Environmental tax and productivity in a decentralized context: new findings on the Porter hypothesis

Abstract

This paper studies, for the first time, the effects of environmental taxes on efficiency gains and the growth of the regions. To this end, we have estimated a dynamic panel-data model to the context of the Spanish regions that reflects the effects of environmental taxation and regulation separately. The results provides further empirical evidence in favour of the Porter hypothesis, to the extent that a strict environmental policy implemented via green taxes rather than regulation may raise productivity, which may be because they drive organizational and technological change in firms seeking to reduce their tax payments.

1 Introduction

There has been much debate in recent decades over the possible impact of regulation on productivity and the superiority of economic instruments, particularly environmental taxes,¹ as a weapon in the fight against pollution. It was Porter (1991) who drew attention to the first of these issues, arguing against the conventional view that strict regulation properly implemented would encourage productivity growth by creating benefits capable of offsetting regulatory costs. The second issue came to the fore, meanwhile, as awareness of environmental issues rose in the developed nations in parallel with increasing concerns about the inefficiency of the command and control instruments that had traditionally been used to combat pollution. The result has been increasing use of green taxes in environmental policy in recent decades, given their advantages in terms of efficiency and economic growth.²

Based on a review of the existing literature, we may identify various factors that are directly or indirectly associated with productivity and economic growth: human capital, infrastructure, technology gap, openness and absorptive capacity, … even environmental aspects. In particular, we might mention the surveys by Isaksson and Ng (2006) and Isaksson (2007), although numerous other papers also examine the determining factors of productivity including, for example, Aghion and Howitt (1998), Hall and Kramarz (1998), Easterly and Levine (2002), Keller and Yeaple (2003), Fisman and Love (2004), and Bloom et al. (2004).

According to the conventional view defended by scholars like Barberá and McConnell (1990), Gollop and Roberts (1983), Gray (1987), and Palmer et al. (1995), a strict environmental policy imposes costs on firms that affect their competitiveness, resulting in adverse socio-economic outcomes for jobs and living standards. This is so because regulation almost always requires firms to allocate a part of their inputs (labour, capital, etc.) to reduce pollution, which is unproductive from a business standpoint. Moreover, regulation may depress investment if it raises the price of energy (a supplementary input to capital), as Ambec and Barla (2006) point out.

However, two decades ago Porter (1991) proposed a different approach to the analysis, arguing that a strict, effectively implemented environmental policy could have the opposite result, fostering productivity and comparative advantages that would enhance the competitiveness of regulated firms in such a way as to offset the costs initially entailed. To put this another way, additional benefits may be generated that are not detected by conventional theory, which fails to consider the dynamic nature of the problem. This hypothesis is defended in Porter and Van der Linde (1995), Shrivastava (1995), Faucheux et al. (1998), Mohr (2002), and Ambec and Barla (2006).

There are three different readings of the Porter hypothesis (Brännlund 2008). The first maintains that regulated firms cut their costs by eliminating internal inefficiencies. The second refers to relative gains in competitiveness (in relation to other firms), which Porter calls the “early mover advantage”: despite the costs inherent in regulation, these costs will be greater for firms that are regulated later. The third interpretation is based on enhanced competitiveness through the increase in demand for products associated with environmental regulation. Thus, the gains do not come from the regulated firms themselves, but from firms supplying them with the materials and equipment required to comply with environmental regulations.³

As Brännlund (2008) notes, Porter’s ideas are controversial and they have spurred considerable theoretical research (Simpson and Bradford 1996; Xepapadeas and de Zeeuw 1999; Mohr 2002; Feichtinger et al. 2005; Popp 2005, and Greaker 2006, among others).⁴ Our approach is to test whether environmental taxes can have a positive impact in terms of productivity gains and/or economic growth, given that such taxes provide an alternative instrument offering considerable advantages over conventional regulation, in terms of cost-effectiveness and dynamic efficiency.⁵ The objective is not to show environmental effectiveness but some kind of prerequisite in the sense of economic effectiveness and productivity. It can be argued that environmental policy instruments which put a burden on those who need to comply actually may fail to achieve their environmental effectiveness for that reason. If, however, an environmental policy instrument achieves the opposite and tends to increase the productivity of those who need to comply, one can predict or expect that it may also be environmentally effective.

This hypothesis can be reinforced by others like the theory that defend that environmental taxes can even be used to undertake green reform processes capable of generating what has been called the “double dividend” (see Pearce 1991; Goulder 1994; Bovenberg 1999; Bosquet 2000; Hoerner and Bosquet 2001; Schöb 2003; and, more recently, Fullerton et al. 2008, for an explanation of the double dividend hypothesis).⁶ Furthermore, even if this “double dividend” does not occur, it is not unthinkable that more intensive use of environmental taxes would further tax decentralization through revenues, which could also boost economic growth. This renewed interest in decentralization has various roots, including the conviction that it provides a useful tool to improve the efficiency of the public sector (Martínez-Vázquez and McNab 2003). Its potential effect on growth is based on two assumptions. The first is the Leviathan model proposed by Brennan and Buchanan (1980), which implies that the public sector will vary, ceteris paribus, inversely to the scope of decentralization. The second is the hypothesis that sub-central tiers of government have access to privileged information about citizens’ needs and preferences given their proximity, and they are therefore better placed to provide public services than central government. The corollary to this is geographical mobility and competition between different administrations, resulting in enhanced living standards and more uniform income distribution.

In view of the Porter hypothesis and the superiority of environmental taxes over command and control measures, we ask whether the environmental taxes established in the regions of Spain have had any impact on productivity and the growth of regional economies.⁷ In contrast to the few other papers that have looked at this relationship in Spain (Gil and López Laborda 2005; Carrion-i-Silvestre et al. 2007; and Pérez and Cantarero 2009), the scope of the study described here is confined to regional decentralization and green taxes, an issue that is also largely absent from the international literature. Hence, the approach we have taken is new.⁸

The paper is structured as follows. The next section describes the model and examines the factors that may be expected to affect productivity and economic growth, paying special attention to regulation and green taxes. An econometric test is also carried out to validate the hypotheses proposed. The paper ends with a section containing our conclusions.

2 Estimation of the relationship between environmental policy and regional development

Much empirical research exists with regard to the Porter hypothesis, as shown in the recent surveys by Volleberg (2007), and Brännlund and Lundgren (2009). Table 1 summarises a representative cross-section of the existing literature. This literature generally analyze the manufactured industries of America (USA and Canadá), although there are several papers concerning European countries, especially Swedish. There are also some paper about Japan, Mexico, Argentina, Chile, Philippiness and India.

Table 1 International empirical evidence for and against the Porter hypothesis

As it can be seen in Table 1, many papers focus on the effects of environmental regulation on investment, innovation and R+D (Nelson et al. 1993; Xepapadeas and de Zeeuw 1999; Gray and Shadbegian 1998; Brunneheimer and Cohen 2003; and De Vries and Withagen 2005), while a second group examines effects on efficiency and productivity gains (Gollop and Roberts 1983; Berman and Bui 2001; Gray and Shadbegian 2003; Marklund 2003; Hamamoto 2006; and Van der Vlist et al. 2007). Finally, a third group of studies examines the effects of regulation on benefits and other type of financial impacts (Brännlund et al. 1995; King and Lenox 2001; Nicoletti and Scarpetta 2003; Filbeck and Gorman 2004; and Gupta and Goldar 2005).

This literature does not find clear evidence to validate the Porter hypothesis, because none of these three groups of studies gives clear answers to what extent regulations affects innovation, productivity, efficiency, or benefits. There are positive relationship between these variables and the environmental regulations, as it can see in the first part of Table 1, but many studies show the contrary effect (second part of Table 1), particularly, most studies indicate negative productivity effect from environmental regulations. It is therefore not possible to affirm that the Porter hypothesis is true, either from a theoretical or from an empirical standpoint. However, this does not mean that there are no specific cases and circumstances in which progress may be made in terms of business, productivity and economic growth after the implementation of an environmental policy.⁹

The question of whether or not the Porter hypothesis is in fact applicable cannot be answered definitively based on the existing research. This is presumably the result of several different factors. Perhaps the most important of these is the fact that existing studies fail to apply a formal hypothesis test to Porter’s idea, at least in part because there is no general consensus about what should be tested. Measurements and definitions are also problematic. What is meant by the terms “competitiveness” and “environmental regulation”, and how can they be measured? Brännlund and Lundgren (2009) also note that most studies fail to distinguish clearly between regulatory measures and instruments, even though Porter is relatively clear that only certain specific types of regulation can actually neutralize the initial costs, suggesting that a better approach might be to try and classify regulations into groups or categories and then analyze the differences in effects.

In light of the above, we shall differentiate between regulation and environmental taxation. The intention is to learn whether both types of instruments have a specific effect on productivity, not of individual entities like enterprises which have to implement the policies, but regions. Thus, our approach will be to examine whether environmental taxes can have a positive impact in terms of productivity gains and/or economic growth, given that such taxes offer considerable advantages over conventional regulation. Finally, we have employed a dynamic specification in our econometric estimates, as it may be that the discrepancies in the findings from existing empirical studies stem from the failure to consider the dynamic nature of the analyzed problem.¹⁰

2.1 Empirical model

The aim of this paper is to test, for the first time, the possible efficiency gains in terms of productivity and economic growth that may be generated by green taxes at the regional level in Spain, employing a dynamic specification in our econometric estimates. Consequently, our empirical analysis adapts the proposals contained in Jorgenson and Wilcoxen (1990), Nicoletti and Scarpetta (2003), De Vries and Withagen (2005), and Loayza et al. (2005) to the situation of the Spanish regions. We begin with the multi-factor productivity equation. Thus, if we denote the regions using the subscript i = 1, …, 17, then the value added for a given year “t” in each region will be produced by labour (H) and physical capital (K) applying a standard neo-classical production technology:

$$ {\text{Y}}_{\text{it}} = {\text{A}}_{\text{it}} {\text{F}}\left( {{\text{H}}_{\text{it}} ,{\text{K}}_{\text{it}} } \right) $$

(1)

where F(·) is grade one homogenous and displays decreasing returns for each factor of production, and A_it is a technology efficiency or multifactor productivity (MFP) index.

We extend the conventional endogenous growth model, in which MFP is generally expressed as a function of knowledge and a residual set of influences (Aghion and Howitt 1998), by assuming that efficiency depends on country characteristics as well as technological and organizational transfers from the technology-leader region (i = L). This implies that MFP growth in the leading region leads to faster MFP growth in catch-up regions by expanding production options.¹¹ We assume that the technology gap between a given region and the technological leader measures the extent of technology transfers. The leading region is defined as that displaying the highest level of MFP. Hence, multifactor productivity growth for a given region may be modeled as follows¹²:

$$ \Updelta {\text{lnA}}_{\text{it}} = \updelta_{\text{it}} \,\Updelta {\text{lnA}}_{\text{Lt}} - \upsigma_{\text{it}} \,{ \ln }\left( {{\text{A}}_{\text{i}} /{\text{A}}_{\text{L}} } \right)_{{{\text{t}} - 1}} + \upxi_{\text{it}} $$

(2)

where δ_it captures the immediate effect of changes in growth in the leader region; σ_it denotes the pace of technology transfer; ln(A_i/A_L)_t−1 is the technology gap between region “i” and region “L”, the technology leader; and ξ_it represents all other factors involved in MFP growth, including those related with inter-regional differences in environmental regulation and taxation.¹³

We also assume that environmental policy, whether in the form of green taxes or regulation, can affect opportunities and incentives for the adoption of cutting-edge technologies. Following Nicoletti and Scarpetta (2003), this link between environmental policy (POLAMB) and the rate of technology transfers between non-leading regions can be formulated as follows:

$$ \upsigma_{\text{it}} = \upsigma_{{ 1 {\text{it}}}} + \upsigma_{{ 2 {\text{it}}}} \;{\text{POLAMBit}} - 1 $$

(3)

Substituting [3] in [2], we obtain the following specification:

$$ \Updelta { \ln }\,{\text{A}}_{\text{it}} = \updelta_{\text{it}} \,\Updelta { \ln }\,{\text{A}}_{\text{Lt}} - \upsigma_{{ 1 {\text{it}}}} \,{ \ln }\left( {{\text{A}}_{\text{i}} /{\text{A}}_{\text{L}} } \right)_{{{\text{t}} - 1}} - \upsigma_{{ 2 {\text{it}}}} \;{\text{POLBAMB}}_{{{\text{it}} - 1}} \,{ \ln }\left( {{\text{A}}_{\text{i}} /{\text{A}}_{\text{L}} } \right)_{{{\text{t}} - 1}} + \upxi_{\text{it}} $$

(4)

In deriving a specification of the MFP equation that can be estimated empirically, it is important to observe that Eq. (4) could be considered an error correction equation derived from a first order lagged autoregressive specification, in which the level of MFP in each region is cointegrated with that of the leader region, as follows:

$$ {\text{lnMFP}}_{\text{it}} = \upbeta_{ 1} \,{\text{lnMFP}}_{{{\text{it}} - 1}} + \upbeta_{ 2} \,{\text{lnMFP}}_{\text{Lt}} + \upbeta_{ 3} \,{\text{lnMFP}}_{{{\text{Lt}} - 1}} + \omega_{\text{it}} $$

(5)

Reordering Eq. (5) under the assumption of long-run homogeneousness (1 − β₁ = [β₂ + β₃]), we obtain:

$$ \Updelta {\text{lnMFP}}_{\text{it}} = \, \upbeta_{ 2} \,\Updelta {\text{lnMFP}}_{\text{Lt}} - \, \left( { 1- \upbeta_{ 1} } \right){\text{ RMFP}}_{{{\text{it}} - 1}} + \, \omega_{\text{it}} ;\quad {\text{where}}:{\text{RMFP}}_{{{\text{it}} - 1}} = { \ln }\left( {{\text{MFP}}_{{{\text{it}} - 1}} /{\text{MFP}}_{{{\text{Lt}} - 1}} } \right) $$

(6)

Equation (6) is equivalent to Eq. (4), in which the relative MFP coefficient is a function of environmental policy. We have also imposed the constraint that the leader region’s MFP growth coefficient (β₂) and the technology transfer coefficient (1 − β₁) do not change.

In addition, the error term in Eqs. (5) and (6) can be broken down into a vector of variables (V_it) including structural aspects (e.g. human capital) and environmental policies with potential effects at the MFP level, as well as unobserved regional effects (f_i), national macroeconomic shocks (d_t) and a non-correlated error term (η_it).

$$ \upomega_{\text{it}} = \Upsigma_{\text{k}} \upgamma_{\text{k}} {\text{V}}_{{{\text{kit}} - 1}} + {\text{ f}}_{\text{i}} + {\text{ d}}_{\text{t}} + \, \upeta $$

(7)

From Eq. (6) it is clear that the MFP gap coefficient measures the (conditional) speed of convergence with the long-run stationary state of MFP. In the presence of technological convergence, moreover, the technology gap between each region and the leader converges on a constant value. This implies that the vector of variables (V_it) and the fixed effects for each region only translate into differences in MFP levels but not into permanent differences in MFP growth rates.

Following the theoretic approach, the productivity model (6)–(7) has been estimated using as explanatory hypotheses the variable representing the growth of the logarithm of the MFP in the leading region (ΔLLEADER_it), the logarithm of technology gap (LGAP_it−1), the interaction between environmental policy (POLAMB) and the LGAP_it−1 (calling this interaction IRLGAP_it−1 and ITLGAP_it−1, when regulation and taxes, respectively, interact with the logarithm of technology gap) and a vector of variables including both structural aspects (ESTFACTORS_it, e.g. human capital, stock of public capital per employee, economies of scale and changes in the production structure) and environmental policies (POLAMB_it, e.g. regulation and tax), with potential effects at the MFP level.¹⁴ All of these variables will be explained in the next section. This equation has been estimated using a dynamic panel-data model, which also includes the dependent variable lagged by one period (ΔLMFP_it−1) to capture the inertia behaviour of MFP, with fixed effects estimator for each Spanish region (f_i), as well as dummy time variables to control for common aggregate shocks affecting changes in MFP across all regions (d_t). Consequently, the model we estimate is the following:

$$ \Updelta {\text{LMFP}}_{\text{it}} = {\text{ f}}(\Updelta {\text{LLEADER}}_{\text{it}} ,{\text{ LGAP}}_{{{\text{it}} - 1}} ,{\text{ IRLGAP}}_{{{\text{it}} - 1}} ,{\text{ITLGAP}}_{{{\text{it}} - 1}} , \, \Updelta {\text{LMFP}}_{{{\text{it}} - 1}} ,\;{\text{POLAMB}}_{\text{it}} ,{\text{ ESTFACTORS}}_{\text{it}} ,{\text{ f}}_{\text{i}} ,{\text{ d}}_{\text{t}} ) $$

2.2 Data and empirical estimation

The period considered in the study runs from 1989 to 2001, for which all the relevant information is available. Thus, we have a sample comprising Spanish regions that have enacted environmental regulations and green taxes, although we have also included regions that made use of neither in order to avoid any possible sample selection bias.

MFP growth, the endogenous variable, was measured as follows: ΔLMFP_it = Δy_it − α_it Δl_it − (1 − α_it) Δk_it. where y, l and k are logarithms of actual aggregate or added value, total jobs and capital stock, respectively. Under conditions of perfect competition, “α_it” may be proxied as the participation of labour in GVA. The main source of data at the level of regional disaggregation for the calculation of productivity as we have defined it consists of the Regional Accounts published by the Spanish National Institute of Statistics (INE in its Spanish acronym),¹⁵ which contain comparable regional data on aggregate value and jobs. Data on the stock of fixed capital was obtained from publications by BBVA (Bank of Bilbao, Vizcaya, Argentaria), and the Valencia Institute for Economic Research (http://www.ivie.es/banco/stock.php).

As our objective is to test whether the environmental policies implemented by the Spanish regions have had any impact on efficiency in terms of productivity and economic growth, we have identified the different green taxes enacted by the Spanish regions during the period of the study (1989–2001). This variable has been included as the ratio of the green tax revenues to the total tax take of each region (ENVTAX). These data were supplied by the Ministry of Economy and Finance in territorial statistical publications,¹⁶ and it is not negligible as a percentage of the own tax of the region, as we have pointed out in footnote 7. Table 2 shows the environmental taxes levied by each region in this period, revealing that the most common tax across all regions is some kind of Waste Water Effluent Tax.¹⁷ Some other regions have other taxes, such as on Fuel—Canary Islands, on Hunting—Extremadura, on electricity—Extremadura and on Activities Affecting the Environment—Balearic Islands, …. This table also reveals that ever more regions have legislated in this regard, although some regions still no green taxes in 2001.

Table 2 Environmental taxes levied by the Spanish regions, 1989–2001

The database of the Spanish Institute of Public Law¹⁸ (Instituto de Derecho Público) has been used to obtain information on regional environmental regulation (ENVREG). This database provides a complete record of parliamentary activity in the Spanish regions, and thus the total of regulatory acts has been considered, as in the case of Holcombe and Sobel (1995), Rogers (2002, 2005) and Vallés and Zárate (2012b), and as the principal international organisms recommend for the analysis of regulation processes.¹⁹ It contains 22,607 different acts of the regional parliaments (see Table 6 in the “Appendix”), which we have grouped in 22 different types of regulatory acts (Table 7 in the “Appendix”). We have then made use of one of these types, i.e. the environmental acts, for constructing the ENVREG variable. For this, we have counted the number of environmental legislation passed by the Spanish regions in the study period, and which is presented in Table 3. Again, it may be observed that ever more regions enacted environmental regulations in this period.

Table 3 Number of environmental legislation enacted by the Spanish regions (1989–2001)

As the impact of these environmental policies is not necessarily immediate or permanent, we have lagged these variables. We have also included variables that measure the interaction between environmental regulations or taxes and the technology gap (IRLGAP and ITLGAP) in order to capture any possible influence of environmental policy on convergence between regions in the period considered. This interaction is also measured with a lag of up to three periods.²⁰

As mentioned in the description of the model, we have also considered other possible explanatory hypotheses for MFP growth in addition to these variables. All of them are shown with the expected sign in Table 4. Thus, the explanatory variables in the model estimated include the technology gap lagged by one period, which we have defined as the coefficient between the MFP of each region and the MFP of the leader region (LGAP_t−1) as the theoretical model suggests, MFP growth in the leader region (ΔLLEADER), and the stock of public capital per employee (PUBKW). We also consider the different levels of human capital between the regions to capture the influence of variations in the quality of labour, based on INE data, which we include as the percentage of illiterate members of the population according to the National Institute of Statistics’ Active Population Survey (ILLITERATE); changes in the production structure, measured as the coefficient of the relative share of services and farming in regional product (PRODUCTEST); and economies of scale or the rate of growth in gross added value in real terms based on the Regional Accounts for Spain published by the National Institute of Statistics (ΔRGVA). The model also includes certain temporal dummy variables to control for any common aggregate shocks that might affect the variation in the MFP of all regions (fixed effects).²¹

Table 4 Explanatory variables included in the model

As it seems reasonable to expect that the behavior of MFP will display some inertia, the method used to estimate the multifactor productivity model presented consisted of a dynamic panel-data model, which includes the dependent variable lagged by one period. Consequently, we opted to estimate the model using the generalized method of moments (GMM), which provides a consistent and efficient estimator (see Hansen 1982 and Arellano and Bond 1991). As the unobserved factor, f_i, (representing the specific characteristics of each region), may be correlated with the rest of the variables in the model, we shall estimate the model based on first differences to eliminate individual effects (f_i). All of the estimations therefore relate to first differences and two-stage estimators with standard errors that are robust to heteroscedasticity.

A sufficient number of valid instruments are required to employ the GMM method. In principle, any variable that might be correlated with the regression variables in period t would be classified as a valid instrument wherever it is orthogonal to the error term (otherwise the over-identifying restrictions would be rejected). In our case, we shall use all available lagged levels of the dependent variables, as well as the variables representing human capital, public capital per employee and changes in the production structure.²² We validate the instruments using the Sargan test for over-identifying restrictions.²³ Tests were also included for the absence of first and second order serial correlation, allowing us to analyse the consistency of estimators. If errors are not autocorrelated, evidence of first-order autocorrelation of differenced residual errors must exist, but no evidence of second-order autocorrelation.

Based on the results obtained, reflected in Table 5, we may conclude that both the dependent variable lagged by one period (ΔLMFP_t–1) and the technology gap lagged by one period (LGAP_t−1) is displaying a significant and negative effect at conventional levels. In view of the definition of the gap in the theoretical model, this outcome for the LGAP_t−1 might suggest that the regions furthest from the technological forefront have higher rates of productivity growth, which is to say long-run technological progress exists based on imitation of the leader, as Gual et al. (2006), Nicoletti and Scarpetta (2003) and Vallés and Zárate (2012a) appear to find for regulation indicators, although the former did not employ econometric techniques. Short run technological progress (i.e. the coefficient for the leading region) is also significant (ΔLLEADER), so the leader region therefore exerts a drag effect on the other regions.

Table 5 Results of the dynamic panel data model with lag effects for the estimation of multifactor productivity

The estimation also indicates that green taxes lagged by one period (ENVTAX_t−1) has a positive effect on productivity, which may be because taxes of this kind drive organizational and technological changes in firms seeking to reduce their tax payments, causing a positive effect on multifactor productivity growth. Furthermore, green taxes do not seem to influence the effect of the technology gap on productivity by accelerating technology transfer, as the impact of the gap on productivity growth did not prove to be significant when the two variables interact (ITLGAP).

Environmental regulation (ENVREG), meanwhile, displays a negative, significant coefficient, and it may therefore be deduced that this factor slows multifactor productivity growth, suggesting that policy has not been well instrumented, as Porter argued (1991). As Griffith and Harrison (2004) and Griffith, Harrison and Simpson (2006) argue, these costs do not really represent the economic or efficiency cost of regulation, because regulation also creates significant but unquantifiable burdens for government, the agents regulated and society in general, such as adverse effects on competition, flexibility and innovation (potentially acting as a drag on productivity and economic competitiveness), as well as uncertainties and private incentives that may affect decision making (distorting or overstimulating investment, or reducing the support required for certain activities). Nevertheless, the empirical evidence does not allow any clear conclusions to be drawn with regard to the impact of environmental legislation on technology transfer, as the coefficient for the interaction between the two variables is not significant (ITLGAP).

The influence of public capital per employee (PUBKW) and economies of scale (ΔRGVA) is positive and significant, as was expected. This result reflects the importance of the stock of public capital for productivity gains, as well as the significant role played by economies of scale in the growth achieved by Spain after joining the European Union. The reallocation of resources between productive sectors is measured through the variable (PRODUCTEST). The estimation displays a significant, negative coefficient, and the change in the structure of production may therefore be said to have had a negative influence on productivity growth, which is consistent with the increase in the relative size of the service sector and the expansion of construction, both sectors in which productivity tends to be low. Meanwhile, the results suggest that human capital (ILLITERATE) has an indeterminate effect on the variation in MFP. As Islam (1995) argues, this non-significant result can basically be attributed to the fact that the variables selected are not good proxies for the theoretical concept of human capital. Or it may be that the channels through which human capital affects productivity growth are more complex than can be reflected through the mere inclusion of a multiplier in the equation. There is another possible explanation associated with the fact that human capital presents important shortcomings as was evidenced by the Pisa report OECD (2004b) for Spain.

3 Conclusions

Economic growth and productivity may to some be extent influenced by a series of factors (such as human capital, the technology gap, etc.), and many studies have sought to establish whether environmental policy affects economic growth in the ways predicted either by Porter or by his detractors. Given the successive reforms of direct taxation undertaken by the central and regional government of Spain, and the policy of introduce green taxes in some regions, this study examines whether the environmental policy implemented by the Spanish regions could have had any influence on efficiency in terms of productivity and economic growth. To this end, we have adapted the multifactor productivity model proposed by Jorgenson and Wilcoxen (1990), Nicoletti and Scarpetta (2003), De Vries and Withagen (2005), and Loayza et al. (2005) to the context of the Spanish regions, and we have estimated a dynamic panel-data model for the period 1989–2001 that reflects the effects of environmental taxation and regulation separately. The model also allows dynamic treatment in our econometric estimate, given the possibility that failure to consider the dynamic nature of the problem analyzed may explain the discrepancies appearing in the different empirical studies.

The results from the model, which was estimated using the generalized method of moments estimator (GMM), suggest that short-run technological progress (i.e. the coefficient for the leading regions) is positive and significant. Hence, the leader region exerts a drag effect on the others, and the most technologically backward regions grow faster than the leader, which is to say they achieve long-run technology gains by imitation. This last fact together with the significance of the dependent variable lagged one period, might suggest that history seems to explain quite a lot of the present. Public capital per employee and economies of scale also have a positive influence on productivity growth, as was expected, while changes in the productive structure have had a negative impact on productivity growth, which is consistent with the increasing size of the service sector and construction, both low-productivity industries.

The results concerning the two “key” parameters (e.g. environmental taxes and environmental regulation) would provide further empirical evidence in favour of the Porter hypothesis, to the extent that a strict environmental policy implemented via green taxes rather than regulation may raise productivity or enhance comparative advantage, improving the competitiveness of the firms subject to the environmental policy, thereby offsetting the costs it initially entailed. However, this affirmation should be made with certain reservations, because the results for these lagged variables are not relevant.

The obtained results could also support the hypothesis that tax decentralization on the revenue side via more intensive use of environmental taxation may boost economic growth, and they could form the basis for a study of the validity of the double dividend theory in this context. Future research could develope in these lines.

Appendix

See Tables 6, 7, 8 and 9.

Table 6 Parliamentary acts contained in the database of the Spanish Institute of Public Law

Table 7 Types of parliamentary acts contained in the database of the Spanish Institute of Public Law

Table 8 Descriptive statistics for the main variables

Table 9 Correlation matrix for the main variables