Environmental and innovation policies for the evolution of green technologies: a survey and a test

Abstract

The aim of this paper is to provide an updated survey of the “state of the art” in environmental policies, with a particular focus on those policies fostering the generation, diffusion and exploitation of technologies improving environmental performances, i.e. eco-innovations. In particular, we provide a critical appraisal of the existing policy schemes, by emphasizing how the demarcation between environmental and technology policies gets more and more blurred. The case for a systemic approach to the implementation of policy measures is finally discussed. Econometric evidence about the importance of environmental policies is also provided in the “Appendix”.

Appendix

This Appendix provides empirical evidence on the inducement effects played by policies to support the discussion on the role of policies for environmental innovations outlined in the paper. It is worth noting that the estimated relations can only be considered as informative of the presence of conditional correlation between regulation and eco-innovation, and that only weak claims of causality can be done.

A country-sector level analysis based on Eurostat and OECD data sources is proposed, coherently with previous contributions that motivate the appropriateness of such a focus (e.g. Gilli et al. 2013). The aim of the appendix is to empirically test for the role of policies, measured in several alternative ways, in inducing the adoption of EI by firms.

The core data on environmental innovations adoption are taken from the Eurostat Community Innovation Survey for the year 2006–2008. As this survey introduced for the first time a section on EI adoption, the analysis is conducted on those countries that included such section in the survey and that made results available for consultation. Those are the following: Belgium, Bulgaria, Cyprus, Croatia, Czech Republic, Estonia, Finland, France, Germany, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherland, Poland, Portugal, Romania, Slovakia and Sweden. As for the sectors included in the analysis, we focused on the following aggregates (Table 2):

The baseline model we estimate is the one reported into Eq. 1.

$$ EI_{{\left( {ECOCO2 - ECOSUB} \right)i,j}} = \alpha + \beta_{1} IS_{i,j} + \beta_{2} RD_{i,j} + \beta_{3} POLICY_{i, j} + \gamma_{i} $$

(1)

Nine typologies of environmental innovations adopted constitute our dependent variables (EI). Previous literature highlighted the need to open the box of EI and to look into the heterogeneities that each typology of EI brings about (e.g. Ghisetti and Rennings 2014). For this reason we estimate Eq. 1 on nine different dependent variables, each of which represents the share of firms of the country-sector adopting a specific type of EI: reducing carbon dioxide emissions (EOCO₂), energy use per unit of output (ECOEN) or after sales use (ECOENU), material use (ECOMAT), soil, water, noise, or air pollution per output (ECOPOL) or after sales use (ECOPOS), improving recycling per output (ECOREC) or after sales use (ECOREA) and replacing materials with less polluting or hazardous substitutes (ECOREC). Estimations are carried out though Ordinary Least Squares estimator with robust standard errors to cope with heteroscedasticity problems. As EI adoption has been surveyed in the CIS 2006–2008, we are only able to run a cross-sectional analysis in which dependent and explanatory variables refer either to the period 2006–2008 (EI variables, RD, REG, FUT_REG, GRANTS) or to 2008 (POL_index, lCo2_va).

As for the explanatory variables, we drew on already established literature to derive a set of core control variables to be included (e.g. Ghisetti et al. 2015; Horbach et al. 2012). Given the low number of observations, we had not enough degrees of freedom to include additional variables. In particular, we chose to control for the role of external R&D activities (RD) and country dummies to capture structural country heterogeneities and we included our core policy variable, which has been constructed in alternative ways. As policy does incorporate an induction effect on internal R&D, we have opted to include the external R&D variable to avoid potential biases. Whereas our sample equals 134 observations in the case of a policy variable constructed from CIS data, it falls to 68 and 71 when exploiting alternative policy measures, as data are not available for all the countries of the first dataset, more precisely Eastern EU countries are excluded.

Table 3 reports the variables short description together with their descriptive statistics, while Table 4 outlines the pairwise correlations among them.

Table 3 Variables descriptive statistics

Table 4 Variables pairwise correlation matrix

At first, we built a policy variable able to capture a broad measure of environmental regulatory stringency. Drawing on previous literature (Costantini and Crespi 2008) it has been built as the logarithm of the ratio between carbon dioxide emissions and value added (lCO2_va). The results of this first model are reported in Table 5 and support is found for the inducement effect played by environmental regulation stringency on the share of adoption of EI, though with some exceptions in which regulatory stringency does not display any significant effect.

Table 5 Results on environmental policy built as an environmental policy stringency variable

Table 6 Results on environmental policy built from CIS data

Table 7 Results on composite policy index

As an alternative to this broad measure of policy intervention, we conducted further estimations in which the policy variable has been replaced with alternative measures.

In the second round of estimations (Table 6), the policy-related variable has been extracted from the CIS 2006–2008 survey. It refers to the motivations that drove firms to adopt EI. In particular, we selected among the available options the adoption of EI consequently to existing regulations or taxes on pollution (REG), to expected forthcoming regulations (FUT_REG) or to the availability of subsidies and grants (GRANTS). In the second model, we thus jointly included REG, FUT_REG and GRANTS to test for the role of these motivations in driving the rate of EI’s adoption. As it emerges from Table 5 GRANTS do actually drive EI’s adoption choices in the case of ECOMAT, ECOPOL, ECOPOS and ECOREC, while the expectations over an upcoming environmental regulation, is a significant driver of EI in most of the cases, with the only exceptions of ECOENU, ECOPOL and ECOREC. Surprisingly instead the role of existing regulation and taxes of pollution is found to play only a rather marginal role, as it determines EI only in the case of recycling related innovations (ECOREA and ECOREC).

We then exploited the OECD database on instruments used for environmental policy and natural resources management that collects information on environmentally related taxes, fees and charges, tradable permit systems, deposit refund systems, environmentally motivated subsidies and voluntary approaches used in environmental policy. In the third round of estimations (Table 7), we built the policy variable as an index that counts the number of instruments in place for each country-sector (POL_index) among environmentally related taxes, fees and charges, tradable permit systems, voluntary approaches, deposit refund schemes and subsidies. This variable emerges to be as expected a significant driver for most the typologies of EI, with the only exception of ECOENU and ECOREA.

Overall, we can conclude that, coherently with the literature scrutinized in the paper, policies are found to play a crucial role in supporting or even spurring the adoption of environmental innovations. We then differentiated among (1) typologies of policy instruments and (2) typologies of innovations. We found support that actually the inducement effects depend on the type of instrument under scrutiny: this is coherent with the qualitative overview of available policy instruments provided in the paper. We also found empirical support to the argument linking inducement effects of a policy and the specific type of EI one may be willing to spur, as heterogeneous EI differently react to the array of policy instruments scrutinized. Given the cross-sectoral nature of the available dataset and the consequently limited empirical setting, it is needed to acknowledge as the main limitation of this exercise the fact that we can interpret the relation between EI and policy as a (significant) correlation, rather than as a proper causality effect of policy on EI.