Analysis
Innovation complementarity and environmental productivity effects:Reality or delusion?Evidence from the EU
Marianna Gilli a ,Susanna Mancinelli a ,⁎,Massimiliano Mazzanti a ,b
a University of Ferrara,Italy b
CERIS CNR,Milan,Italy
a b s t r a c t
a r t i c l e i n f o Article history:
Received 16July 2013we are family
Received in revid form 3January 2014Accepted 6April 2014
Available online 23May 2014JEL classi fication:L6O3Q55
Keywords:
时Complementarity Innovation
Climate change Sector performance
Innovation is a key element behind the achievement of desired environmental and economic performances.Regarding CO 2,mitigation strategies would require cuts in emissions of around 80–90%with respect to 1990by 2050in the EU.We investigate whether complementarity,namely integration,between the adoption of environ-mental innovation measures and other technological and organizational innovations is a factor that has supported reduction in CO 2emissions per value added,that is environmental productivity.We merge new EU innovation and WIOD data to asss the innovation effects on ctor CO 2performances at a wide EU level.We find that jointly adopting different innovations is not a widespread factor behind increas in environmental productivity.Neverthe-less,even though complementarity is not a low hanging fruit,a ca where ‘innovation complementarity ’aris is for manufacturing ctors that integrate eco-innovations with product innovations.One example of this integrated ac-tion is a strategy that pursues energy ef ficien
cy with product value enhancement.We believe that the lack of inte-grated innovation adoption behind environmental productivity performance is a signal of the current weakness economies face in tackling climate change and green economy challenges.Incremental rather than more radical strategies have predominated so far.The latter have been con fined to industrial ‘niches ’,in terms of the number of involved firms.This is probably insuf ficient when we look at long-term economic and environmental goals.
©2014Elvier B.V.All rights rerved.
1.Introduction
The ful fillment of EU strategy goals on emissions and greenhou targets chie fly depends upon the economic and technological evolution of its industrial ctors.Technological development and composition effects are pillars of sustainability in production since they both coun-terbalance the growth scale effect as the IPAT (Impact-Population-Af fluence-Technology)model shows (York et al.,2003).Long run sus-tainability targets need to undergo radical changes in the EU economy.The ctor's evolution is pivotal to the ‘greening ’of the economy,since,as the neo Schumpeterian tradition emphasizes,innovation is id-iosyncratic at a ctor level.Sector and national systems of inno
vation must both be recognized (Breschi et al.,2000).Various analys have re-cently focud on economic and environmental dynamics at a ctor level,by placing innovation at the center of their reasoning (Costantini and Crespi,2008;Costantini and Mazzanti,2012,2013;Marin and Mazzanti,2013).
Environmental innovations are a relevant part of the innovative dynamics that should support the integration of competitiveness and sustainability (Cainelli et al.,2012;De Marchi,2012;Horbach,2008;Kemp and Pontoglio,2011).We here focus on innovation rather than
invention given the importance of diffusion and adoption of innovation practices throughout the economy (Costantini and Mazzanti,2013).Patent data and invention bad analys are nevertheless an important part of the related literature,which we do not address here for reasons of conciness and space (Costantini and Crespi,2013;Dechezlepretre et al.,2011;Hafner et al.,2012;Johnstone et al.,2010).
De finitions of eco-innovation (Kemp,2000)highlight the ecological attributes of new individual process,products and methods from a technical and ecological perspective (Kemp,2010).Along the lines,the drivers of EI have been analyzed both inside and outside a firm's boundary,within the institutional and economic features of the territo-ry (Horbach et al.,2012).
Relevant to this paper,various streams of literature within the inno-vation framework have placed attention on the role of complementarity among innovation practices (Hall et al.,2012;Mancinelli and Mazzanti,2009;Mohnen and Roller,2005).Nevertheless,despite some advance-ment even in the framework of environmental innovation,the comple-mentarity hypothesis has been ldom analyzed,if at all,as a factor behind the achievement of desired economic and environmental per-formances (Antonioli et al.,2013).Complementarity is a key strategic element of a firm's organizational capabilities.It is also a somewhat irre-producible ‘not patented ’ast which nevertheless delivers appropria-ble rents (Dosi et al.,2006).
Building on the theoretical framework of Topkis (1998)and follow-ing the approaches of Milgrom and Roberts (1990,1995),we wish to
Ecological Economics 103(2014)56–67
⁎Corresponding author at:Department of Economics and Management,University of Ferrara,Via Voltapaletto 11,44121Ferrara,Italy.
E-mail address:susanna.mancinelli@unife.it (S.
Mancinelli).dx.doi/10.lecon.2014.04.0040921-8009/©2014Elvier B.V.All rights
rerved.
Contents lists available at ScienceDirect
Ecological Economics
j o u r n a l h o me p a g e :ww w.e l s e v i e r.c o m /l o c a t e /e c o l e c on
first analyze if there is complementarity between different kinds of ,product innovation,process innovation,environmental innovation)behind the reduction of CO2emissions,with a focus on en-vironmental productivity(value added on CO2)as a key indicator.We investigate whether innovation complementarities are evident for the economy as a whole,as well as for sub ctor groups,specifically manufacturing,ETS(Emission Trading System)ctors and geographi-cally divided groups(North/South EU,to test whether the innovation gaps prent in southern countries might be relevant in environmental terms).We aim to asss if regulated ctors,namely ETS ctors,adopt a greater level of environmental innovation to comply with regulation and are able to u complementarities among different kinds of innova-tion,following the hypothesis of Porter and Van der Linde(1995).Calel and Dechezleprêtre(2012)have stated that the EU-ETS has actually had effects on the increa in the introduction of environmental innovation, in this ca low-carbon innovation;however,in pha one of EU-ETS, process innovation is found to be more likely to occur with respect to product innovation.There is a high level of uncertainty nevertheless on ETS-related inducement of innovation(Borghesi et al.,2012; Cainelli and Mazzanti,2013).
This attempt is somewhat original given that literature on comple-mentarity has mainly focud on the drivers of innovation rather than its effects.Secondly,as regards performances,apart from few excep-tions(Crespi,2013),the literature about the effects of environmental in-novations on economic performance has expanded along the Porter hypothesis(Mohnen and Van Leeuwen,2013).We here take a specific and original direction by analyzing the recent effects of innovations and their complementarity on environmental productivity,which we here define as economic value on CO2(Repetto,1990).We focus on the EU economy.
To investigate the issues that revolve around the notion of comple-mentarity within innovation practices and its effects on environmental productivity,we merge data from the EU Community Innovation Survey–at the ctoral level(available at EUROSTAT website)–with data on ctoral CO2emissions(2009and2010)available from the WIOD1.We thus merge and exploit new EU ctor datats that cover ctor,environmental innovation adoption and emission performances to investigate whether innovation determines better environmental performances.Various econometric techniques are implemented to as-ss this relationship,taking into account the specific features of ETS ctors,the complementarity among various innovations and the dy-namic contents of the innovation–emission relationship at meso level. Wefirst asss the effect of innovations taken alone and their‘integrat-ed’effect with a view to complementarity.
The paper is structured as follows:Section2prents a review of the empirical literature about complementarity;Section3discuss the complementarity conceptual framework that we adopt and prents main rearch hypothes;Section4prents the empirical analysis about complementarity,discussing various econometric analys; Section5concludes.
2.Measuring Complementarity:the Relevant Literature
A relationship of complementarity between two activities implement-ed by afirm exists when the‘doing more’of‘one of them’increas the attractiveness of‘doing more’on the part of the other.Systemic effects ari,“with the whole being more than the sum of the parts”(Roberts, 2006,p.37).This has obvious implications onfirms'strategies,since a firm's efforts should be targeted toward all the complementary activi-ties.In fact,the change of just some choice variables may result ineffec-tive if other complementary variables remain unchanged.
Economic literature esntially distinguishes three methods of mea-suring complementarity(Galia and Legros,2004a,b;Mohnen and Roller,2005).Thefirst examines whether the correlation between two variables is positive and conditioned by other(exogenous)elements. In other words,one establishes whether or not empirical evidence sup-ports the hypothesis of a relationship of complem
entarity between two variables,while controlling for other parameters,but with a substantial difference compared to simple correlations which do not provide any in-formation about potential complementarity(Arora and Gambardella, 1990;Ichniowski et al.,1997).The“advantage”of this method can be found in the fact that it does not specify an objective reference variable in the analysis of productivity).Rather,it focus on the variables being examined for complementarity,which can be de-fined as the“dependents”in the model(Galia and Legros,2004b).The other two approaches in contrast treat variables which are potentially part of a relationship of complementarity as explanatory variables in an empirical model where the dependent variable is usually a perfor-mance variable(productivity,profitability,innovation).
The cond approach(the reduced form approach)analyzed by Arora (1996)is bad on the notion that if an activity of thefirm has an effect on any given objective variable,it will not be correlated to another activ-ity unless the activities are complementary.Analysis of complemen-tarity is esntially founded on an analysis of interaction/correlation between two factors,in relation to any chon dependent variable in the empirical model.The limit here is on the focus placed on only two potentially complementary variables,as Arora(1996)and Athey and Stern(1998)have highlighted.The limits lead us to the third ap-proach,which we can consider as more general in nature.
Defined in literature as the productivity approach,the third approach rembles the last and is bad on the identification of an objective var-iable defined as dependent in the regression model,with an explanatory vector which could contain discrete or continuous variables of interest, their interactions of complementarity defined in different terms,and other external control factors.Especially when dealing with discrete variables,this approach reveals to beflexible,general,and relatively simple,even when more than two activities of thefirm are being ana-lyzed.Inside this third,most prevalent approach,developments in em-pirical multivariate analysis can be broken down into two basic trends in application.Thefirst and most diffu technique verifies complemen-tarity by testing the significance of interaction variables created from factors of rearch interest,controlling for exogenous factors and possi-bly omitted variables2.The cond technique,on the other hand,re-quires either structurally discrete variables,or variables empirically proven to be discrete,or a dichotomization of continuous variables.Dis-crete variables of interest allow for the identification of afinite ries of combinations,which,in other words,indicate different states of the world.The states of the world are either attributable to cas of com-plementarity(prence or abnce of all factors)or to cas of substitut-ability(other states,with at least one factor missing).The goal is to examine whether the impact on the performance of cas of comple-mentarity surpass,or at least is equal to,the effect of substitutable states.The added value of the cond analytical practice is in its higher
degree offlexibility,even if it lies within a statistical context of incread complexity as regards testing for complementarity,since it involves ex-amination of the vectors of two,three or even more elements of interest.
All three approaches outlined above can be attributed to conceptual schemes that are modular in nature,where the organization or system analyzed can be broken down into explanatory factors and exogenous elements/parameters.
Concerning the framework of discrete analys within the more recently developed productivity approach,we cite the contributions of Galia and Legros(2004a),Mohnen and Roller(2005),and Carree et al. (2011)as the most reprentative.
1World Input Output Datat,stemming from the WIOD EU project funded under the
Seventh Framework Programme FP7.It is a ctor bad economic environmental ac-counting datat.
2For a clo examination of problems related to the estimation of the reduced forms, e the contributions of Arora(1996)and Carree et al.(2011).
ie是什么意思
57
M.Gilli et al./Ecological Economics103(2014)56–67
Since Mohnen and Roller's(2005)minal applied work devoted to testing empirical evidence for complementarities in national innovation policies,a great deal of economic literature has revolved around empir-ical analysis in order to test complementarities infirms'innovation practices3.Firms'innovation activity is a complex outcome deriving from the influence of many factors that are interrelated through com-plementary relationships which might give ri to systemic effects.
As regards literature on innovative strategies and performance,one significant work is that of Miravete and Pernias(2006),in which they show the prence of complementarity between process and product in-novations in the Spanish ceramic tile industry.They u a structural dis-crete choice model of production and innovation decisions through which it is possible to distinguish if obrved correlations among innova-tions is due to a relationship of complementarity or if it is only induced by firms'unobrved heterogeneity.Finally,the contribution of Quatraro (2011)is also pretty interesting.He investigates the role of ctorial com-plementarities in the impact of Information Communication Technologies (ICTs)on productivity.Using an empirical strategy bad on a general Cobb–Douglas pr
oduction function,he compares the estimation deriving from the ca in which the ICT industries are considered as complemen-tary production factors with the estimation from the ca in which indus-tries are considered as substitute production factors,to demonstrate that ICTs capital and rvices are complementary when considering their im-pact on growth process.Though we do not treat here ICT due to CIS data limitations,the analysis of complementarity between EI and ICT is a key fact behind dematerialization and competitiveness achievements.
The literature shows that the issue of complementarity,in its various aspects,has gained momentum over the years.It is relevant to be explored given that management strategies and good practices have in-creasingly emphasized that competitiveness relies upon how different innovations are quantitatively and qualitatively combined more than on single investments.
In the following ctions we do explore how EI(Environmental In-novation)integrates with other innovations.Building upon the afore-mentioned literature,our study offers original and insightful evidence through a focus at a wide EU level.
nervous怎么读3.Environmental Productivity and Complementarity Among Innovations:the Conceptual Framework
Remaining within the innovation sphere,we believe that deepening the empirical analysis of comple
mentarity among differentfirms' innovation practices is particularly relevant when environmental inno-vations are involved,especially in the increasing need to adopt integrat-ed and more complex green strategies and not only“end of pipe”technologies.
This consideration strictly descends from the definition of Envi-ronmental Innovation itlf.In the MEI(Measuring EI)rearch pro-ject(Kemp and Pearson,2007),EI is defined as“the production, assimilation or exploitation of a product,production process,rvice or management or business method that is novel to the organization (developing or adopting it)and which results,throughout its life cycle,in a reduction of environmental risk,pollution and other negative impacts of resource u(including energy u)compared to relevant alternatives”4(Kemp,2010,p.2).
The definition of EI is not limited to specific technologies;it also in-cludes new organizational methods,products,rvices and knowledge-oriented innovations5.It is worth stressing that the CIS data we here exploit,though prenting EI for thefirst time within a wide coverage survey(the EU),do not dintangle between process and product EI. Product EI tend to increa the value of the product(the attached willing-ness to pay in monopolistic competitive markets),while process EI gener-ally reduce gy efficiency one key example).This is a slight data limitation that opens windows for further rearch.Nevertheless,we do believe that EI and other innovations largely
euphemismbelong to distinct realms —e.g.the adoption shares are different.It is also worth noting that while product innovation generally aims at increasing value added,process in-novations in the EI realm might be pretty radical compared to the‘non-EI’counterpart.Most innovative changes in energy u that are necessary to cope with climate change mitigation are production process restructuring to enhance energy efficiency and change the energy mix.
The importance of adopting integrated strategies for innovation is particularly relevant in complexfirms'technologies such as tho pertaining to CO2abatement,compared to cuts in emissions such as SOx–NOx(Cainelli et al.,2013;Marin and Mazzanti,2013).The latter might occur through end of pipe technologies while CO2abatement de-pends upon a radical change in the energy–technological framework. Various internal and external drivers(Horbach et al.,2012)are relevant in triggering decarbonization.The costly process of business decarbonization might be mitigated by the occurrence of complemen-tarity,which,for example,generates increasing returns to scale.
We are particularly interested in analyzing the relationship between firms'environmental performance and different innovation practices, including environmental and process,product and organizational inno-vations.More specifically,the agent of our analysis is not thefirm,but the ctor,for two reasons:t
hefirst resides in the availability of data (which is ctorial);the cond is that the meso level is the level in which we can fully understand how specific innovation,environmental and economic performances behave and interact(Costantini and Mazzanti,2013).
In the prent specific ca,we assume that there is afinite t of economic ctors,indexed by j=1,…,J.In each ctor there are a large number of atomistic identicalfirms;we can therefore assume that each ctor features one reprentativefirm.
We consider the environmental(productivity)performance of ctor j (EP j)as the ctor's objective function and we focus on two innovation practices that can affect the ctor's EP function.One of the two innova-tion practices is Environmental Innovation(EI)and the other one is either the product,or process or organizational innovation itlf(PI)6.
EP j¼EP j EI;PI;θj
ðÞ∀jð1Þ
The problem with ctor j resides in choosing a combination of inno-vation practices,(EI,PI)∈I,which maximize its EI function.θj repre-nts the ctor's exogenous parameters(such as ctor-specific environmental policies,or the ctor's geographical locations).
We are particularly interested in analyzing whether a relationship of complementarity exists between EI and PI.
2016年4月2日Since innovation practices are typically investigated in discrete adopting or not,adopting at an intensity higher than the average,etc.),we study complementarity between the forms of ac-tions through the properties of supermodular functions(Milgrom and Roberts,1990,1995;Milgrom and Shannon,1994;Topkis,1995,1998).
This technical approach has the benefit of focusing on a purely eco-nomic analysis,without the need to dwell on more mathematical issues, such as particular functional forms that ensure the existence of interior optima.For example,no divisibility or concavity assumptions are need-ed,so that increasing returns are easily encompasd.
In our specific ca,complementarity between the two different inno-vation practices may be analyzed by testing whether EP j=EP j(EI,PI,θj)is
3Among others,e Bocquet et al.(2004),Cozzarin and Percival(2006,2008),Gomez and Vargas(2009),Schmiedeberg(2008).
4Results of the MEI project can be found it.unu.edu/MEI/.
5The importance of deepening analysis of the relationship between EI and other inno-vation practices has already been stresd in Antonioli et al.(2013),even if at a narrower regional level.
6The relationship of complementarity may involve more than two variables simulta-neously through a chain reaction that starts from a complementarity relationship between two variables and in turn involves a complementarity relationship between one of the two variables and a third variable and so on.
58M.Gilli et al./Ecological Economics103(2014)56–67
supermodular in EI and PI.Since each ctor is characterized by speci fic exogenous parameters (θj ),even if the maximization problem is the same for all the ctors,the EP function may result supermodular in EI and PI for some ctors but not for others.
In our empirical analysis,the ctor's environmental performance that we want to test is related to an index of environmental productivity.More speci fically,in agreement with Repetto's (1990)de finition of a “single factor measure of environmental productivity ”(Repetto,1990,p.36)7,we consider each ctor's value added per unit of CO 2emissions.Obviously,the lesr the ctor's CO 2emission is with respect to its value added,the better is its environmental performance,and the higher is its envir
onmental productivity (EP j ).Environmental innovations (EI)that reduce environmental damages of cour contribute to environmental productivity.What we want to verify is if EI is complementary to other innovation practices (either product,process,or organizational)when the ctor's objective function is its environmental productivity.
If a ctor choos to adopt none of the two practices in its EP maximizing problem,namely EI =0,PI =0,the element of the t I is EI ∧PI ={00}.If a ctor choos to adopt both practices,we have EI =1,PI =1and the element of the t I is EI ∨PI ={11}.Including mixed cas as well,we have four elements in t I that form a lattice:I ={{00},{01},{10},{11}}.
From the above we can asrt that EI and PI are complements and hence that the function EP j is supermodular,if and only if:EP j 11;θj þEP j 00;θj ≥EP j 10;θj þEP j 01;θj ;ð2Þ
or:
EP j 11;θj −EP j 00;θj ≥⌊EP j 10;θj −EP j 00;θj ⌋þ
þ⌊EP j 01;θj −EP j 00;θj
⌋
ð3Þ
That is to say,the changes in the Environmental Productivity of ctor j that are brought about when both Environmental Innovation and process/product/organizational innovations increa together are more than the changes resulting from the sum of the parate increas of the two kinds of innovations.8Speci fically,increas in EP due to an increa of both EI and PI from {00}to {11}are greater (or at least equal)than the sum of the increas in EP due to parate increas of EI and PI from {00}to {10}({01}).
salute是什么意思
To sum up,complementarity between the two decision variables (EI and PI)exists if the EP j function is shown to be supermodular in the two variables and this happens when either inequality (2)or in-equality (3),or other derived inequalities are satis fied.
As mentioned above,different ctors'exogenous parameters (θj )may imply different degrees of complementarity between the two innovation practices (EI j and PI j )9.
In our speci fic analysis,we are particularly interested in verifying whether the different ctors and geographical speci ficity and also the strength of environmental regulations to which ctors are expod may play a role in the exploitation of complementarity relationships
between environmental innovations and other innovation practices 10.We will then narrow the analysis to some sub ctors of the economy and geographical areas.As regards policy,we asss whether a joint im-plementation of EI/PI strategies can improve environmental productivity especially when situations of more stringent environmental regulations are prent.We will therefore focus on ETS ctors in some speci fic anal-ys 11.More stringent environmental standards may indeed foster firms'adoption of product,process or organizational innovation,which in turn could lead to further environmental innovation.The conceptual frame-work refers somewhat to the Porter idea of competitive firm advantages that reside in the firm value chain,within which ‘strategy is manifested in the way activities are con figured and linked together ’(Porter,2010).
Building upon the aforementioned discussion,we can thus t out two main rearch hypothes:
[H1].Complementarity between environmental innovations 12aimed at abating CO 2on the one hand,and product,process,and organization-al innovation on the other hand is crucial to increasing environmental productivity.
[H2].Manufacturing might prent more evident signs of innovation complementarity than non-ETS ctors,given (i)the higher (compared to rvices)innovation intensity and (ii)since tho ctors are
presd to find more radical solutions in order to remain both competitive and sustainable by regulatory tools that put a price on carbon.
The cond hypothesis is relevant even becau the EU is currently rethinking its industrial development.The ‘Mission for growth ’states that ‘Europe's economy cannot survive in a sustainable way without a strong and profoundly reshaped industrial ba.New technologies have dramatically changed our life and our economy in the past 20years ’(ec.europa.eu/enterpri/initiatives/mission-growth ).The EU aim is to increa the industry GDP to 20%by 2020from the current 16%.
This opens two considerations.First,the issue is relevant becau it readdress the old structural change fact that rvices are intrinsically less innovative (Baumol's dia).This might be critical for the long run growth of productivity,which largely depends upon innovation.The lower innovation we witness in the rvices for innovations in many countries,including tho of environmental oriented flavor,is one crit-ical fact when asssing the prospected environmental and innovation performances:a rvice oriented economy is not per a greener econ-omy (Cainelli and Mazzanti,2013;Marin et al.,2012).Various rearch windows open out of this consideration.
Second,it is evident that in the short run this ‘remanufacturing ’target undermines the environmental performances at EU level.In the medium long term nevertheless the greater innovative capacity of industrial ctors might counterbalance the structural effect toward the achievement of 2030and 2050targets.
The H1and H2are then tested by focusing on different geograph-ical areas of the EU.The main reason is that northern EU is an area where carbon pricing and climate change policies are historically more stringent (Johnstone et al.,2010;Mazzanti and Musolesi,2013)13.
7
For extensive discussion on environmental productivity measures and their conceptual background we refer to Mazzanti and Zoboli (2009a,b).Here we simply remark that the IPAT framework and its ‘statistical ’counterpart (STIRPAT)are a general conceptual umbrella (York et al.,2003)to study the economic and innovation determinants of environmental performances.8
From Eqs.(2)and (3)it is evident that complementarity is symmetric:increasing EI rais the value of increas in PI.Likewi,increasing PI rais the value of increas in EI.On the other hand,it is possible to asrt that a substitutability relationship exists if:EP j (11,θj )−EP j (00,θj )≤[EP j (10,θj )−
EP j (00,θj )]+[EP j (01,θj )−EP j (00,θj )],that is,the changes in the ctor's environmental productivity when both forms of innovation practices (EI and PI)are incread together are less than the changes resulting from the sum of the parate increas of the two kinds of practice.9
In Cassiman and Veugelers (2006),great emphasis is given to the analysis of the con-textual variables affecting the supermodularity of the performance function that allows one to understand the conditions under which innovation strategies are complementary.
10
A few examples of stringent environmental standards are:the EU 2003Directive on emission trading;the 2008Directive IPPC on emission abatement and environmental technology together with its 2010revision;and the EU Waste Packaging Directives of 1994and 2003.11
The EU Emission Trading System (ETS),which followed a proposal for a Directive that had been discusd since 2001,was launched by the related 2003EU ETS Directive.It is currently the major EU policy aimed toward achieving Kyoto and EU 2020targets.It allo-cates tradable CO 2permits to firms in ctors such as metallurgy,ceramics,paper and cardboard,chemical,and coke and re finery,as far as manufacturing is concerned.The in-novation effects of (the EU)ETS (Ellerman et al.,2
010),though having been extensively analyzed and compared to other environmental policies at the theoretical level,so far have not found consolidated empirical testing.12
Process and product technological innovations,following the EU CIS survey.13
Given the shrinking of the datat when focusing on regional sub areas,we give prior-ity to testing H1and H2.‘Regional ’investigations are corollary analys that might open windows for future rearch.
59
M.Gilli et al./Ecological Economics 103(2014)56–67
4.The Empirical Framework
4.1.The Data
The data ud in this analysis comes from two main different sources:the CIS Eurostat innovation data and WIOD environmental–economic accounting.
Thefirst of the is data on innovation practices(eco-innovation14, organizational innovation,product and process innovation15)as well as data on ICT adoption are from the sixth Community Innovation Sur-vey(CIS),16who ctoral level is available on EUROSTAT website. The Community Innovation Survey is a ries of surveys produced by the national statistical offices of the27European Union member states, also covering the European Free Trade Association countries and the EU candidate countries.The surveys have been implemented since1993, on a biannual basis and are designed to obtain information on the inno-vation activities of enterpris,including various aspects of the innova-tion process,such as innovation effects,cost and sources of information ud.Data is collected at the micro level,using a standardized question-naire developed in cooperation with the EU Member States to ensure comparability across countries.The sixth CIS(2006–2008)collects data on environmental innovation for thefirst time17.Though it is a cross ction datat,it captures a3-year time span of EI and is the first CIS survey ever to include EI at the EU level.Community Innovation statistics-bad data is the main data source for measuring innovation in Europe and is ud in the academic rearch as in Horbach et al.(2012), Borghesi et al.(2012),and Veugelers(2012),which exploit data for Germany,Italy,Belgium,respectively.
The cond source of data is the World Input Output Databa (WIOD),which results from a Europe
an Commission funded pro-ject as part of the venth Framework Programme and was devel-oped to analyze the effects of globalization on socio-economic variables and trade,in a wide range of countries(the27EU member states and other13major countries in the world,from1995to 2009).The WIOD is made up of four different accounts(World Tables,National Tables,Socio Economic Accounts and Environmen-tal Accounts).For the purpo of this work,we ud the Socio-Economic and Environmental Accounts,both providing a wide range of economic variables such as value added,employment and CO2emissions.
Table1shows summary statistics and gives a description of the variables considered in this analysis.Building on the concept of environmental productivity(Repetto,1990)the dependent vari-ables VA/CO2_09and VA/CO2_10are obtained as the ratio between ctorial value added and ctorial CO2emission in2009and2010 respectively.We note that VA/CO2is higher in2010.This means, taking into account the GDP collap in2009,that the GDP increa in2010was lower overall than the related CO2emission increa (with respect to2009).
Innovation practice indicators,originally prented by Eurostat as the share offirms introducing innovation per ctor have been di-chotomized to obtain an innovation adoption indicator.To compute the binary variable,we compare the country's ctorial value(name-ly the share offirms ado
pting EI)to the average CIS sample ctorial value18:if the specific country/ctor value is above the CIS average, the adoption indicator values are1and0otherwi;however,since the average is nsible to outliers,to test if our empirical analysis was robust,we computed the innovation indicator also using the Appendix A value and the third quartile value when statistically fea-sible with our data size for dichotomization.Notwithstanding this, we did not obtain generally different results.
In order to test for complementarity,we ud the dichotomized innovation practice indicators to create four states of the world for each joint adoption of innovation.For example,concerning the introduc-tion of both eco-innovation and organizational innovation(Table2)we obtained an‘index’19for joint adoption(EI/OI(11)),two indexes for the adoption of only one of the practices(EI/OI(10)stands for EI adop-tion only;EI/OI(10)stands for organizational innovation adoption only)and,finally we obtained the index EI/OI(00)when none of the practices were introduced.
The following tables from2to4show the distribution of the states of the world for the adoption of EI and organizational innova-tion,EI and product innovation,EI and process innovation for the whole of EU(Table2),for Northern Europe(Table3)and for South-ern Europe(Table4).
4.2.Econometric Evidence
man hub
The empirical model we rely upon is a cross ction framed regres-sion wherein the dependent variable‘environmental productivity’(VA/CO2in2009or2010)is diachronic with respect to lagged innova-tion adoption(2006–2008).This rules out the simultaneity between innovation and productivity which might generateflaws at empirical level.
The regression we test is:
VA=CO2
t
¼β0þβ1vaemp2008þβ2ICT2008þ
þβ3EIPI11
2008
þβ4EIPI10
2008
þβ5EIPI01
2008
þβ6EIPI00
2008
þεt;
ð6Þ
where t refers to2009and2010respectively while PI reprents inno-vation practices other than ,product innovation or process innovation or organizational innovation,respectively). Labor productivity(vaemp)and ICT are picked in2008,while inno-vation practices from CIS-VI cover the three years between2006 and2008.
The inclusion of labor productivity as a main covariate follows Mazzanti and Zoboli(2009a,b)and aims at capturing ctor heteroge-neity and general heterogeneity in economic conditions.ICT invest-ments are included to further control for a‘new economy’factor that can absorb relevant cross
ction heterogeneity.The four factorsfinally introduce the states of the world for which EI and other‘innovations’are both prent(11),neither are adopted(00),or they are adopted in iso-lation of each other(10,01).We u OLS as an estimation procedure and we correct for heteroskedasticity in a usual fashion.The parsimoni-ous regression aims to mitigate collinearity(e the Appendix A for correlations).Since labor productivity and ICT are not correlated–this recalls the‘Solow productivity paradox’–we can inrt both as main factors.Other controls then contribute to mitigate unobrved hetero-geneity:as example,heterogeneity is further controlled by geographical
典故是什么意思14We only consider CO
2
abatement innovation for the purpo of this work.In the CIS-VI eco-innovation module,afirst t of questions asks respondents if they have introduced an innovation with one or more environmental benefits(ECO).Six types of environmental ben-efits are listed that can occur during the enterpri's u of the innovation(ECOOWN):lower u of materials(ECOMAT),lower energy u(ECOEN),lower CO2emissions(ECOCO),less u of pollutants(ECOPOL),less soil,water,air or noi pollution(ECOSUB),and recycling (ECOREC).
15We do not exploit in the work the information on EI that pertain to organizational strategies,such as EMS and ISO.
uropa.eu/portal/page/portal/microdata/cis.Data is available at uropa.eu/portal/page/portal/science_technology_innovation/data/ databa.
17Information taken from the Eurostat website(uropa.eu/ portal/page/portal/microdata/cis).
18The CIS ctorial average for each country is adjusted by omitting the country ctorial value when making the comparison.For example,for the manufacturing ctor in Italy we compared the Italian manufacturing value to the CIS manufacturing value computed without Italy.
19A state of the world.
60M.Gilli et al./Ecological Economics103(2014)56–67treat
