation output in CIS II and CIS
III, contrasting with previous research on innovation that has used patent data or data on R&D
expenditures as a proxy for innovation output or focused on high technology intensive
manufacturing industries.
In this framework, Frenz (2002) compared the number of self-declared innovators in CIS II
with CIS III in UK and reached the following main results: 1) the number of firms reporting
innovation output in CIS III is substantially below the proportion of enterprises declaring to
have had some innovation output in CIS II and 2) there was a decrease of 19 percent in the
proportion of product (goods and services) innovators.
Examining firms’ innovation inputs (intramural and extramural R&D, personnel in R&D), the
author made a different picture. The number of enterprises engaging in some forms of R&D
related expenditures, such as the acquisition of machinery in connection with innovation, is
increasing, at the same time as firms’ engagement in other areas of R&D, such as the
acquisition of external knowledge, is decreasing. Also significant is the drop in high-tech
enterprises engaging in internal R&D and the large fall of innovation outputs in the high-tech
sector.
After analyzing the raw data, some factors contributing to a decline in the proportion of
enterprises with innovation output were identified.
The first factor was the increase in the number of enterprises reporting unsuccessful
innovation projects by approximately ten percent. Despite the fact that this figure indicates
that obstacles to successful innovation output have increased between CIS II and CIS III, the
author did not assessed the proportion of firms affected by specific innovation constraints
because of differences in the questionnaires. However, it was identified a change in the
distribution of firms affected by prevalent innovation constraints towards financial constraints
to innovate, such as direct innovation costs and costs of finance.
The second aspect was the concentration of innovation activities taking place between the CIS
II and the CIS III. A larger proportion of enterprises in the CIS III generated a higher
36
proportion of turnover from new or improved products than in the CIS II. One potential
explanation for this is that fewer enterprises generated higher levels of innovation output in
the CIS III than in the CIS II.
Finally, because the sample of firms is different between CIS II and CIS III since the CIS
methodology is based in a sampling method that produces different samples for the different
versions of the questionnaire, some results might be explained by these sampling differences
(Frenz 2002). In this context and despite the fact that a non-response analysis is made, the rate
of response has to be considered when analyzing the CIS data in order to analyze more
correctly the data. In the Portuguese CIS III the rate of response was 45.8 % (Bóia 2003).
Frenz (2002), in comparing innovation activities, subdivided the information of the CIS II and
CIS III surveys into high and lowtech enterprises and into production, versus distribution and
services. This procedure was taken in order to see whether these groups differ in terms of their
innovation patterns and to find evidence that technology and knowledge intensity and main
industry sector are innovation determinants. In addition, Frenz (2002) analyzed the input
factors of innovation using the following variables: proportion of firms engaging in intramural
and extramural R&D, the number of personnel working in R&D and the proportion of
enterprises continuously conducting R&D.
In a final section, he looked at various factors hindering innovation, dividing them in three
categories: economic factors, internal factors and other factors. The economic factors
identified were: the economic risks of innovation perceived by the responding enterprises, the
direct costs of innovation, the costs of finance and the availability of finance. The internal
factors that were recognized as constraints on an enterprise’s ability to innovate were:
organizational rigidity, lack of qualified personnel, lack of information on technology and
markets. The other obstacles identified were the impact of regulations or standards and the
lack of customer responsiveness to new goods or services.
Tether et al. (2001) conducted a comparative study on innovation in the service sector, using
the CIS II data. They reported the main results of the comparative analysis, discussed
methodological issues and made some policy conclusions. Tether et al. (2001) stated that
services firms are more dependent on the knowledge and skills of their employees than is
generally true for manufacturing firms, as service workers tend to be more closely implicated
in the development of services than is usually the case with manufacturing workers. Taking in
consideration that services are likely to be the principal source of economic and employment
growth into the future, these authors consider vital that European policies promote the
development of high quality services, which allows scope for innovation and reform.
37
Mohnen and Röller (2001) developed a framework for testing discrete complementarities in
innovation policy using European data (CIS) on obstacles to innovation. In this context, they
proposed a discrete test of supermodularity in innovation policy and applied it to two types of
innovation decisions: to innovate or not, and if so, by how much.
They found evidence that the existence of complementarity in innovation policies depends on
the phase of innovation that is targeted (getting firms innovative or increasing their innovation
intensity) as well as on the nature of policies implemented. In addition, they stated that the
diverse innovation process phases are subject to different constraints and that there appears to
be a need to adopt more general policies to make firms innovate, while a more targeted choice
among policies is necessary to make them more innovative (Mohnen and Röller 2001).
Dachs et al. (2004) used data from the third wave of the Community Innovation Survey (CIS
III) to analyze co-operative behavior of innovative firms in Finland and Austria. Analyzing
the descriptive statistics, they found that the rate of innovators is quite similar in Austria and
Finland, while the number of co-operating enterprises is considerably higher in Finland. The
econometric analysis revealed that most of factors that decide co-operative arrangements are
only significant in one country, so the authors concluded that co-operative behavior in the two
countries is much more dependent on national factors and much deeper rooted in the
underlying innovation systems than the existing literature may assume.
Castellacci (2003) investigated the reasons behind innovative performance differences across
manufacturing industries in Europe using the CIS-SIEPI database, which includes CIS II data
on the innovative activity of firms in 22 manufacturing industries in ten European countries.
With this data, Castellacci (2003) presented a model where each sector’s absorptive capacity
determined the technological trajectory, the intensity and type of innovation expenditures, and
the intensity of interactions between firms and the users of new technologies.
The econometric model showed that innovative capacity in a specific sector is positively
related to the share of R&D expenditures, to the intensity of user-producer interactions, to the
ability to imitate, and to the sector overall innovative intensity.
3.1.3 - CIS Methodology Limitations
Despite the advantageous characteristics of the CIS described above and the several studies
using it, this data have some gaps when used to study the relationship between productivity
and innovation that must be cited in order to contextualize the results obtained in this thesis.
38
The first problem was already referred and is associated to the innovation variable captured in
this survey. Regardless of the fact that the survey is accompanied by an annex where the
concept of innovation is explained through examples and the existence of a controlling
variable that asks the description of the innovation activities of the firm, the response to this
question is very dependent of pre-concepts that the respondents have about innovation.
In addition, this variable does not differentiate between the innovation intensity and relevance
of different firms: firms that introduced only one innovation with little impact and firms that
introduced several crucial technologies in their production scheme are treated as equals. This
fact, that is almost inevitable when using data from surveys, introduces a bias when studying
the effect of innovation activities on productivity. Possible ways to overcome or mitigate this
specific methodological problem may be the realization of case studies, a methodology that
implies a loss of representativeness of the study and a change of the firm level character of the
study or the introduction of some specificity in the question regarding innovation in future
surveys.
The little information on the production structure included in the CIS data (a survey can only
include quantifiable information), only permits the use of a specific productivity variable:
logarithm of the change in the ratio of total revenues (firm turnover) over the number of
workers from the end of 1998 to the end of 2000. This variable is reductive because does not
give information on the specifities of the different firms and so, may bias the productivity
measure (for example, service firms have a completely different production scheme than
manufacturing firms and so it is difficult to compared levels of productivity only by looking
to a single number). Once again this problem is difficult to overcome when using survey data,
since it is very difficult to ask very firm specific questions when enquiring a large amount of
firms.
In spite of these gaps and as referred above, it has to be stated that the data from the CIS is a
good instrument to understand the relationship between productivity and innovation.
39
Chapter 4 – Model and Methods
4.1 – The Model
4.1.1 - Productivity Equation
After the analysis of productivity studies that discuss the different firm characteristics
influencing this variable and taking in consideration the work done by Conceição et al. (2003)
that used the CIS II data to produce a similar study, we are able to propose a model that aims
at comprehending the link between productivity and innovation in the short run.
Log (Pr dG ) i = α 0 + α 1 Inov i + α 2 Exp i + α 3 NFi + α 4 GPi + α 5 EDi + α 6 CS + S i + ε i (1)
Where: PrdG is a measure of productivity; Inov is the innovation dummy variable; Exp is the
export variable; NF is a dummy that measures the entry of the firm in the market; GP is a
dummy for whether or not the firm is part of a group of companies; ED is the share of the
company’s workforce engaged in specialized highly qualified tasks; CS is log of the gross
investments in tangible goods and S are sector dummies, constructed taking in consideration
the 2 digit NACE of each firm and including manufacturing, service and mining sectors.
The predicted variable employed is a measure of labor productivity growth and is calculated
as the logarithm of the change in the ratio of total revenues (firm turnover) over the number of
workers from the end of 1998 to the end of 2000. On the right hand side, the model includes
predictors that have been found to be critical determinants of productivity growth, as
reviewed in a previous section, with special attention to the innovation variable.
Exports (Exp) will be considered as the share of total turnover of the company that is
transacted in foreign markets. Firm entry (NF) is a dummy that will be one if the company has
been created within the last two years and zero otherwise. To account for the skill structure of
the workforce, a variable denoting the share of the company’s workforce engaged in
specialized highly qualified tasks is used (ED).
40
In order to measure the influence of the role of management and strategy of the firm in
productivity, we considered the log of the gross investments in tangible goods (CS), a variable
that exposes the investment strategy of the firm. These kinds of investments are a possible
indicator of a firm’s strategy towards enhancing productivity since productivity growth is
often linked to investments in capital goods. Since it is expected that this kind of investment
will increase the level of productivity because, the variable is included as a regressor.
Nevertheless, this indicator as to be analyzed carefully given that it is not a direct measure of
management and firm strategy. It provides only an indication of it (Jorgenson 1988; De Long
and Summers 1991; Caselli and Wilson 2003).
Besides these variables, a set of controls was also considered. The first is a dummy for
whether or not the firm is part of a group of companies (GP). The others (S) are sector
dummies, constructed taking in consideration the 2 digit NACE of each firm and including
manufacturing, service and mining sectors (see Table 2).
Sub-sector
Mining and Quarrying
Manufacturing
Electricity, Gas and Water Supply
Wholesale Trade
Transport, Storage and Communication
Financial Intermediation
Computer and Related Activities
Research and Development
Architectural and Engineering Activities
Technical Testing and Analysis
NACE
Code1
10 to 14
15 to 37
40 to 41
51
60 to 64
65 to 67
72
73
74.2
74.3
Table 2 - Surveyed Sub-sectors
(Bóia 2003)
It is clear from the objective stated in previous sections of this thesis that the main
independent variable is innovation (Inov). So it is essential to define correctly this variable. In
order to do so and following the methodology proposed by Conceição et al. (2003), we used
two different variables to measure this firm characteristic: a broad definition and a narrow
definition. The broad definition establishes a value of one for the variable in companies that
have tried to innovate and succeeded and also for those firms that attempted to innovate but
1
Statistical Classification of Economic Activities (NACE, Rev. 1) for the European Community
41
were never capable to really innovate. In the narrow definition, the value of one is established
only to firms that have really innovated. The option of considering two variables to measure
the same economic reality is justified by the zero-one characteristics of both variables that do
not permit a complete analysis of the innovation attitude of the firm when using only the
narrow definition. The broad definition includes firms that, although have not finished any
innovation project, have engaged some resources in this kind of activities and is a way to
solve, at least partially, the selection problem raised by Crépon et al. (1998), which led them
to use a Heckman correction. These firms may have a more innovative than non-innovative
character and so the narrow definition may reduce the explicative power of our model.
4.1.2 - Endogeneity
As referred above, this model has a structural characteristic that makes the Ordinary Least
Squares (OLS) estimator unable to yield efficient and unbiased estimates of the coefficients.
The inclusion of two variables (innovation and productivity growth) that are generally though
to be endogenous (Crépon et al. 1998; Lööf and Heshmati 2004) would lead to biased
estimates. To statistically test the endogeneity hypothesis, it is common to use a Hausman
test. We provide an illustration in annex II that only partially covers the analysis made here
for illustrative purposes, but we proceed assuming that the endogeneity is present on
theoretical grounds.
So, it is necessary to find a mechanism to expunge this contemporary correlation between the
relevant independent variable and the error term. To solve the endogeneity that occurs when
one model is embedded within another it is frequently convenient to estimate the model in
two stages, where the second stage model includes variables defined from parameters
estimated in the first stage equation. In the case of the model developed in this paper, the
covariance matrix of the second stage estimator includes noise induced by the first-stage
estimates since the equations have different structural forms (logit and OLS).
This issue has been subject of research. Amemiya (1978) derived the asymptotic covariance
for two-stage estimation of multinomial logit models when both stages use the same
observations. Heckman (1979) determined the correct asymptotic covariance matrix of the
two step estimator that absorbs the sample selection bias by using a subset of the first stage
observations in the second stage observations. Further work by Murphy and Topel (1985)
corrected the covariance for two-stage estimation with maximum likelihood and least squares
42
using the same observations in the two stages. Based on this paper Greene (2000) showed that
its results can be applied under various specific models, and Newey and McFadden (1994)
and McFadden (1999) extend them to generalized method of moments (Karaca-Mandic and
Train 2003).
Karaca-Mandic and Train (2003) applied a method to correct the asymptotic covariance
matrix of the second-stage estimates to the Petrin-Train application of households’ choice of
TV reception, where the different TV prices are correlated with other attributes of the
different alternatives. To correct for the endogeneity, it was specified a first stage linear
regression of prices on exogenous variables and instrumental variables using market level
data and a second stage were a multinomial logit model is applied.
In this context, the method used in the present paper follows a two-stage estimation method
for mixed models that include limited dependent variables. This method takes into
consideration the fact that the variable innovation is a dichotomous variable and thus it is
impossible to use a two stages least squares (2SLS) model to overcome the endogeneity issue.
The procedure follows the following steps:
1) the innovation variable is regressed on instrumental variables – a probit model;
2) the resulting regression is employed to obtain estimates of the predicted probability
distribution for dichotomous innovation variable;
3) the initial model is ran, but with the predicted values of the first stage regression,
generating consistent estimates for the values of coefficients of the independent
variables.
Despite the consistency of the results, the fact that the endogenous variables have different
structures (one discrete and one continuous) implies that it is complex to derive the joint
distribution of the model. This problem is reflected in the asymptotic covariance matrix
originated by the second stage regression that it is not consistent and has to be corrected to
account for the use of instrumental variables.
As described by Murphy and Topel (1985), the two-step procedure only replaces the
unobserved components with their estimated or predicted values from the auxiliary model.
Despite the fact that this procedure yield consistent estimates of second-stage parameters
under fairly general conditions, it is unwise to treat these values as if they are known for
purposes of estimation and inference in the second-step model. As a consequence, the secondstep estimated standard errors and related test statistics based on these procedures are
incorrect. This fact is commonly ignored in studies that contain unobservable, though
43
estimable, variables (models used very commonly in several areas of applied econometrics)
and examined as a mere problem of efficiency of the estimator.
To correct the covariance matrix we used an application, referred above, detailed by Murphy
and Topel (1985) that presented a general method of calculating the correct asymptotic
covariance matrix for the two-stage estimation procedure. Murphy and Topel (1985)
identified the failure of the two-stage method to account for the fact that the unobservable
regressors have been estimated in calculating second-step coefficients and standard errors.
This failure means that the second step of the method is not consistent because the imputed
applied in the second step are measured with sampling error. So, they assumed that the
auxiliary model for the unobserved variables creates consistent estimates of both first-step
parameters and their asymptotic covariance matrix and thus the sampling error of the
unobserved variables disappears in the limit. As a result, the second-step parameters are
consistently estimated and the limiting distribution of this error may be used to consistently
estimate the variances of the second-step parameter estimator.
This empirical method gives more explanation power to the model results since it corrects for
the fact that innovation and productivity change are simultaneously determined in the sample.
4.1.3 - Innovation Equation
Within this framework and in turn to obtain results that avoid the problems associated to
endogeneity, we run a first-stage Probit equation in order to predict the innovation variable.
Pr( Inov i = 1) = β 0 + β 1 Log _ Turn _ Inic i + β 2 NFi + + β 3 GPi + β 4 S i + η i (2)
The instrumental variables2 used in this equation, that precedes equation (1), are: the turnover
levels at the beginning of the two year period (1998-2000) in log form (Log_Turn_Inic); two
binary variables, one that reports if the firm is part of a group (GP) and one that indicates for
whether the firm is new or not (NF); and controls for industrial sector (S).
2
“In an equation with an endogenous explanatory variable, an Instrumental Variable is a variable that does not
appear in the equation, is uncorrelated with the error in the equation, and is (partially) correlated with the
endogenous explanatory variable” (Wooldridge 2003)
44
As explained previously, this equation, in conjunction with the productivity equation (that
was detailed above) permits an estimation using instrumental variables. With this procedure,
consistent and asymptotically normally distributed estimates for the values of coefficients of
the independent variables in the original equation are generated. Although, the success of this
estimation is dependent of the correct definition and operability of the critical identification
variable: the turnover at the beginning of the two year period (1998). This variable is essential
since it is not correlated with productivity and thus permits the identification of all model
parameters (coefficients and covariances).
The option of the turnover to be the critical instrumental variable was based on the thought
that it is a good predictor of innovation. This variable is an indicator of the capability of firms
to go around liquidity constraints and, simultaneously, it is not a good predictor of
productivity. Even so, the use of this variable as the critical instrumental variable can imply
some econometric problems, since we only have one lagged turnover value. With this
structure, it is possible that this variable may not give the right idea of the real capability of
firms to face liquidity constrains because only one year may not be representative. Thus, the
ideal scenario would be to have the same variable lagged two or three time-periods, which is
an impossibly situation given that the CIS methodology is not based in the inquiry of a fixed
panel of firms (it is impossible to lag the variable because the information was not collected
from the same firms in the previous surveys).
So, it is advisable to test the robustness of our results using alternative critical identification
variables. Conceição et al. (2003) proposed two alternatives: the initial productivity level of
the firm and the average innovation of the remaining firms in the same NACE level 5 sector.
Even though they assumed that these variables are not perfect instruments, they obtained
similar results across these different specifications and concluded that these results are an
indication of the robustness of the results.
4.2 - Methods
This section aims at describing in a more insightful manner the econometric method used in
this work to overcome the endogeneity problem associated to the data: the Murphy Topel
application.
45
4.2.1 - Murphy Topel Application
As described above, after the endogeneity test, it was applied a method that, when making the
calculation of the estimation results, corrects the covariances matrices of the two steps
estimation model: the Murphy-Topel application. The aim of this section is to complement the
report done on the model description section.
As referred above, to answer to the endogeneity issue that occurs in a system of equations it is
advisable to estimate the model in two stages (using instrumental variables). In case of the
model developed in this thesis, the covariance matrix of the second stage estimator includes
noise induced by the first-stage estimates since the equations have different structural forms
(logit and OLS).
The method of instrumental variables assumes that a subset Xw of the independent variables
is correlated with the error term in the model and that, in consequence, a matrix Xs of
independent variable is correlated with Xw. In addition it is assumed that X, Y and Xw are
uncorrelated.
Based on these premises, it is possible to construct an approximately consistent estimator.
This estimator is obtained by performing a regression for each of the independent variables of
Xw on the instruments and the independent variables not correlated with the error term (Xz
Xs). Predicted values are obtained from each regression and substituted for the associated
column of Xw in the analysis of the generalized linear model (GLM) of interest. This
structure provides an approximately consistent estimator of the coefficients in the GLM (it is
consistent in the linear case) (Murphy and Topel 1985).
Taking this method as a starting point, Murphy and Topel (1985) described the two-step
estimator (T-S) using a set of two theorems.
Theorem 1 stating that the correct covariance matrix for the second-step estimators in the T-S
procedure exceeds the commonly reported asymptotic covariance matrix, σ2Q0-1, by a
positive-definitive matrix, and that, consequently, standard errors from the initial two-step
procedure are understated.
Theorem 2 reflects two aspects of the method: that the result may be easily generalized, via
conformable derivations, to situations in which either model or both models are nonlinear and
that second-stage standard errors of structural parameter estimates must account for the fact
that reduced-form parameters are estimated in the first stage (Murphy and Topel 1985).
46
To validate their work, Murphy and Topel (1985) compared their estimator (T-S) with the
usually used instrumental variables (IV) estimation. In the first place, they stated that the twostep procedure and the standard error corrections presented in this article contain IV as a
special case. In addition, they defend that the T-S estimator is appropriate in many cases in
which IV estimation is not or in which IV is simply infeasible because the T-S estimator has
the additional advantage of considering the model’s structure in estimating the first stage or
reduced form.
To illustrate these points, these authors proposed a simple reflection about the functioning of
the methods.
Assuming that a second-step dynamic model contains, in addition to variables based on
present information, an unobserved expectation based on past information, the current
variables become invalid instruments for the second-stage equation and so they will be
correlated with the prediction error. In the hypothesis of an adequate number of instruments
remain; this will lead to less efficient estimation of the second-step parameters. This fact
occurs given that the values of current variables would be substituted by predicted values
based on the last period’s information when estimating the second stage. If the loss of current
variables reduces the number of available instruments below that needed for identification,
then the IV method could not be realized.
On the other hand, they also defended that the T-S procedure avoids these problems by
considering the model’s structure when creating the predicted values for second-stage
estimation. Therefore, the two-step procedure uses current variables to predict the current
variables in the second step while imposing the constraint on the estimated reduced form that
these current variables do not affect the unobserved expectation. These facts expose the
fundamental advantage of the T-S estimator over the IV estimation: it utilizes the model’s
structure and associated constrains in estimating the reduced form (Murphy and Topel 1985).
In this context, Murphy and Topel (1985) presented the method that was used in this thesis to
correct for the fact that two variables are correlated in a two step model with two equations of
different types. In the next paragraphs, we will describe numerically the methodology,
following the work of Hardin (2002) that describes this methodology in a brief and
understandable manner.
Various models have been proposed in the literature in which one model is embedded in
another. Such models are broadly known as two-step estimation problems and are
characterized by the following general form:
47
Model 1: E {y1|x1, θ1}
Model 2: E {y2|x2, θ2,E(y1|x1, θ1)}
The model structure indicates that there are two parameter vectors to estimate: the parameter
vector θ1 that is present in both models and the second parameter vector θ2 that appears only
in the second model.
As referred above, this kind of models can be estimated using two methods. The first
approach is a full information maximum likelihood, FIML, model in which the joint
distribution f(y1, y2|x1, x2, θ1, θ2) is specified and the joint log-likelihood function is
maximized. The second alternative is to adopt a limited information maximum likelihood,
LIML, two-step procedure, in which, the first model is estimated, since it does not include the
second parameter vector. In this context, the second parameter vector is estimated taking in
consideration the results of the first step estimation; to, in sequence, maximize the conditional
log-likelihood L given by:
n
^
⎧
⎛
⎞⎫
L = ∑ ln f ⎨ y 2i | x2i ,θ 2 , ⎜ x1i θ1 ⎟⎬
⎝
⎠⎭
⎩
i =1
Where it is assumed that there are n observations, x1i is the ith row of the X1 design matrix, x2i
^
is the ith row of the X2 design matrix, and θ1 is the maximum likelihood estimate obtained
from the estimation of Model 1.
After identifying the problems associated to the estimation of the parameter θ2, Hardin (2002)
described the method developed by Murphy and Topel (1985).
The method is based on a general formula of a valid variance estimator for θ2 in a two-stage
maximum likelihood estimation model. This LIML estimation fits one model, which is then
used to generate covariates for a second model of primary interest.
So, the calculation of a variance estimate for the explanatory variables θ2 in the primary
model must take in concern the information that one or more of the regressors have been
^
calculated through (x1, θ1 ).
Following Greene (2000), the formula for the Murphy–Topel variance estimate for θ2 is given
by:
48
[
]
V2* = V2 + V2 CV1C ' − RV1C ' − CV1 R ' V2
Where
⎡∧ ⎤
V 1 = Asy .Var ⎢θ 1 ⎥
⎣ ⎦
based on ln L1
⎡∧⎤
V2 = Asy.Var ⎢θ 2 ⎥ based on ln L2|θ1
⎣ ⎦
⎡⎛ ∂ ln L2
C = E ⎢⎜⎜
⎣⎢⎝ ∂θ 2
⎞⎛ ∂ ln L2
⎟⎟⎜⎜
'
⎠⎝ ∂θ1
⎡⎛ ∂ ln L2
R = E ⎢⎜⎜
⎢⎣⎝ ∂θ 2
⎞⎛ ∂ ln L1 ⎞⎤
⎟
⎟⎟⎜⎜
' ⎟⎥
∂
θ
⎠⎝
1 ⎠⎥
⎦
⎞⎤
⎟⎟⎥
⎠⎦⎥
In this method V1 and V2 are calculated as the inverse matrix of negative second derivatives, a
technique that is not essential. So, the asymptotic equivalence of these estimators is given by:
⎧ ∂2L ⎫
⎧⎛ ∂L ⎞⎛ ∂L ⎞⎫
E ⎨⎜ ⎟⎜ T ⎟⎬ = − E ⎨
T ⎬
⎩⎝ ∂θ ⎠⎝ ∂θ ⎠⎭
⎩ ∂θ∂θ ⎭
The component matrices of the Murphy–Topel estimator are estimated by evaluating the
^
^
formulae at the maximum likelihood estimates θ1 and θ 2 . The presentation assumes the
existence of a log likelihood for the first model L1(θ1) and a conditional log-likelihood for the
second (primary) model of interest L2(θ2|θ1) (Hardin 2002).
The correction of the asymptotic covariance matrix at the second step requires some
additional calculation, since the matrices R and C are obtained by summing the individual
observations on the cross products of the derivatives. These are estimated with,
⎛
1 n ⎜ ∂ ln f i 2
C = ∑⎜
n i =1 ⎜ ∂ θ∧
2
⎝
∧
⎞
⎞
⎞⎛
⎛
⎞⎛
∧
1 n ⎜ ∂ ln f i 2 ⎟⎜ ∂ ln f i1 ⎟
⎟⎜ ∂ ln f i 2 ⎟
⎟ and R = n ∑ ⎜
⎟
∧
∧
∧
⎟⎟⎜
⎟⎟⎜
i =1 ⎜
⎜ ∂θ ' ⎟
⎜ ∂θ ' ⎟
θ
∂
2
⎠⎝
⎝
⎠⎝
1
1 ⎠
⎠
49
Chapter 5 – Results and Conclusions
5.1 - Results and Discussion
After the theoretical framework, the presentation of the data and the description of the model,
we introduce the descriptive statistics and the results obtained with the model previously
presented and some conclusions are drawn.
5.1.1 - Descriptive Statistics
Prior to analyzing the regressions results and in order to characterize the analyzed sample and
ease the results interpretation, we present some of the descriptive statistics of the CIS III
database relevant for this study.
Table 3 – Descriptive Statistics
50
Table 4 – Descriptive Statistics (continuation)
The first significant fact is that, on average, the firms inquired by the survey have improved
their productivity by 24%. This productivity growth is inflated by the minority sectors
(E.G.W. – Electricity, Gas and Water and Mining) that have rates of growth much higher than
the average and, thus, may bias the result from the real economic value. Despite the fact that
the industry sector has a more uniform profile (the standard deviation is almost half the value
from the standard deviation of the service sector), the manufacturing and services firms have
similar rates of improvement in productivity (19%). In conclusion, the average productivity of
Portuguese firms has improved significantly between 1998 and 2000, but with a skewed
distribution that influence the generalization of these findings. These results are coherent with
the results obtained with the CIS II database and that where shown by Conceição et al. (2003).
In what concerns innovation, almost 44% of the firms have introduced some new or
significantly improved product in the market or improved or introduced a new production
process. Almost 46% have engaged in some form of innovation activity (complete or
incomplete). Unlike the CIS II (Conceição et al. 2003), in the CIS III the services sector (47%
innovative firms and 48% firms tried to innovate) is slightly more innovative than the
manufacturing sector (43% innovative firms and 45% firms tried to innovate).
Only 2.7% of the surveyed firms were founded between 1998 and 2000, but the services
sector (3.7%) has a higher percentage than the manufacturing sector (2.1%). This fact is
51
justified by the tendency of the Portuguese economy to be more service dependent, following
the trend verified in many countries and that constitutes a global tendency. These results also
reflect the increasing entrance of multinational firms in national markets that contributes to
the terciarization of markets.
Almost a third of the firms are part of a group, a figure that is not equal to the two main
sectors – manufacturing and services. The share of service firms (44%) that are part of a
group is almost twenty percentage points above the share of manufacturing firms (25%) in the
same situation.
As expected, the manufacturing firms (27%) are much more export oriented, having a bigger
average share of turnover resulting from sales to foreign countries than the services firms
(7%). The overall average is 21%. Portugal does not provide a considerable amount of
services to foreign countries and so, the manufacturing sector, despite the loss of relative
weight in the Portuguese economy, is still the most exporting sector in Portugal.
Also expected is the difference between the shares of high qualified workers in the services
and manufacturing sectors. With an overall average of 34%, the services sector has an average
of 43.5% high qualified workers against a 31% average in the manufacturing sector. This
dissimilarity gives an indication about the more capital intensive characteristic of the
Portuguese secondary sector relatively to the tertiary sector that is more skill intensive.
The investment in tangible goods per employee is very similar in the two sectors and it is not
a structural difference between them.
In what concerns the response rate, it has to be referred that does not exist a considerable
difference between sectors: the services sector had an initial response rate of 44.9% and the
manufacturing sector of 46.16%. So, the difference between response rates does not explain
possible bias in the econometric analysis.
Doing this analysis, we have to report that some variables from this database have missing
values which reduce the observations in the regressions. So, since imputation was not
possible, the weights were not applied and thus the representativeness of the sample could not
be confirmed.
We could go around this problem by analyzing the distribution of the missing values and
testing if its distribution is uniform. If the results of this test were positive, the missing values
would not bias the regression coefficients.
Unfortunately, we could only make this test to the innovation variables since these were the
only variables that were enquired in the non-response questionnaire and, therefore, the
52
extrapolation of the better part of the variables used in the present paper could not be
completed.
5.1.2 Regression results
5.1.2.1 - Complete Sample
Table 5 summarizes the regression results obtained when the model described above was
tested with the complete CIS III database. These are the main results of this thesis since they
are obtained using the CIS complete sample, giving an approximate estimation of the short
run relationship between productivity and innovation for the all of the Portuguese economy.
Table 5 – Regression results for the complete sample stage
Note: * Significant at 10%; ** 5%; *** 1%; Sector Dummies included but not reported
The first main result is that the innovation coefficients are negative and significant for all the
models tested. This result corroborates the theoretical assumptions exposed in the second part
of the literature review chapter (see Table 1), since these theories suggest that, in the short
run, innovating activities can lead to productivity losses. The first theory focuses on the
learning lags that result from introducing a new technology (Jovanovic and Nyarko 1996;
Ahn 1999, 2001). The second is centered in the fact that new technologies, in the first steps of
the diffusion process in the economy, are not as perfected and developed as older ones
(Leonard-Barton 1988, 1992; Young 1991, 1993; Utterback 1994; Christensen and Bower
1996; Christensen 1997; Tripsas and Gavetti 2000; Benner and Tushman 2002). The last is
53
concerned with the adjustments costs associated with the introduction of innovation
(Bernstein et al. 1999; Bessen 2002; Hall 2002; Leung 2004).
The variables share of exports and share of high qualified workers are not significant and
when they are removed from the model, it gains robustness (the coefficient of the innovation
variable becomes more significant). The behavior of the exports variable is unexpected since
exports were considered to be a determinant of productivity in empirical work conducted by
another authors. Despite the fact that the qualification variable also has a theoretical
background, the result concerning this variable is not surprising since the quality of data
obtained in the community innovation survey has some consistency problems (missing values
and possible problems of mis-interpretation of the question by the respondents). Another
hypothesis that can explain the non-significance of the qualification variable is the nature of
some economic sectors that are capital intensive and its success is not based on the
qualification of its workforce.
The other coefficients are consistent with the theoretical framework described above and
significant, more precisely, be a new firm (this variable has to be analyzed carefully since the
group of new firm represent only 2.7% of the sample) and part of a group are characteristics
that enhance productivity growth of firms. The variable that measures the management and
strategy (investment in tangible goods per employee) is significant and positive in all the
models, which indicates that, the investment in capital goods may contribute to a higher
productivity growth (despite the fact that the coefficient is very small).
Although the sector dummies coefficients are not reported in result tables, it is pertinent to
make a brief analysis of these results. The majority of the sector dummies coefficients are not
significant at 10% and the significant ones are all negative, with the exception of the dummy
associated to a very singular economic sector, water distribution, which is a natural monopoly
and has to be considered as an outlier.
These results are similar to the two innovation variables used in this study: actual innovation
and attempt of innovation.
5.1.2.2 - Complete Sample (process and product innovations)
Based on some theoretical arguments (described in the literature review) that defend that
product and process innovation have dissimilar profiles, Table 6 describes the results obtained
when the analysis was restricted to only one kind of innovation: product or process. The
54
results for these two scenarios are very similar and coherent with the results of the overall
regression. In terms of productivity dynamics, there is not any substantial difference between
the behavior of process innovators and product innovators.
Table 6 – Regression results for the complete sample (process and product innovations)
Note: * Significant at 10%; ** 5%; *** 1%; SectorDummies included but not reported
5.1.2.3 - Model with the level of productivity in 2000
Table 7 - Regression results for the model with the level of productivity in 2000 as dependent variable
Note: * Significant at 10%; ** 5%; *** 1%; SectorDummies included but not reported
Table 7 presents a two-stage regression that is identical to our original model but with the
variable level of labor productivity in 2000 (in log form) as the dependent variable instead of
the labor productivity growth. This regression was made in order to partially test the adequacy
of these empirical findings to the different theories that try to justify the negative relation
between innovation and productivity in the short run.
55
Since the coefficients on the variable innovation are positive and significant for all the
scenarios, the results obtained advocate that firms that are engaged in innovation activities, on
average, have higher levels of labor productivity when compared to those that did not
introduce innovations. These results are consistent with the adjustment cost hypothesis
(described above) and with several studies that predict that firms that innovate have
productivity levels higher than average (Bernstein et al. 1999; Bessen 2002; Hall 2002; Leung
2004).
The other two theories described in this paper defend that productivity laggards are more
innovative than productive firms.
So, we can state that in the CIS III firm sample, the adjustment cost hypothesis explain a
better part of the negative relationship between productivity and innovation in the short run.
When using sales per worker as a variable as opposed to value added, we are facing an
important limitation. Portuguese firms, despite the fact that they have enrolled a recent
process of modernization, have customarily focused their activity on low labor.
Consequently, we can interpret these findings as labor saving innovations and not as
productivity enhancement innovations. This interpretation attributes the positive coefficients
on innovation in Table 7 to a more capital intensive characteristic of innovative firms than the
one observed in non-innovators and not to productivity growth.
A fact that can rivet this possible interpretation of the results is that, under this context, we
would expect to find innovative firms increasing labor productivity, even on a contemporary
basis. In addition, the responses to another question of the CIS III that asked firms what was
the labor impact of their innovations, reveal that most firms had maintained or increased their
labor force in the period under analysis. So, we can refute this limitation with empirical
background.
5.2 - Conclusions
The study of the relationship between productivity and innovation is a very active research
field. Within it, two topics are crucial in order to understand this relationship: the discussion
about the real impact of technological breakthroughs in productivity and the time period that
must be considered to study the real effect of innovation on productivity.
Overall, the relationship between innovation and productivity is expected to be positive in the
long run and at the macro level (countries and regions), but the scenario may be different in
56
the short run. Some work, including this thesis, indicates that in the short time and at the firm
level, the relationship between innovation and productivity may be expected to be negative.
The short run analysis of this relationship is an important issue since the understanding of the
effects of the first steps of innovation on the firms’ functioning can give explanations to
momentary productivity and efficiency losses. So, the revision of the determinants of
productivity growth in the short run can prevent precipitated judgments about the
effectiveness of a specific innovation, since it supports the idea that innovation effectiveness
is only correctly evaluated after a time period.
In this thesis, in order to contextualize the results, we described three theoretical arguments
(learning, technology and organizational rigidities, and adjustment costs) that justify the
negative relationship between productivity and innovation in the short run. The first theory
focuses on the learning lags that result from introducing a new technology. The second is
centered in the fact that new technologies, in the first steps of the diffusion process in the
economy, are not as perfected and developed as older ones. And the last is concerned with the
adjustments costs associated with the introduction of innovation.
In addition, we made a broader literature review where factors that may influence productivity
besides innovation were identified (youth of the firm; human capital; exports level; being part
of a group; and management and strategy of the firm) and the concepts of innovation and
productivity were characterized.
In order to illustrate the short run relationship between innovation and productivity, we used
the CIS III database. The CIS III is a nation-wide firm-level survey that measured directly
innovation by asking if firms have introduced any new process or product in the context of the
firm. In this context and considering that this survey inquired a representative sample of the
Portuguese economy and that provides complete information at the firm level of the period
1998-2000, the database is a good instrument to analyze the relationship between productivity
and innovation in the short run.
The model developed in this thesis was constructed assuming that innovation and productivity
change are simultaneously determined in the CIS III sample. Thus, in order to avoid possible
biases that result from this fact, the proposed model (that builds on the Conceição et al. (2003)
approach) is a system of two equations: one predicting innovation and other predicting
productivity. Several tests were done in order to correct this endogeneity, all indicating the
robustness of the results and proving that productivity growth is affected negatively by
innovation.
57
So, the major result that spring out from this thesis is that, in the sample of Portuguese firms
inquired in the CIS III, innovation leads to productivity losses when measured in a two year
period. To complement these results and in order to test the three theoretical trends exposed
above, it was found that more productive firms are likely to innovate more than the average.
So, this outcome is coherent with adjustment costs/liquidity constraints explanations of a
negative relationship between innovation and productivity growth.
A novelty in this study was the inclusion of a variable that measures the management and
strategy of firms, gross investments in tangible goods. Although the productivity literature
states that management and investment strategy influence the level and the dynamics of
productivity, these aspects of firm behavior are exceptionally difficult factors to quantify and
to measure. The inclusion of this variable can bring some new light to the comprehension of
the productivity/innovation relationship. By obtaining significant results in a productivity
econometric model that includes a proxy of firm strategy as an independent variable, this
thesis makes an important contribution to prove an idea defended in the literature but that
lacked empirical testing: management and firm strategy influences productivity.
In order to measure the influence of the role of management and investment strategy in
productivity, we considered the log of the gross investments in tangible goods (CS), a variable
that exposes the investment strategy of the firm. These kinds of investments are seen as an
indicator of a firm’s strategy towards enhancing productivity since productivity growth is
often linked to investments in capital goods. Since it is expected that this kind of investment
will increase the level of productivity, the variable is included as a regressor. Nevertheless,
this indicator, when used, has to be analyzed carefully given that it is not a direct measure of
the management and firm strategy.
From the results of this thesis some policies implications can be drawn. The most relevant is
that the evaluation of innovation efficiency and its impact cannot be done immediately:
technology adoption is a complex process that does not render results instantaneously. In
order to understand the real costs of adoption, it has to be considered that the adopter will take
time to fully take advantage of the technology. This time period is dependent on the firm and
technology characteristics since technology adoption is a very complex and sensitive process.
Therefore, when evaluating a new technology, decision makers at the firm and state level
have to consider this time lag between adoption and productivity impact: a technology that is
inefficient in the short run can raise productivity in the long run.
T complement the present analysis, it could be productive to conduct case studies in
Portuguese firms aiming at capturing specific factors that are associated to the short run
58
relationship between productivity and innovation. This more qualitative analysis could be
done by creating a panel of firms that would be accompanied for several years, permitting a
more dynamic approach of the innovation / productivity relationship.
A final suggestion would be to carry out this kind of analysis in other countries to do an
international comparison. With the data form the CIS from other countries, this could be done
in order to understand if these results are specific to the Portuguese context or not.
59
References
Acemoglu, D., and Zilibotti, F. (2001). “Productivity Differences”, Quarterly Journal of
Economics, 115(3), pp. 563-606.
Ahn, S. (1999). “Technology Upgrading With Learning Costs: A Solution for Two
"Productivity Puzzles"”, OECD Economics Department Working Papers, 220.
Ahn, S. (2001). “Firm Dynamics and Productivity Growth: A Review of Micro Evidence
from OECD Countries”, OECD Economics Department Working Paper, 297.
Amemiya, T. (1978). “On two-step estimation of multivariate logit models”, Journal of
Econometrics, 8, pp. 13–21.
Archibugi, D., Evangelista, R., Perani, G., and Rapiti, F. (2000). “Expenditure, outcomes, and
the nature of innovation in Italy”, in Conceição, P., Gibson, D., Heitor, M., and Shariq, S.
(eds.), Science, Technology, and Innovation Policy - Opportunities and Challenges for the
Knowledge Economy, pp. 243-268. Westport: Quorum Books.
Barua, A., Kriebel, C., and Mukhopadhyay, T. (1995). "Information Technology and Business
Value: An Analytic and Empirical Investigation", Information Systems Research, 6 (1), pp. 323.
Bartelsman, E. J., and Doms, M. (2000). “Understanding Productivity: Lessons from
Longitudinal Microdata”, Journal of Economic Literature, 37 (September), pp. 569-594.
Basu, S., Fernald, J., and Kimball, M. (1998). "Are Technology Shocks Contractionary?",
International Finance Discussion Papers, 625.
Bender, D. (1986). “Financial impact of information processing”, Journal of Management
Information Systems, 3 (2), pp. 22-32.
60
Benner, M., and Tushman, M. (2002). “Process management and technological innovation: a
longitudinal study of the photography and paint industries”, Administrative Science Quarterly,
47, pp. 676-706.
Bernard, A., Eaton, J., Bradford J., and Kortum, S. (2000). “Plants and Productivity in
International Trade”, NBER Working Paper, 7688.
Bernstein, J., Mamuneas, T., and Pasharades, P. (1999). “Factor Adjustment, Quality Change,
and Productivity Growth for U.S. Manufacturing”, NBER Working Paper, 6877.
Bessen, J. (2002). “Technology Adoption Costs and Productivity Growth: The Transition to
Information Technology”, Review of Economic Dynamics, 5(2), pp. 443-469.
Bessen, J., and Maskin, E. (2000). “Sequential Innovation, Patents, and Imitation”, MIT
Department of Economics Working Paper, 00-01.
Beyer, H., and Vergara, R. (2002). "Productivity and Economic Growth: the Case of Chile",
Central Bank of Chile Working Papers, 174.
Bóia. M. J. (2003). “Determinants of Innovation in Portugal: Designing, Implementing and
Analyzing Evidence from the Third Community Innovation Survey”, Master Thesis in
Engineering Policy and Management of Technology.
Bresnahan, T., and Trajtenberg, M. (1995). “General Purpose Technologies “Engines of
Growth?””, Journal of Econometrics: Annals of Econometrics, 65(1), pp. 83-108.
Brynjolfsson, E., and Hitt, L. (1995). “Information technology as a factor of production: the
role of differences among firms”, Economics of Innovation and New Technologies, 3, pp.
183-199.
Brynjolfsson, E., and Hitt, L. (1996). "Paradox Lost? Firm-Level Evidence on the Returns to
Information Systems", Management Science, 42 (4), pp. 541-558.
61
Brynjolfsson, E., and Hitt, L. (2000). "Beyond Computation: Information Technology,
Organizational Transformation and Business Performance", Journal of Economic
Perspectives, 14(4), pp. 23-48.
Brynjolfsson, E., and Hitt, L. (2003). "Computing Productivity: Firm-Level Evidence",
Journal of the European Economic Association, 85(4), pp. 793-808.
Caselli, F., and Wilson, D. (2003). “Importing Technology”, NBER Working Paper, 9928
(forthcoming Journal of Monetary Economics).
Cassiman, B., and Veugelers R. (2002). "R&D Cooperation and Spillovers: Some Empirical
Evidence from Belgium", American Economic Review, 92 (4), pp. 1169-1184.
Cassiman, B., and Veugelers, R. (2003). "In Search of Complementarity in the Innovation
Strategy: Internal R&D and External Knowledge Acquisition", IESE Research Papers, D/457.
Castellacci, F. (2003). “Why innovation differs across sectors in Europe? Evidence from the
CIS-SIEPI database”, presented at the International Workshop on Empirical Studies on
Innovation in Europe, University of Urbino, Italy, 1-2 December.
Castells, M. (1996). The Information Age: Economy, Society and Culture: Vol. 1 – The Rise of
the Network Society. Oxford, UK and Cambridge: Blackwell Publishers.
Christensen, C. M. (1997). The Innovator’s Dilemma: When New Technologies Cause Great
Firms to Fail. Boston, MA: Harvard Business School Press.
Christensen, C. M., and Bower, J. L. (1996). “Customer power, strategic investment, and the
failure of leading firms”, Strategic Management Journal, 17, pp. 197-218.
Comín, D. (2004). “R&D: A Small Contribution to Productivity Growth”, NBER Working
Paper, w10625.
Conceição, P., Heitor, M., and Oliveira, P. (1998). “University-based technology licensing in
the knowledge-based economy”, Technovation, 18 (10), pp. 615-625.
62
Conceição, P. and Ávila, P. (2001). A Inovação em Portugal: II Inquérito Comunitário às
Actividades de Inovação (in Portuguese). Oeiras: Celta Editora.
Conceição, P., Heitor, M., and Veloso, F. (2003). “Innovative Shocks and Productivity”,
presented at the conference What do We Know about Innovation? A Conference in Honour of
Keith Pavitt; University of Sussex, Brighton, U.K, 13-15 November.
Crafts, N. (2002). “The Solow Productivity Paradox in Historical Perspective”, CEPR
Discussion Papers, 3142.
Crépon, B., Duguet, E., and Mairesse, J. (1998). “Research, Innovation, and Productivity: An
Econometric Analysis at the Firm Level”, NBER Working Paper, 6696.
Criscuolo, C., and Haskel, J. (2003). “Innovations and productivity growth in the UK:
Evidence from CIS2 and CIS3”, CeRiBa Discussion Paper.
Dachs, B., Ebersberger, B., and Pyka, A. (2004). "Why do Firms Co-operate for Innovation? A comparison of Austrian and Finnish CIS 3 results", Universitaet Augsburg, Institute for
Economics Discussion Paper, 255.
David, P. (1990). "The Dynamo and the Computer and Dynamo: A Historical Perspective on
the Modern Productivity Paradox," American Economic Review Papers and Proceedings,
80(2), pp. 355-361.
De Long, J., and Summers, L. (1991). “Equipment Investment and Economic Growth”, The
Quarterly Journal of Economics, 106, pp. 445-502.
Dedrick, J., Gurbaxani, V., and Kraemer, K. L. (2001). “Information Technology and
Productivity Growth at the Firm and Country Level”, University of California Working
Papers.
Doms, M., Dunne, T., and Troske, K. (1997). “Workers, Wages, and Technology”, The
Quarterly Journal of Economics, 112 (1), pp. 253-290.
63
Frenz, M. (2002). “A Comparison of the Second and Third UK Community Innovation
Survey”, TESE Economics and Statistics Reports, 6.
Galarneau, D., and Maynard, J-P. (1995). “Measuring productivity”, Perspectives - Statistics
Canada, 75-001E.
Gera, S., Gu, W., and Lee, F. (1999). “Information Technology and Labour Productivity
Growth: An Empirical Analysis for Canada and the United State”, Canadian Journal of
Economics, 32(2), pp. 384-407.
Greene, W. (2000). Econometric Analysis. New Jersey: Prentice Hall.
Hall, R. (2002). “Industry Dynamics with Adjustment Costs”, NBER Working Paper, 8849.
Haltiwanger, J. (2000). “Aggregate Growth: What Have we Learned from Microeconomic
Evidence?”, OECD Economics Department Working Papers, 267.
Hardin, J. (2002). “The robust variance estimator for two-stage models”, The Stata Journal, 3,
pp. 253–266.
Hausman, J. (1978). “Specification tests in Econometrics”, Econometrica, 46(6), pp. 1251-71.
Heckman, J. (1979). “Sample selection bias as a specification error”, Econometrica, 47(1), pp.
153–161.
Hipp, C., Tether, B., and Blind, K. (2003). “A Comparative Analysis of CIS-2 Data on
Innovation in the Service Sector - Empirical Results and Methodological Issues”, presented at
the International Workshop on Empirical Studies on Innovation in Europe, University of
Urbino, Italy, 1-2 December.
Hubbard, R. G. (1998). “Capital-Market Imperfections and Investment”, Journal of Economic
Literature, 36(1), pp. 193-225.
64
Inklaar, R., O’Mahony, M., and Timmer, M. (2003). “ICT and Europe’s Productivity
Performance Industry-level Growth Account Comparisons with the United States”, GGDC
Research Memorandum, GD-68.
Islam, N. (1999). “International Comparison of Total Factor Productivity: A Review” Review
of Income and Wealth, 45(4), pp. 493-518.
Jaffe, A. B., Trajtenberg, M., and Fogarty, M. S. (2000). “The Meaning of Patent Citations:
Report on the NBER/Case Western Reserve Survey of Patentees”, NBER Working Paper,
7631.
Jorgenson, D. W. (1988). “Productivity and Postwar U.S. Economic Growth”, Journal of
Economic Perspectives, 2(4): 23-41.
Josephson, C., and Schön, L. (2002). “Determinants of total factor productivity and phases in
economic growth: Swedish manufacturing industry 1950-1994”, presented at the Congress of
the International Economic History Association: Session Technology and Human Capital in
Historical Perspective at the XIII, Buenos Aires, Argentina, 22-26 July.
Jovanovic, B., and Nyarko, Y. (1996). “Learning by Doing and the Choice of Technology”,
Econometrica, 64 (6), pp. 1299-1310.
Karaca-Mandic, P., and Train, K. (2003). “Standard error correction in two-stage estimation
with nested samples”, The Econometrics Journal, 6(2), pp. 401-407.
Kline, S. J., and Rosenberg N. (1986). “An Overview of Innovation”, in Landau, R., and
Rosenberg, N. (eds.), The Positive Sum Strategy: Harnessing Technology for Economic
Growth, pp. 275-304. Washington D.C.: National Academy Press.
Kocher, M., Luptacik, M., and Sutter, M. (2004). “Measuring productivity of research in
economics: A cross-country study using DEA”, Socio-Economic Planning Sciences
(forthcoming).
65
Kuznetsov, P., and Muraviev, A. (2001). “Ownership Structure and Firm Performance in
Russia: The Case of Blue Chips of the Stock Market”, Economics Education and Research
Consortium Working Paper Series, 01/10.
Lanjow, J. O., and Schankerman, M. (1999). “The Quality of Ideas: Measuring Innovation
with Multiple Indicators”, NBER Working Paper, 7345.
Leiponen, A., and Helfat, C. (2003). “Innovation Objectives, Knowledge Sources, and the
Benefits of Breadth”; presented at the conference What do We Know about Innovation? A
Conference in Honour of Keith Pavitt; University of Sussex, Brighton, U.K, 13-15 November.
Lehr, B., and Lichtenberg, F. (1999). “Information Technology and Its Impact on
Productivity: Firm-Level Evidence from Government and Private Data Sources, 1977-1993”,
Canadian Journal of Economics, 32(2), pp. 335-62.
Leonard-Barton, D. (1988). “Implementation as mutual adaptation of technology and
organization”, Research Policy, 17(5), pp. 251-267.
Leonard-Barton, D. (1992). “Core capabilities and core rigidities: A paradox in managing new
product development”, Strategic Management Journal, 13, pp. 111-125.
Leung, D. (2004). “The Effect of Adjustment Costs and Organizational Change on
Productivity in Canada: Evidence from Aggregate Data”, Bank of Canada Working Papers,
04-1.
Licandro, O., Maroto, R., and Puch, L. (2004). “Innovation, Investment and Productivity:
Evidence from Spanish Firms”, EUI Working Paper, ECO2004/7.
Lööf, H. and Heshmati, A. (2004). “On the relationship between innovation and performance:
a sensitivity analysis”, Economics of Innovation and New Technology (forthcoming).
Loveman, G. W. (1994). “An Assessment of the Productivity Impact of Information
Technologies”, in Allen T.J., and Morton, M. S. (eds.), Information Technology and the
Corporation of the 1990s: Research Studies, pp. 81-117. Cambridge: MIT Press.
66
McFadden, D. (1999). “Economics 240b”, Department of Economics Lecture Notes,
University of California.
Mohnen, P., and Dagenais, M. (2002). “Towards an Innovation Intensity Index: The Case of
CIS 1 in Denmark and Ireland”, in Kleinknecht, A., and Mohnen, P. (eds.), Innovation and
firm performance: Econometric explorations of survey data, pp. 3-30. Hampshire and New
York: Palgrave.
Mohnen, P., and Röller, L. (2001). “Complementarities in Innovation Policy”, CIRANO
Working Papers, 2001s-28.
Mulkay, B., Hall, B. H., and Mairesse, J. (2000). “Firm Level Investment and R&D in France
and the United States: A Comparison”, NBER Working Paper, 8038.
Murphy, K. M., and Topel, R. H. (1985). “Estimation and Inference in Two-Step Econometric
Models”, Journal of Business and Economic Statistics, 3, pp. 370-379.
Newey, W., and McFadden, D. (1994). “Large sample estimation and hypothesis testing”, in
Griliches, Z., and Intriligator, M. (eds.), The Handbook of Econometrics, Volume 4, pp. 21112245. Amsterdam: Elsevier Science Publishers.
Nicoletti, G., and Scarpetta, S. (2003). “Regulation, Productivity and Growth: OECD
Evidence”, OECD Economics Department Working Paper, 347.
OECD (1992, 1996). “Proposed Guidelines for Collecting and Interpreting Technology
Innovation Data - Oslo Manual”, Paris: OECD.
Oliner, S., and Sichel E. (1994). “Computers and Output Growth Revisited: How Big is the
Puzzle?”, Brookings Papers on Economic Activity, 2, pp. 273-334.
Oliner, S., and Sichel, E. (2000). “The Resurgence of Growth in the Late 1990s: Is
Information Technology the Story ?”, Journal of Economic Perspectives, 14(4), pp. 3-22.
67
Paganetto, L., Becchetti, L., and Bedoya, D. (2003). “ICT investment, productivity and
efficiency: evidence at firm level using a stochastic frontier approach”, Journal of
Productivity Analysis (forthcoming).
Parisi, M., Schiantarelli, F., and Sembenelli, A. (2002). “Productivity, Innovation Creation
and Absorption, and R&D: Micro Evidence for Italy”, Boston College Working Papers in
Economics, 526.
Pavitt, K. (1982). “R&D, Patenting and Innovative Activities. A Statistical Exploration”,
Research Policy, 11(1), pp. 33-51.
Power, L. (1998). “The Missing Link: Technology, Investment and Productivity”, Review of
Economics and Statistics, 80(2), pp. 300-313.
Roach, S. (1989). “America's White-Collar Productivity Dilemma”, Manufacturing
Engineering, 104.
Rogers, M. (1998). “The Definition and Measurement of Productivity”, Melbourne Institute
Working Paper, 9/98.
Sakellaris, P. (2001). “Patterns of Plant Adjustment”, Board of Governors of the Federal
Reserve System Finance and Economics Discussion Series, 2001-05.
Sichel, A. (1997). The Computer Revolution: An Economic Perspective. Washington, DC:
Brookings Institution Press.
Singh, N., and Trieu, H. (1996). “The Role of R&D in Explaining Total Factor Productivity
Growth in Japan, South Korea and Taiwan” UCSC Dept of Economics Working Paper, 362.
Sirilli, G. (2002). “Trends and Opportunities for R&D and Innovation Surveys”, in
Conceição, P., Gibson, D., Heitor, M., Sirilli, G., and Veloso, F. (eds.), Knowledge for
Inclusive Development, pp. 287-300. Westport: Quorum Books.
StataCorp (2003). Stata Base Reference Manual. College Station, TX: Stata Corporation.
68
Steindel, C., and Stiroh, K. J. (2001). “Productivity: What Is It, and Why Do We Care about
It?”, Business Economics, 36(4), pp. 13-31.
Stoneman, P. (2002). The Economics of Technological Diffusion. Oxford: Blackwell
Publishers.
Strassmann, P. (1990). The Business Value of Computers: An Executive's Guide. New
Canaan, CT: Information Economics Press.
Syverson, C. (2004). “Market Structure and Productivity: a Concrete Example”, NBER
Working Paper, 10501.
Tether, B., Miles, I., Blind, K., Hipp, C., DeLiso, N., and Cainelli, G. (2001). Innovation in
Services – An Analysis of CIS-2 data on Innovation in the Service Sector. Report for the
European Commission (under CIS Contract 98/184).
Timmer, M., Ypma, G., and Van Ark, B. (2003). “IT in the European Union: Driving
Productivity Divergence?”, Groningen Growth and Development Centre Research
Memorandum, GD67.
Tripsas, M., and Gavetti, G. (2000). “Capabilities, cognition, and inertia: Evidence from
Digital Imaging”, Strategic Management Journal, 21, pp. 1147-1161.
Tybout, J. R. (2000). “Manufacturing Firms in Developing Countries: How Well do They Do
and Why?”, Journal of Economic Literature, 38(1), pp. 11-44.
Utterback, J. (1994). Mastering the dynamics of innovation. Cambridge: Harvard Business
School Press.
Van Leeuwen, G., and Van der Wiel, H. (2003). “ICT, Innovation and Productivity”,
presented at the Comparative Analysis of Enterprise (micro) Data (CAED) Conference,
London, UK, 15-16 September.
69
Vivero, R. (2002). “The impact of process innovations on firm's productivity growth: the case
of Spain”, Applied Economics, 34 (8), pp. 1007-1116.
Womack, K., Jones, D., and Roos, D. (1989). The Machine that Changed the World.
Cambridge, MA: MIT Press.
Wolff, E. (2002). “Productivity, Computerization, and Skill Change”, Economic Review,
87(3), pp. 63-87.
Wooldridge, J. (2003). Introductory Econometrics. Manson:, Thomson, South-Western.
Young, A. (1991). “Learning by doing and the dynamic effects of international trade”,
Quarterly Journal of Economics, 106(2), pp. 369-405.
Young, A. (1993). “Invention and bounded learning by doing”, Journal of Political Economy,
101(3), pp. 443-472.
Zagler, M. (2002). “Innovations, Economic Growth and Productivity in the New Economy”,
in Huebner, K. (ed.), The Regional Divide, Promises and Realities of the New Economy in a
Transatlantic Perspective., London and New York: Routledge (forthcoming).
70
Annex I – The Community Innovation Survey
http://www.oces.mces.pt/docs//doc36/lang1/CISIII.pdf
71
Inquérito Comunitário à Inovação(CIS III)
Questionário relativo ao processo de inovação na empresa referente a 1998-2000
MINISTÉRIO DA CIÊNCIA E DA TECNOLOGIA
OBSERVATÓRIO DAS CIÊNCIAS E DAS TECNOLOGIAS
Terceiro Inquérito Comunitário à Inovação
INQUÉRITO DO SISTEMA ESTATÍSTICO NACIONAL (Lei 6/89 de 15 de Abril) DE RESPOSTA OBRIGATÓRIA, REGISTADO NO INE SOB O Nº 9328, VÁLIDO ATÉ 31/12/2001
Este question rio constitu do por um conjunto de quest es relativas introdu o, ou a actividades ligadas introdu o, de produtos e processos novos ou significativamente melhorados na ind stria e nos servi os no per odo
1998-2000. fundamental que todas as empresas respondam s perguntas indicadas, independentemente de terem
introduzido ou n o inova es. S assim se poder o comparar n veis de inova o das empresas portuguesas com as
empresas dos outros pa ses comunit rios.
Agrade emos que leia calmamente o question rio pergunta a pergunta antes do respectivo preenchimento.
Para esclarecimentos contactar:
>
ISABEL SOUDO
Telefone:
Fax:
E-mail:
21 8452090
21 8463432
[email protected]
A preencher pela empresa:
Nome do Responsável pelo preenchimento _________________________
Função na Empresa _____________________________________________
Telefone _______________________________________________________
Fax ___________________________________________________________
E-mail _________________________________________________________
Notas importantes de preenchimento
1. Todos os campos de preenchimento com n meros devem ser preenchidos colocando os algarismos da direita
para a esquerda deixando em branco os espa os que ficarem livres.
126
ex.
2. Os valores monet rios podem ser dados quer em contos, quer em euros, devendo, no entanto, utilizar-se a
mesma unidade monetária ao longo de todo o questionário. Sempre que a pergunta exige uma resposta com
valores monet rios tal indicado atrav s da coloca o das palavras "contos" e "euros" a seguir ao campo de
preenchimento, devendo riscar-se a unidade monetária que não se utiliza. Indique, em seguida, a unidade
monet ria escolhida:
contos
euros
Legislação
O OCT (Observat rio das Ci ncias e das Tecnologias)
rg o delegado do INE para a rea estat stica da Ci ncia e da
Tecnologia (Despacho Ministerial Conjunto 265/97 de 31 de Julho), passando a integrar o Sistema Estat stico
Nacional (SEN) e sujeitando-se como tal legisla o que estipula o seu funcionamento (Lei 6/89 de 3 de Abril).
SEGREDO ESTATŒSTICO
Consiste no dever que impende sobre o OCT de guardar reserva absoluta em rela o informa o estat stica de
car cter individual de pessoas singulares e colectivas por ele recolhida (art… 5… da Lei 6/89).
OBRIGATORIEDADE DE RESPOSTA
obrigat ria a presta o das informa es pedidas pelos funcion rios e agentes do INE enquanto encarregados da
recolha directa de informa es estat sticas (art… 19… da Lei 6/89).
00
Informação geral sobre a empresa
Define-se empresa como uma organiza o definida juridicamente, com balan o pr prio, submetida a uma direc o
que pode ser tanto uma entidade jur dica como uma entidade f sica e constitu da com o fim de exercer, num ou
v rios locais, uma ou v rias actividades de produ o de bens e servi os.
Nome da empresa _______________________________________________________________________________
Morada ________________________________________________________________________________________
Código postal
-
Localidade _____________________Concelho ______________________Distrito _________________________
Actividade Principal (CAE Rev. 2)
0.1
Número de pessoa colectiva
A sua empresa é parte de um grupo de empresas?
Sim
Em que país se localiza a sua sede? ______________________
Não
0.1.1
Qual o ano de fundação da empresa (no nosso país)?
0.2
Entre 1998 e 2000 ocorreu na sua empresa alguma das seguintes mudanças significativas?
Sim
Não
Aumento do volume de vendas em 10 % ou mais devido a fusão com outra empresa
Redução do volume de vendas em 10 % ou mais devido à venda ou encerramento de
parte da empresa
0.3
Indique o tempo médio de vida do produto (bem ou serviço) mais importante para a sua empresa antes
de ser substituído ou significativamente melhorado:
Menos de 1 ano
0.4
1-3 anos
4-6 anos
7-9 anos
Mais de 9 anos
Qual o mercado geográfico mais importante para a empresa? Escolher a alternativa mais apropriada
Local/ regional (até uma distância de cerca de 50 km) em Portugal
Local/ regional (até uma distância de cerca de 50 km) mas incluindo Espanha
Nacional (para além de 50 km)
Internacional (para além de 50 km)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Impossível dizer
Informação económica sobre a empresa
0.5
Volume de negócios 1
vendas de bens e serviços
em 1998
em 2000
em contos ou euros
em contos ou euros
em 1998
em 2000
em contos ou euros
em contos ou euros
em 1998
em 2000
em contos ou euros
em contos ou euros
em 1998
em 2000
em contos ou euros
em contos ou euros
em 1998
em 2000
em contos ou euros
em contos ou euros
em 1998
em 2000
em 1998
em 2000
em 1998
em 2000
em 1998
em 2000
(incluido exportações e excluindo apenas o I.V.A.)
0.6
0.7
Volume de exportações (apenas)
Investimento bruto em capital fixo2
(excluíndo o I.V.A.)
0.8
Valor acrescentado bruto 3
0.9
Custos com o
0.10
Características do pessoal ao serviço5
pessoal 4
0.10.1
Total de pessoal ao serviço
0.10.2
Pessoal ao serviço que completou ensino superior
Escolaridade
(concluiram cursos universitários ou politécnicos)
0.10.3
Pessoal ao serviço que completou o 12.º ano
0.10.4
Qualifica o/Fun es
Pessoal ao serviço cuja qualificação corresponde a:
quadros, profissionais altamente qualificados e
profissionais qualificados
1
2
3
4
5
6
7
8
1
2
3
4
5
Para instituições de crédito: receitas de juros e similares; para seguradoras: receitas brutas de prémios recebidos.
Aquisição de equipamentos, ter renos e construções.
Obtido pela diferença entre as vendas e o consumo intermédio, isto é, os bens e serviços consumidos pela empresa no seu processo produtivo.
Inclui todas as despesas efectuadas que revertem a favor do pessoal ao serviço.
Média anual. Se não for possível indicar a média anual, indicar os valores para o final de cada ano. O pessoal ao serviço inclui as pessoas que, no período de referência, participaram na actividade da empresa qualquer que tenha sido a duração dessa participação, nas
seguintes condições: a) pessoal ligado à empresa por um contrato de trabalho, recebendo em contrapartida uma remuneração; b) pessoal ligado à empresa/instituição, que por não estar vinculado por um contrato de trabalho, não recebe uma remuneração regular pelo
tempo trabalhado ou trabalho fornecido (p.ex.: proprietários-gerentes, familiares não remunerados, membros activos de cooperativas);
c) pessoal com vínculo a outras empresas/instituições que trabalharam na empresa/instituição sendo por esta directamente remunerados.
9
10
11
12
13
14
15
16
Inovação
A inova o corresponde, no mbito deste question rio, introdu o no mercado de um produto (bem ou servi o)
novo ou significativamente melhorado, ou introdu o por parte da empresa de processos novos ou significativamente melhorados. A inova o pode ser baseada em novos desenvolvimentos tecnol gicos, em novas combina es
de tecnologias existentes, ou na utiliza o de outro tipo de conhecimento adquirido pela empresa.
Atenção: ler “Anexo” sobre inovação antes de continuar a preencher o questionário.
01
Inovação de produto
A inova o de produto corresponde introdu o no mercado de um produto (bem ou servi o) novo ou significativamente melhorado relativamente s suas caracter sticas fundamentais, s suas especifica es t cnicas, ao software
ou outros componentes imateriais incorporados, s utiliza es para que foi concebido, ou facilidade de utiliza o.
A inova o tem que ser nova para a empresa; n o tem que ser necessariamente nova no mercado servido pela
empresa. A inova o pode ter sido desenvolvida tanto pela empresa como fora dela. Modifica es de natureza unicamente est tica e a mera venda de inova es totalmente produzidas e desenvolvidas por outras empresas n o se
podem considerar inova es.
Em “Anexo” apresentam-se exemplos de inovações.
1.1
Durante o período de 1998-2000, a sua empresa introduziu no mercado algum produto (bem ou serviço)
novo ou significativamente melhorado no contexto da empresa?
Sim
Quem desenvolveu esses produtos?
Indique apenas a alternativa mais adequada
Principalmente a sua empresa ou grupo a que pertence
A sua empresa em cooperação com outras empresas ou instituições
Principalmente outras empresas ou instituições
Não
1.2
passar para a questão 2 (página seguinte)
Por favor, faça uma descrição sucinta do mais importante produto (bem ou serviço) novo ou significativamente melhorado (por favor escreva com letra de imprensa).
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
1.3
Por favor estime a distribuição da percentagem de volume de negócios6 de 2000 entre:
Vendas de produtos novos ou significativamente melhorados introduzidos no período de
1998–2000.
%
Vendas de produtos não modificados ou apenas marginalmente melhorados e introduzidos
no período de 1998-2000 7.
Volume de Negócios Total em 2000
1
2
3
4
1.4
%
1 0 0%
Durante o período de 1998-2000, a sua empresa introduziu no mercado algum produto (bem ou serviço)
novo ou significativamente melhorado no contexto do mercado servido pela empresa?
Sim
Não
Contribuição da venda destes produtos para o volume de negócios em 2000:
5
6
7
8
9
10
11
12
13
14
15
16
6
7
Para instituições de crédito: receitas de juros e similares; para seguradoras: receitas brutas de prémios recebidos.
Bens e serviços totalmente desenvolvidos e produzidos por entidades terceiras devem ser incluídos nesta rubrica.
%
02
Inovação de processo
A inova o de processo corresponde adop o de m todos de produ o novos ou significativamente melhorados,
assim como de meios novos ou significativamente melhorados de fornecimento de servi os e de distribui o de produtos. O resultado da inova o de processo ter que ter um impacte significativo na produ o, qualidade dos produtos (bens ou servi os) ou custos de produ o e de distribui o.
A inova o tem que ser nova para a empresa; n o tem que ser necessariamente nova no mercado servido pela
empresa. A inova o pode ter sido desenvolvida tanto pela empresa como fora dela. Modifica es de natureza unicamente organizacional ou de gest o n o se podem considerar inova es.
Em “Anexo” apresentam-se exemplos de inovações.
2.1
Durante o período de 1998-2000, a sua empresa adoptou processos de produção novos ou significativamente melhorados, incluindo meios de fornecimento de serviços ou de distribuição de produtos?
Sim
Quem desenvolveu esses processos?
Indique apenas a alternativa mais adequada
Principalmente a sua empresa ou grupo a que pertence
A sua empresa em cooperação com outras empresas ou instituições
Principalmente outras empresas ou instituições
Não
2.2
passar para a questão 3
Por favor, faça uma descrição sucinta do mais importante processo novo ou significativamente melhorado
(por favor escreva com letra de imprensa).
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
03
Actividades de inovação incompletas ou abandonadas
3.1
Até ao fim de 2000, a sua empresa desenvolveu, mas ainda não concluíu, projectos orientados para o
desenvolvimento ou introdução de produtos (bens ou serviços) ou processos novos ou significativamente
melhorados, incluindo actividades de investigação e desenvolvimento8 (I&D)?
Sim
Não
3.2
Durante o período 1998-2000, a empresa abandonou actividades em curso orientadas para o desenvolvimento ou introdução de produtos (bens ou serviços) ou processos novos ou significativamente melhorados, incluindo actividades de investigação e desenvolvimento(I&D)?
Sim
1
Não
2
ATENÇÃO:
Empresas que responderam não ao conjunto das questões 1.1, 2.1, 3.1 e 3.2, devem passar para a questão 10.1.2 (pág. 13)
3
4
5
6
7
8
8
A I&D na empresa compreende todo o trabalho criativo empreendido numa base sistemática com vista a aumentar a reserva de
conhecimentos da empresa, assim como a utilização dessa reserva no desenvolvimento de novas aplicações, tais como produtos
(bens/serviços) ou processos novos ou significativamente melhorados (incluindo investigação em software).
9
10
11
12
13
14
15
16
04
Despesa em actividades orientadas para a inovação em 2000
4.1
A empresa esteve envolvida nas seguintes actividades de inovação em 2000?
Por favor indique, assinalando "sim", se a sua empresa esteve envolvida durante 2000 nas seguintes actividades orientadas para a introdução
de produtos (bens/serviços) ou processos novos ou significativamente melhorados baseados em ciência, tecnologia ou outras áreas de saber.
Subsequentemente, estime as despesas correspondentes em 2000, incluindo as despesas associadas a actividades abandonadas ou não concluídas. Assinale "não" para as actividades em que a empresa durante 2000 não esteve envolvida. Se sim, por favor estime a despesa em 2000,
incluindo despesas com pessoal e investimento (sem depreciação) – em contos ou euros.
Sim Não
Investigação e desenvolvimento realizados na
empresa
(I&D interna)
A I&D na empresa compreende todo o trabalho criativo
empreendido numa base sistemática com vista a aumentar a reserva de conhecimentos da empresa, assim como
a utilização dessa reserva no desenvolvimento de novas
aplicações, tais como produtos (bens/serviços) ou processos novos ou significativamente melhorados (incluindo
investigação em software).
Aquisição de serviços de
I&D (I&D externa)
As mesmas actividades mencionadas acima, mas executadas por outras empresas (mesmo que sejam do grupo
da sua empresa) ou por entidades públicas ou privadas de
I&D.
Aquisição de maquinaria Maquinaria avançada, hardware ou outros equipamentos
e de equipamento
ligados especificamente a produtos (bens/serviços) ou
processos novos ou significativamente melhorados.
Aquisição de outros con - Aquisição de conhecimento externo, sob a forma de
hecimentos externos
patentes, licenças, know-how, marcas, software e outros
tipos de conhecimento externo para implementar as inovações da sua empresa
Formação
Formação interna ou externa especificamente orientada
para o desenvolvimento ou introdução de inovações
Introdução de inovações Actividades de marketing internas ou externas à empresa
no mercado (marketing) directamente orientadas para a introdução no mercado
dos produtos (bens/serviços) novos ou significativamente
melhorados (pode incluir estudos de mercado, testes de
mercado, publicidade de lançamento; deve excluir a constituição de redes de distribuição para comercializar as
inovações)
1
Projecto industrial e
outros tipos de
preparação para a produção ou distribuição de
inovações
Outros procedimentos e preparações técnicas não contemplados acima, necessários para a introdução de produtos (bens/serviços) ou processos novos ou significativamente melhorados
em contos ou euros
em contos ou euros
em contos ou euros
em contos ou euros
em contos ou euros
em contos ou euros
em contos ou euros
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Despesa total em inovação em 2000
em contos ou euros
05
Investigação e desenvolvimento realizados na empresa (I&D)
Se a sua empresa teve actividades de I&D internas:
5.1
Qual o pessoal ao serviço na empresa que esteve afecto a investigação e desenvolvimento em 2000?
(inclui tanto pessoas do departamento de investigação e desenvolvimento como fora dele, desde que envolvidas em investigação e desenvolvimento)
em total de ETI (ver definição e exemplo em seguida):
,
Definição e Exemplo:
ETI: "equivalentes a tempo integral"; calculam-se a partir da fracção (calculada em %) do tempo que cada indivíduo dedicou a actividades de I&D na empresa; o total resulta do somatório das fracções de cada pessoa.
As actividades de I&D da empresa foram desenvolvidas por pessoal a tempo integral e a tempo parcial.
Como proceder ao cálculo do Equivalente a Tempo Integral (ETI)?
Investigadores :
Um indivíduo A ocupa-se a 100% em actividades de I&D durante todo o ano na Empresa - Tempo Integral
Um indivíduo B ocupa-se a 100% em actividades de I&D durante 6 meses (1/2 ano) na Empresa - Tempo Parcial
Um indivíduo C ocupa-se a 25% em actividades de I&D durante todo o ano na Empresa - Tempo Parcial
Um indivíduo D ocupa-se a 30% em actividades de I&D durante 4 meses (1/3 ano) na Empresa - Tempo Parcial
Indivíduo
Percentagem de
tempo em I&D
Percentagem de
tempo em I&D no ano
Tempo Integral
Tempo Parcial
ETI
100%
100% x 1ano = 100%
1
-
1,0
100%
100% x 1/2ano =50%
-
1
0,5
25%
25% x 1ano = 25%
-
1
0,25
30%
30% x 1/3ano =10%
-
1
0,1
1
3
1,85
A
B
C
D
total
O total de ETI deve ser arredondado para uma casa decimal. Assim, o preenchimento do campo no caso do
exemplo será:
em total de ETI
5.2
1,9
No período de 1998-2000, de que forma se desenrolaram as actividades de I&D na empresa?
Continuadamente
Ocasionalmente
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
06
Efeitos das inovações introduzidas na empresa durante 1998-2000
A inova o pode ter v rios efeitos nas actividades da empresa. Indique, para as v rias alternativas em seguida, o grau
de impacte verificado no fim de 2000 fruto das inova es introduzidas no per odo de 1998-2000.
Grau de impacte
Alto
Efeitos associados
aos produtos
Médio
Baixo Irrelevante
Alargamento da gama de produtos
(bens/serviços)
Entrada em novos mercados ou aumento
da quota de mercado
Melhoria da qualidade dos produtos
(bens/serviços)
Efeitos associados
Melhoria da flexibilidade de produção
aos processos
Aumento da capacidade de produção
Redução dos custos de trabalho por unidade
produzida
Redução do consumo de energia e de materiais
por unidade produzida
Outros efeitos
Melhoria do impacte ambiental ou de outros
aspectos associados à segurança ou saúde
Cumprimento com regulamentações e normas
07
Financiamento público da inovação
O financiamento p blico inclui apoio financeiro sob a forma de subs dios ou empr stimos, assim como de garantias
banc rias. As vendas a entidades p blicas n o devem ser consideradas.
7.1
No período de 1998-2000 a sua empresa recebeu algum tipo de apoio financeiro público para apoiar actividades orientadas para a inovação?
Sim
Apoio de:
Não
Autoridades locais ou regionais
Governo
União Europeia
7.2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
A sua empresa recebeu financiamentos quer do 4º (1994-98) quer do 5º (1998-2002) Programa Quadro da
União Europeia para investigação e desenvolvimento?
Sim
Não
08
Cooperação na área da inovação entre 1998-2000
A coopera o na rea da inova o significa a participa o activa em actividades de I&D e em outras actividades de
inova o com outras organiza es (tanto empresas como outras entidades). A coopera o n o implica que ambos
os parceiros retirem benef cios comerciais imediatos. A simples contrata o ao exterior da empresa, sem qualquer
colabora o activa da empresa, n o considerada coopera o.
8.1
A sua empresa estabeleceu algum acordo de cooperação para actividades de inovação com outras empresas ou instituições durante o período 1998-2000?
Sim
Não
8.2
passar para a questão 9 (página seguinte)
Por favor indique o tipo de organização com quem colaborou e respectivo país ou região de origem
Admitem-se várias respostas
Tipo de parceiros
Nacional UE*/ EFTA** UE-PC***
EUA
Japão
Outra
Outras empresas do grupo
Fornecedores de equipamento,
de materiais, de componentes ou de software
Clientes
Concorrentes
Consultores
Laboratórios comerciais ou empresas de I&D
Universidades ou outras instituições
de ensino superior
Laboratórios do Estado, institutos de I&D
governamentais ou instituições
privadas sem fins lucrativos
8.3
Por favor indique a importância dos parceiros para o desenvolvimento de actividades de inovação
Tipo de parceiros
Alta
Média Baixa
Nenhum parceiro
Outras empresas do grupo
Fornecedores de equipamento,
de materiais, de componentes ou de software
Clientes
Concorrentes
Consultores
Laboratórios comerciais ou empresas de I&D
Universidades ou outras instituições de ensino superior
Laboratórios do Estado, institutos de I&D governamentais
ou instituições privadas sem fins lucrativos
1
2
3
4
5
6
7
8
*
União Europeia (Bélgica, Dinamarca, Alemanha, Grécia, Espanha, França, Irlanda, Itália, Luxemburgo, Holanda, Áustria, Portugal,
Finlândia, Suécia e Reino Unido)
**
EFTA- European Free Trade Association (Islândia, Liechtenstein, Noruega, Suiça)
*** UE Países Candidatos (Bulgária, Chipre, República Checa, Estónia, Hungria, Letónia, Lituânia, Malta, Polónia, Roménia, Eslováquia,
Eslovénia e Turquia)
9
10
11
12
13
14
15
16
09
Fontes de informação para a inovação entre 1998-2000
Esta quest o diz respeito identifica o das principais fontes de informa o das quais resultaram sugest es para
projectos de inova o ou que contribuiram para a implementa o de inova es. Por favor indique a import ncia
atribu da s diferentes fontes de informa o mencionadas em seguida.
Se utilizada, importância
Fonte de Informação:
Fontes internas
Alta
Dentro da própria empresa
Outras empresas do grupo
Fontes de mercado
Fornecedores de equipamento,
de materiais, de componentes ou
de software
Clientes
Concorrentes
Fontes
institucionais
Universidades ou outras instituições de
ensino superior
Laboratórios do Estado, institutos de I&D
governamentais ou instituições privadas
sem fins lucrativos
Outras fontes
Conferências, reuniões e publicações
científicas ou profissionais
Feiras, mostras de produtos
Empresas de consultoria
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Média Baixa
Não utilizada
10
Factores que dificultam a inovação
10.1.1
Durante o período de 1998-2000, houve actividades orientadas para a inovação que:
Sim
Não
foram seriamente atrasadas?
nem sequer foram iniciadas?
Passar para a questão 10.2
foram canceladas?
10.1.2
Ausência de actividades orientadas para a inovação
(Apenas para Empresas que responderam não ao conjunto das questões 1.1, 2.1, 3.1 e 3.2. O questionário deve ser preenchido até ao fim,
incluíndo a questão 10.2 )
Durante o período de 1998-2000, alguma das razões seguintes foi relevante para que a empresa não
tivesse tido quaisquer actividades orientadas para a inovação?
Sim
Não
Não se justificavam actividades orientadas para inovação,
dado que havia inovações introduzidas anteriormente
Não se justificavam actividades orientadas para inovação,
dadas as condições do mercado da empresa
Existiram factores que dificultaram a inovação
10.2
Factores que dificultaram a inovação
Se a empresa sentiu dificuldades no desenvolvimento de actividades de inovação ou nem sequer as iniciou
entre 1998-2000, por favor indique a importância de cada um dos factores de impedimento.
Factores de impedimento:
Factores
económicos
Grau de importância
Alto Médio Baixo
Não relevante
Percepção de riscos económicos excessivos
Custos de inovação demasiado elevados
Falta de fontes de financiamento apropriadas
Factores
internos
Estrutura organizacional pouco flexível
Falta de pessoal qualificado
Falta de informação sobre tecnologia
Falta de informação sobre mercados
Outros
factores
Regulamentação e normas
Falta de receptividade dos clientes às inovações
Reduzida dimensão do mercado
1
2
ATENÇÃO:
Todas as empresas devem responder às questões 11 e 12 da página seguinte.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
11
Patentes e outros métodos de protecção
11.1.1
Durante o período de 1998-2000, a sua empresa, ou alguma outra empresa do mesmo grupo submeteu
pedidos de patentes para proteger invenções ou inovações por elas desenvolvidas?
Total
Bens /serviços/processos
Sim
Não
11.1.2
Por favor indique o número de pedidos9
A sua empresa, ou outra empresa do mesmo grupo, tinha patentes válidas no fim de 2000 para proteger
invenções ou inovações por elas desenvolvidas?
Total
Bens /serviços/processos
Sim
Não
11.1.3
Dos quais:
Apenas bens /serviços
Por favor indique o número de patentes válidas9
Que percentagem do volume de negócios, em 2000, estava protegido por patentes ou patentes pedidas?
Proporção das vendas 2000
11.2
Dos quais:
Apenas bens /serviços
%
Impossível responder
Durante o período 1998-2000, a sua empresa, ou alguma outra empresa do mesmo grupo, recorreu aos
seguintes métodos para proteger invenções ou inovações por elas desenvolvidas?
Sim
Não
Registo de padrões de design
Métodos formais
Marcas Registadas (Trademarks)
Direitos de Autor (Copyright)
Métodos estratégicos
Segredo
Complexidade de concepção
Antecipação face aos concorrentes na introdução da inovação
12
Outras mudanças estratégicas e organizacionais importantes
At esta altura, este question rio tem-se debru ado sobre produtos (bens ou servi os) ou processos novos ou significativamente melhorados. Esta ltima quest o est associada a outros melhoramentos criativos que podem ter sido
desenvolvidos pela empresa.
12.1
Durante o período de 1998-2000, a empresa desenvolveu alguma das seguintes actividades?
Sim
Estratégia
1
2
3
Gestão
Implementação de técnicas de gestão avançadas por parte da empresa
Organização
Implementação de estruturas organizacionais novas ou
significativamente alteradas
Marketing
Mudanças significativas nas estratégias ou conceitos
de marketing da empresa
Mudanças estéticas
Mudanças significativas de carácter estético, de design ou de outro tipo
com carácter subjectivo em pelo menos um produto
4
5
6
(ou outras de
carácter subjectivo)
Não
Implementação de estratégias novas ou significativamente alteradas
7
8
9
10
11
12
13
14
15
16
9
Pedidos de patentes ou patentes concedidas às mesmas invenções em países diferentes devem contar como a mesma (uma única)
patente.
Muito obrigado por ter disponibilizado o seu tempo e colaborado com o Observat rio
das Ci ncias e das Tecnologias (Minist rio da Ci ncia e da Tecnologia).
Agradecemos a devolu o do question rio preenchido, utilizando o envelope de
resposta sem franquia (RSF) junto enviado.
Inquérito Comunitário à Inovação(CIS III)
Questionário relativo ao processo de inovação na empresa referente a 1998-2000
MINISTÉRIO DA CIÊNCIA E DA TECNOLOGIA
OBSERVATÓRIO DAS CIÊNCIAS E DAS TECNOLOGIAS
Anexo
INQUÉRITO COMUNITÁRIO À INOVAÇÃO (CIS III)
MINISTÉRIO DA CIÊNCIA E DA TECNOLOGIA
OBSERVATÓRIO DAS CIÊNCIAS E DAS TECNOLOGIAS
EXEMPLOS DE INOVAÇÃO
Considera-se que a inovação, de acordo com a definição do questionário, pode ser de dois tipos:
inovação de Produto (bens ou serviços) e inovação de Processo.
1 – Inovação de Produto (bens ou serviços), permitindo, designadamente:
um melhor desempenho do produto ou do serviço;
um alargamento das possíveis aplicações do produto ou do serviço.
Exemplos: alteração do tipo de materiais utilizados, introdução de produtos ecológicos numa gama de artigos, incorporação de
"chips" electrónicos, utilização de sistemas de cartão de cliente, recurso a serviços de atendimento telefónico ao cliente, desenvolvimento de actividades bancárias e de seguros electronicamente, utilização de serviços de internet e de comércio electrónico
(embora a mera criação de um site de informação sem serviços on-line não se considere uma inovação).
2 – Inovação de Processo, que se pode manifestar na melhoria do desempenho:
do próprio processo, levando a que:
os processos se tornem melhor integrados ou mais automatizados;
aumente a flexibilidade;
melhore a qualidade;
melhore a segurança ou se reduzam os danos ambientais.
Exemplos: melhorias na logística de armazenagem (como, por exemplo, sistemas de order picking), seguimento e localização de
expedições (tracking and tracing), interligação entre comunicação de dados e transporte de mercadorias, sistemas de código de
barras, processamento óptico da informação, sistemas dedicados (expert systems), software para integração de funções, primeira
utilização de ferramentas CAD/CAE.
Considera-se que a certificação ISO é uma inovação apenas quando está directamente relacionada com a introdução de processos
novos ou melhorados.
dos procedimentos de logística e controlo, permitindo que:
melhore o planeamento e a rota de mercadorias;
aumente a flexibilidade na distribuição;
melhore o controlo de stocks.
Exemplos: sistemas de automatização de pedidos/compras, sistemas de minimização de stocks (just-in-time), sistemas auxiliares
computadorizados para logística.
EXEMPLOS MAIS ESPECÍFICOS DE INOVAÇÃO POR SECTOR
Indústria
Produto
inclusão de produtos ecológicos na gama de produtos existente
garantia perpétua em produtos novos ou usados
alteração de materiais em artigos, por exemplo a "respiração activa" em artigos de vestuário
roupa à prova de água
módulos para a área da ciência da vida produzidos através de engenharia biológica
introdução em produtos de chips electrónicos
uso de telemática em veículos a motor
veículos a motor com redução da emissão de poluentes (ex. carros com baixo consumo, autocarros movidos a gás natural)
programa de estabilização electrónica para veículos a motor (ESP)
novo tipo de papel para impressoras específicas
novos tipos de motores em navios
linhas de alta tensão isoladas com gás
manutenção à distância
filtros de cerâmica para microondas e filtros de radiação nas comunicações móveis
Dar um novo nome ou voltar a acondicionar bens já existentes de forma a alcançar outro mercado não se considera
uma inovação.
Processo
digitalização de processos de impressão
novos tipos de sistemas de lâminas para produção de aparas de madeira (wood chips)
novo modelo de unidade de remoção e recuperação de água
medição de partículas por sensores em exaustão de gases
aplicação em série de "lacas" ou "verniz" em pó para o tratamento e protecção superficial de metais
novos processos de produção de ácidos com diferentes matérias-primas
sistemas de identificação e controlo
novos sistemas de CAD
novos sistemas de distribuição da informação
interligação de sistemas de processamento de dados, software para computadores em rede
introdução de métodos de assistência/auxílio e/ou baseados em computador para desenvolvimento de
produto
introdução de programas de simulação com base em elementos finitos para optimização de componentes
recurso ao comércio electrónico interligado com a produção (ex. bancos electrónicos com ofertas personalizadas, compras via internet)
disponibilização de canais directos de comunicação entre o cliente e o produtor
controlo do tempo e fase de execução na internet
Comércio por Grosso
Produto
1
2
3
4
inclusão de produtos ecológicos na gama de produtos existente
garantia perpétua em produtos novos ou usados
novos tipos de serviços de certificação
inclusão de serviços adicionais
soluções combinadas (ex. serviços técnicos e de consultoria)
teste, exame e certificação de serviços
introdução de sistemas de cartão de cliente
consulta e pedidos de compra no ponto de venda (PoS)
serviço de recolha para clientes
manutenção à distância
venda via internet (comércio electrónico); mas não somente um site de informação sem disponibilização
de serviços online
sistemas de identificação e controlo
venda directa ao cliente final
Processo
leitores ópticos nas caixas registradoras
desenvolvimento e introdução de canais de distribuição digital
computadores portáteis para vendedores, de auxílio directo às vendas
sistemas de identificação e controlo
colocação digital de rótulos ou etiquetas em produtos (ex. uso de códigos de barras)
reconstrução ou reorganização de espaços de venda de forma a proporcionar uma fácil aquisição por
parte dos clientes
recibos por PC que incluem mais informação nas facturas/vendas a dinheiro
catálogos electrónicos (ex. CD-ROM)
soluções de centro de atendimento telefónico (call-centre)
disponibilização de meios físicos (oficina, equipamento) para efectuar serviços em regime de self-service
treino de trabalhadores especializados para oferecer serviços especiais de consultoria para clientes
novos sistemas de CAD
sistemas de distribuição de informação
interligação de sistemas de processamento de dados, software para computadores em rede
disponibilização de canais directos de comunicação entre o cliente e o produtor
centros de atendimento ao cliente para coordenar todas as necessidades do cliente
Serviços Financeiros
Produto
serviços de seguros novos ou significativamente melhorados
introdução de conceitos de seguro de vida por módulos
novos seguros de invalidez ocupacional
introdução de sistemas de seguros de activos e títulos de catástrofes (cat-bonds)
introdução de cartões para acesso directo com identificação e controlo nos hospitais
Processo
bancos online
ferramentas para controlo de chamadas telefónicas
software novo ou melhorado, ou redes de computadores pessoais
aplicação de novos métodos de diversificação de risco
arquivamento óptico-electrónico de documentos
escritório livre de papel
melhoria nos sistemas fundamentais de identificação e controlo
políticas de ponto de venda
introdução de novos métodos de pontuação e classificação ("rating")
1
2
3
4
Outros Serviços
Produto
máquinas de venda de bilhetes com porta-moedas electrónico (PMB) ou cartão de pagamento de
débito/crédito
manutenção remota de software, aconselhamento remoto
novos métodos de análise estatística
desenvolvimento de software flexível para clientes
contratação de serviços nas áreas ambiental e de energia
fornecimento de novas aplicações de multimédia
novos serviços de logística
serviços de resposta por voz
serviços de atendimento telefónico ao cliente
Processo
transferência electrónica de dados
CAD ou projectos em CAD
bancos electrónicos
"caixa de ferramentas" para software específico para clientes (CASE-Tools)
criação computacional de documentos
melhoria da rede de computadores
sistemas de monitorização de redes
sistemas de gestão de chamadas telefónicas
aplicação de sistemas de visualização térmica ("termographics") para avaliar sistemas técnicos
controlo do tempo e fase de execução com base na internet
sistemas de navegação via satélite
novas ferramentas de software para gestão de fornecedores e compras
introdução de gás natural em autocarros
introdução de piso rebaixado em autocarros
1
2
3
4
Inquérito Comunitário à Inovação(CIS III)
Questionário relativo ao processo de inovação na empresa referente a 1998-2000
MINISTÉRIO DA CIÊNCIA E DA TECNOLOGIA
OBSERVATÓRIO DAS CIÊNCIAS E DAS TECNOLOGIAS
Annex II - Endogeneity and the Hausman Test
Endogeneity occurs when a model includes an endogenous explanatory variable. In others
words, when a multiple regression model includes an explanatory variable that is correlated
with the error term, either because of an omitted variable, measurement error, or simultaneity
(Wooldridge 2003). In the present study the endogeneity problem is a result from the
correlation between innovation and productivity.
As referred above and in order to confirm the theory that previews this relationship between
the two principal variables used in this study, we ran a Hausman test.
The Hausman test is defined by the Stata 8.0 Manual as: “the test that compares an estimator
θ1 that is known to be consistent with an estimator θ2 that is efficient under the assumption
being tested; the null hypothesis is that the estimator θ2 is indeed a (consistent and) efficient
estimator of the true parameters; if this is the case, there should be no systematic difference
between the two estimators; if there exists a systematic difference in the estimates, then we
have reason to doubt the assumptions on which the efficient estimator is based” (StataCorp
2003).
This test based on the vector of differences of two estimators was proposed by Hausman
(1978) is now widely used in econometric analysis in order to detect endogeneity between
variables (Greene 2000).
In the specific case of the model analyzed in this thesis we want to compare two regressions:
the regression using the data on innovation that result from a previous estimation (see the
description of the model specification) and the regression that uses the innovation variable
directly taken from the raw data (an one step OLS). From the first regression we take the
consistent estimator that makes possible the testing on the endogeneity between productivity
and innovation since it uses the predicted variable that avoids endogeneity.
After analyzing the results,
91
Table 8 - Hausman Test Stata Output
The null hyphothesis (the difference between the coefficients obtained is not systematic) is
rejected and so the results of the simple OLS estimation are not consistent. Within this
framework, we can conclude that there exists endogeneity between the two analyzed
variables.
The test was realized using a simplified version of the model and with the product innovation
variable as the innovation variable (some controlling variables were not included) because the
Hausman test presented an error when ran with the complete model. Despite this problem, we
think that these results are sufficient to conclude that, in the context of the Portuguese CIS III
data, the variables productivity and innovation present endogeneity.
92