‘Better late than never’: the interplay between green technology and age for firm growth

This paper investigates the relationship between green/non-green technologies and firm growth. By combining the literature on eco-innovations, industrial organisation and entrepreneurial studies, we examine the dependence of this relationship on the pace at which firms grow and the age of the firm. From a dataset of 5498 manufacturing firms in Italy for the period of 2000–2008, longitudinal fixed effects quantile models are estimated, in which the firm’s age is set to moderate the effects of green and non-green patents on employment growth. We find that the positive effect of green technologies on growth is greater than that of non-green technologies. However, this result does not apply to struggling and rapidly growing firms. With fast-growing (above the median) firms, age moderates the growth effect of green technologies. Inconsistent with the extant literature, this moderation effect is positive: firm experience appears important for the growth benefits of green technologies, possibly relative to the complexity of their management.


Introduction
Following the 'Porter hypothesis' and the debate over 'whether it pays to be green', studies have shown that by complying with environmental regulations, adopting sustainable practices and eco-innovating, firms can become more competitive (Porter andvan der Linde, 1995, Ambec andLanoie, 2008;Ambec et al., 2013) if not even more profitable (Horváthová, 2010;Ghisetti and Rennings, 2014). However, the impact of green technologies on firm growth has been minimally investigated, especially given the abundant literature on 'standard' innovation as a growth driver (Sutton, 1998;Botazzi and Secchi, 2006;Lotti et al., 2009;Coad and Holz, 2012). 2 Supportive evidence has been mainly obtained by examining the relationship between eco-innovations and firm growth through the lens of the technology-jobs nexus, usually in a non-longitudinal setting (e.g. Gagliardi et al., 2016;Rennings and Zwick, 2002).
However, with few exceptions, these analyses do not address the 'growth premium' attached to green technologies vis-à-vis the non-green ones, nor do they pay attention to the inner complexity and dynamics of the phenomenon.
The present paper aims to close this gap by addressing two research questions. We first draw on the idea from the field of industrial organisation that the growth effect of technology exploitation varies with the pace at which a firm grows, given that growth opportunities and threats change at different growth rates . Hence, we investigate whether the growth outcome of ecoinnovations depends on the firm's pace of growth and on whether the firm is struggling or rapidly growing. To address this research question, we rely on a novel methodological approach: a quantile regression analysis (e.g. Coad and Rao, 2008;Coad and Rao, 2010;Coad et al., 2013) performed by using a fixed-effects estimation technique (Canay, 2011). This technique captures the potentially heterogeneous effects of green (and non-green) technologies on firm growth across different growth rates, while controlling for unobserved heterogeneity.
The second research question examines whether the firm's age influences the growth impact of green technologies. Again, we are informed by the industrial organisation literature (e.g. Barba Navaretti et al., 2014;Distante et al., 2014): we consider age-dependent mechanisms that characterise the firm's capacity to exploit innovation (Coad et al., 2016) and add to them specific ones related to ecoinnovations. By studying the knowledge-complexity implications of eco-innovations (e.g. in terms of risk and financing) and the higher need for technology experience to grasp it (Carrillo-Hermosilla and Konnola, 2010), we investigate whether age moderates how the firm's growth benefits from green technologies.
technologies. These extra returns can be exploited (e.g. re-invested) and, as we posit, yield an employment growth premium to eco-innovators. This argument finds support in three different research streams. At the outset, by extending the debate on Schumpeterian innovation regimes to the green realm (Malerba, 2005;Oltra and Saint Jean, 2009), we argue that eco-innovators could make a more effective 'creative accumulation' of knowledge (i.e. Schumpeter Mark II) than standard ones, and translate economic returns into higher growth opportunities. Given the irreversibility of complying with environmental regulations, investing in green-specific assets and acquiring internal/external green knowledge (Oltra and Saint Jean, 2009;Mazzanti and Rizzo, 2016) and environmental technologies have actually been found to lead to more persistent (eco-) innovation practices and outcomes than standard technologies, with greater opportunities of increasing returns (see Sàez-Martínez et al., 2016;Chassagnon and Haned, 2015).
A growth premium from green vs. non-green technologies is also supported by the literature on the joint improvements of environmental and economic/financial performances of firms (i.e. their 'winwin' strategies). The green-specific mechanisms that increase firm revenues (e.g. green-differentiation of products, access to green demand segments and sale of environmental-control technologies) and reduce costs (e.g. material and energy efficiency, and recycling initiatives) (see Ambec and Lanoie, 2008), provide eco-innovators with improved financial indicators (e.g. Misani and Pogutz, 2015), greater profits (Ghisetti and Rennings, 2014) and, in general, extra economic returns to be turned into higher growth.
Finally, the regulations and policy actions on which eco-innovations depend (the so-called 'regulatory push/pull' effect) also represent an 'extra' driver of growth. 'Polluting' firms at the end of the value chain are legally forced to improve their environmental performances and, in so doing, 'induce' an additional element of 'derived demand' in the upstream producers of green technologies that fuels the latter's growth (Colombelli et al., 2015;Ghisetti and Quatraro, 2013).
In summary, the extant literature seems to imply a growth effect of green technologies vs. non-green ones, which can depend on two scarcely analysed aspects: (i) the pace at which firms grow and (ii) the firm age.
The pace aspect has been confirmed through the use of quantile regressions for standard innovations. Coad et al. (2016), for example, showed that only the fastest-growing firms benefit from standard innovation in terms of employment growth, while this return is actually negative for the slowestgrowing ones. In general, fast-growing firms have been shown to have crucial advantages in the 'jobcreation argument' (for a review see Almus, 2002). First, they are generally smaller, and thus more prone to commercialising their innovations, and younger, and accordingly more in need of investing in the knowledge they miss at the beginning of their businesses. Second, they often operate in technology-intensive sectors and are thus endowed with a larger knowledge base, qualified human capital and technological skills and experience. They also usually have a limited liability legal form, thus showing greater incentives for riskier but also more rewarding, innovations. Finally, their close connection to suppliers, customers and competitors enables them to benefit from an open innovation approach.
As these aspects do not vary by the nature of the relevant technologies and given the absence of specific literature on green technologies, we maintain that the distinction between rapidly and slowly growing firms could be a relevant factor for eco-innovations too. Although in a non-quantile framework, but rather in a dynamic parametric estimation of Gibrat's law (Gibrat, 1931(Gibrat, , 2003, this is confirmed by Colombelli et al (2015). They find that the growth differential between green and generic technologies is actually greater for firms growing at more than 'the average' rate. We interpret this in light of the 'induced' innovation, which is the 'derived demand' from environmental regulations that fuels green sectors.
With regard to the second aspect of our analysis, in the industrial organisational literature, age has a two-fold effect on growth. On the one hand, it is (along with size) an important determinant of a firm's growth potential, with a large (although not yet conclusive) body of evidence favouring younger firms (Haltiwanger et al., 2013;Lawless, 2014). On the other hand, age (along with other characteristics) is a crucial moderating factor of the impact of innovative activity on firms' growth (Audretsch et al., 2014) and on employment growth, in particular (Coad et al., 2016). However, the role of age in the relationship between green technologies and growth has received little emphasis. A sort of 'sin of youth' seems to emerge from the literature on '(eco-)sustainable entrepreneurship' (Dean and McMullen, 2007), in which the comparative analyses of start-ups (young firms) vs. incumbent (old) firms have been very rare, so far, and specific to some sectors (e.g. green electricity and microfinance) (Hockerts and Wustenhagen, 2010). While 'emerging green Davids' usually show higher environmental commitment and attractiveness to sustainable consumers, they often fail to translate their niche market potential into a broad mass market, mainly because of the competition from incumbent 'greening Goliaths', through their 'inner' form of corporate sustainable entrepreneurship (e.g. Bird et al., 2002;Stenzel and Frenzen, 2008).
Other and more general age-related insights emerge from environmental and eco-innovation studies, all suggesting a greater growth potential of mature eco-innovators. First, an older firm can be expected to have an advantage in terms of learning experience against the multidimensionality and complexity that characterises green knowledge (Carrillo-Hermosilla and Konnola, 2010) and new green product development projects (Tsai, 2012). Second, younger firms may be more averse to the growth exploitation of green technologies, as these are often in the early stage of their life cycles and thus marked by greater uncertainty than non-green ones (Consoli et al., 2016). Similarly, young firms could This is the author's post-print copy of the article published on Small Business Economics https://doi.org/10.1007/s11187-017-9939-6 be disfavoured in benefiting from policy instruments for the adoption of green-tech -such as new practices of green public procurement (Parikka-Alhola, 2008) -as these are still marked by uncertainty and require experience in managing demand-pull policy. Third, given the hard collaterisation and information signalling of green investment projects, older firms could be expected to have better access to financing (Schneider and Veugelers, 2010) and be better prepared to cope with the higher cost of eco-innovations without crowding out other growth-driving investments (Hall et al., 2016). Last, but not least, older firms may have an advantage in strengthening their available resources to increase their economic green returns (e.g. through economies of scale) as well as in forming alliances to tap into external resources (e.g. through reputation and market position) (e.g. Cainelli et al. 2015). Similarly, firm maturity could be beneficial for searching, absorbing and transforming external knowledge (Franco et al., 2014) towards adopting the open eco-innovation mode (Ghisetti et al., 2015), particularly when accessing new and foreign markets (e.g. Autio et al., 2000;D'Agostino, 2015). 3 In light of the above aspects, our study attempts to address the firm growth potential of green vs. nongreen technologies, by providing new empirical evidence for two original research questions: (1) To what extent does the association between green technologies and firm growth vary along the conditional distribution of growth rates? (2) What is the role that a firm's age plays in the relationship between green technologies and growth?

Data
The empirical analysis is based on a dataset that has been obtained by combining three different sources (see Appendix A1 for details): (i) the ASIA database of the Italian National Statistical Office The arguments about the growth potential of mature (young) companies that we have just presented refer to green technologies in general terms (i.e. without distinguishing specific environmental targets or technological realms). In the absence of theoretical backing and/or prior empirical findings on the existence of differences for different green technological realms, we distinguish between green and non-green technologies only, without focusing on specific green technologies.
is to investigate the growth premium, if any, offered by green versus non-green innovations and the moderating effect of age in the relationship between innovative activity and growth, our implications will be valid for innovative firms only. 4

Methodology
To address our research questions (see Section 2), we investigate the following relationship: where d t indicates a series of time controls; z i,t-1 is a vector of firm-specific control variables; denotes the unobserved firm specific effects; and is the error term.
Building upon previous empirical works on the relationship between growth and innovation, and given our focal interest in the role of the pace at which firms grow, we employ a quantile regression approach (Coad and Rao, 2008;Kesidou and Demirel, 2012). As is well known, this approach allows for a richer characterisation of the data: it disentangles the relationships between our independent variables and firm growth at different quantiles of the distribution of the growth rates, rather than at the conditional mean only. Further, as is normally the case when investigating firm growth (Buchinsky, 1998;Bottazzi and Secchi, 2003), quantile analysis is preferable over standard least squares for different reasons linked to the distribution of the growth rates in our sample (see Appendix A2).
Most of the applied literature adopting a quantile regression approach stems from cross-sectional settings, and for this reason, controlling for problems of endogeneity arising from unobserved heterogeneity has been difficult. Conversely, we follow recent developments in a stream of the applied econometrics literature that has attempted to overcome this major limitation (Koenker, 2004;Galvao, 2011;Canay, 2011). Specifically, we implement the procedure suggested by Canay (2011), who developed a method to estimate fixed effects quantile regressions for panel data. The solution proposed consists of a two-step estimator. In the first step, we estimate our previous equation (1) as a standard linear panel regression model via the within estimator (Wooldridge, 2010). From this model, we obtain the predicted value depurated from the unobserved heterogeneity component: Table A1 in the appendix reports the difference in one-year employment growth, employment and age between our sample and the overall population of Italian companies (source: ASIA-ISTAT). We test whether the means for the variables above are statistically different between the two groups. In line with the approach of Gagliardi et al. (2016), who employed similar data, firms in our sample are older and bigger, while there is no significant difference in terms of employment growth.
where ̂= [ ℎ − ℎ � ] is an estimate of the unobserved heterogeneity term. In the second step, a standard quantile regression model is implemented in which the transformed dependent variable above ( � ) is regressed on our relevant independent variables (Koenker and Hallock, 2001).

Variables
Our dependent variable is the growth rate of employees (Growth it ), calculated as the difference between the logarithm of firm i's employees in year t and the logarithm of employees in year t-1 Coad, 2010) (see Appendix A3 for details). In addition to theoretical reasons (see Section 2), this choice has also empirical motivations. Unlike other measures (Delmar et al., 2003) such as sales growth, employment growth can capture growth performance in recently constituted firms (Clarysse et al., 2011).
We have three main independent variables: (i) Pat Green i,t-1 , which is the logarithm of the stock (at time t-1) of environmentally friendly technologies (plus 1), filed by firm i; (ii) Pat Nongreen i,t-1 is the logarithm of the stock of non-environmentally friendly technologies (plus 1); (iii) Age i,t-1 which measures the (log transformed) age of company i at time t-1, with the difference between the current and its constitution date.
Despite its limitations as an innovation proxy, patent data has been used by most of the recent research on eco-innovations because they are, on the one hand, more widely available and more informative than R&D about their environmental nature and, on the other hand, a more robust indicator than questionnaire-based measures (Arundel and Kemp, 2009;Berrone et al., 2013). For the identification of 'green patents' in particular, we have relied on Marin and Lotti (2016) (see Appendix A3). Both green and non-green technological variables are defined as stocks, rather than flows (see Appendix A3 for details). We do so because we expect a firm's rate of growth to be affected by the knowledge cumulated over time and not only by its variation added in the recent and/or current period (Bloom and Van Reenen, 2002;Hall et al., 2005). This also helps reduce, at least partially, the possible confounding effect of the persistency in technological leadership (Denicolo', 2001) on firm growth, which cannot be addressed by controlling for unobserved heterogeneity (Peters, 2009). 5 We control for a set of variables that are often included in growth rate regression models: (i) investments in tangible (Inv Tang i,t-1 ) and intangible (Inv Intang i,t-1 ) assets (at time t-1), recognised by the literature to have an important role in 'accounting' for the firm's capacity to grow (Hall, 1987); (ii) a measurement of size in terms of number of employees of firm i at time t-1 (Emp ,t-1 ), used to retain the implications of the Gibrat's law (e.g. Evans, 1987;Hall, 1987;Calvo, 2006); (iii) an Herfindahl-Hirschman index of industry concentration (Herfindahl index jt ), which has been often found to play a relevant role with respect to firms' performance (Kaniovski & Peneder 2008) (see Appendix A3 for details).
Finally, we include a set of eight dummy variables to control for year effects. Table 1 shows the variables included in the analysis and their sources.
[ Descriptive statistics of the variables employed in the empirical exercise are reported in Table 2. Table 3 reports the bivariate correlations of the variables considered in the analysis. No indication of significant multi-collinearity amongst the independent variables was found (i.e. the variance inflation factor ranges from 1.02 to 2.62, well below the threshold level of 5). [

Results
The results of the quantile fixed effect estimations are presented in Table 4, which shows the baseline model, and in Table 5, which incorporates the role of the firm's age as a moderating factor in the relationship between environmental (and non-environmental) patents and firm's growth. 6 Starting with the controls, as expected (e.g. Coad and Holz, 2012), both tangible and intangible investments significantly drive firm growth. As for size, smaller companies show greater growth opportunities and capacities, in agreement with the entrepreneurship literature (Acs and Audretsch, 2006). An increase in market concentration (Herfindahl index) seems to favour firm growth, although the effect is significant -and positive -only at the 50 th and 90 th percentile. This result resonates well with the characterisation of the 'Schumpeter Mark II' pattern of innovation (Malerba and Orsenigo, 1995) marked by an oligopolistic context with high technological opportunities and appropriability -which 6 Both tables report results of the Parente and Santos Silva (2016) test to determine whether intra-industry correlation affects the standard errors in our estimates. Results show that, apart from the 50th percentile, all other percentiles (10th, 25th, 75th and 90th) are affected by intra-cluster correlation. Results reported in the tables therefore use cluster-robust standard errors at industry level (NACE rev. 1.1 2 digit codes). This is the author's post-print copy of the article published on Small Business Economics https://doi.org/10.1007/s11187-017-9939-6 the former literature actually identified in a section of the Italian national system of innovation (Malerba, 1993), and that here appears to be represented by fast-growing companies.
As far as firm's age is concerned, the results of the standard literature on the growth advantages of newly created companies (Coad et al., 2013;Barba Navaretti et al., 2014) appeared reversed across all quantiles in Table 4: unexpectedly, older companies grow more than younger ones. This result can be only be partially explained by the specificity of our quantile methodology. Most likely, its explanation lies in the characteristics of our sample. Our sample consists of innovation-oriented, manufacturing firms operating in a national context, where new-born firms face structural difficulties in taking off and surviving (Audretsch et al., 1999), and where established incumbents usually obtain the most radical innovation outcomes (Malerba, 1993). The importance of banks in financing innovation (Benfratello et al., 2008) also plays a key role in the Italian context, and mature firms are more capable (e.g. by reputation) of developing borrowing relationships for their innovations (Gregory et al., 2005;Hartarska and Gonzalez-Vega, 2006;Carpenter and Rondi, 2000;Magri, 2009). In the same context, firm internationalisation and innovation often entails a strong increase of competitive pressure and failure risk (Giovannetti et al., 2013), and maturity and foreign market experience increase the chance of postinternationalisation survival (Autio et al., 2000;Carr et al., 2010). Finally, the regime of 'creative accumulation' (Schumpeter Mark II) that characterises the most competitive Italian industries (e.g. motor-vehicles and non-electrical machinery) (Malerba and Orsenigo, 1995) could lend knowledge accumulation and innovation persistence experienced by mature firms a larger impact, also on the growth performance of sectoral systems of environmental innovation (Chassagnon and Haned, 2015; Oltra and Saint Jean, 2009).
We now come to the core of our analysis. As Table 4 shows, the positive and significant coefficients of both Pat Nongreen and Pat Green across the whole set of percentiles confirm the role of green technology as a driver of firm growth. This finding supports and extends the emerging evidence on the businessenvironmental win-win situations enhanced by environmental practices. Indeed, as we expected, the increase of product value and the reduction of production costs they entail (Ambec and Lanoie, 2008), actually seems to translate into higher growth.

A more relevant aspect to consider is the comparison between the coefficients of Pat Green and Pat
Nongreen across the quantiles of the distribution. This comparison is crucial for assessing whether green technologies provide a growth premium with respect to non-green technologies or whether, instead, the effects of green and non-green patents are not different. By running appropriate statistical tests on the difference between the two coefficients (tests are reported in Table 4), we found that the difference between Pat Green and Pat Nongreen is not homogeneous across the quantiles. Specifically, it emerges This is the author's post-print copy of the article published on Small Business Economics https://doi.org/10.1007/s11187-017-9939-6 that for the 25 th , 50 th and 75 th percentiles, green technologies have a significantly larger effect (at a 99% level of confidence) on growth than standard technologies, while for the 10 th and the 90 th percentiles green and non-green patents have statistically similar effects on firm growth. In brief, the growth premium of green over non-green technologies is not unlimited and weakens when innovation efforts are pursued either to survive (struggling firms) or to remain among the growth 'superstars' (gazelles).

[TABLE 4 ABOUT HERE]
The picture becomes more nuanced when we introduce interaction terms to capture the moderating role of age (Table 5). While Pat Nongreen remains positive and significant, except for the 10 th and 25 th percentile, Pat Green, per se, is not positive and becomes negative and significant for the 75 th and 90 th percentiles. However, the effect of Pat Green, as shown in Table 5, should be understood in relation to the age of the company, given the contribution of the interaction between Pat Green and Age, which is always positive and significant, except for the 10 th and 25 th percentiles.

[TABLE 5 ABOUT HERE]
In other words, we find an apparently exclusive capacity of older firms to translate eco-innovation into growth. This is the second important result of our analysis, which is possibly linked to the points of firm maturity and eco-innovations discussed in Section 2. First, as we said, older firms may be better equipped to evaluate the uncertainty/risk and the actual marketability of their eco-innovations, irrespective of their likely disadvantages in terms of organisational inertia and learning impediments (Majumdar, 1997;Sorensen and Stuart, 2000;Criscuolo et al., 2012). Second, owing to better access to finance (Schneider and Veugelers, 2010), older firms can have a higher capacity to cope with the cost of eco-innovating (Gagliardi et al., 2016) and with the resources needed to engage in signalling, labelling and certification efforts, which are often required to extract value from investment in green innovations (Ambec and Lanoie, 2008). Third, older firms might have greater pressures and incentives for renewing their older capital vintages in an eco-sustainable manner -for example in responding to a policy constraint (Ruth et al., 2004) -also in light of their greater capacity to exploit internal economies of scale and external knowledge sourcing (Herriott et al., 1985;Levitt and March, 1988;Ghisetti et al., 2015). Finally, the persistence of the learning and innovation patterns that characterises greentechnologies (Sàez-Martínez et al., 2016, Chassagnon andHaned, 2015) can 'reserve' the growth impact to firms that are capable of reaping the benefits of their path-dependence.
While favouring older firms, the implications of our results for entrepreneurial growth are quite discouraging. When attempting to pursue the heavily uncertain path of growth (e.g. Coad et al., 2013), This is the author's post-print copy of the article published on Small Business Economics https://doi.org/10.1007/s11187-017-9939-6 young companies are able obtain short-term gains only from standard innovations, unable to capture the external benefits associated with environmental protection and arguably less distant from the traditional industrial knowledge base (Ghisetti et al., 2015). Interestingly, these gains occur for the central quantiles of the distribution, as can be noticed from the negative and significant coefficients on the interaction term Pat Nongreen X Age in the 50 th and 75 th percentiles. For rapidly or slowly growing companies, age does not moderate the growth-driving effects of non-green technologies.
We further qualify the additional effect of green technologies compared to non-green technologies for the quantiles where the interaction between Pat Green and Age is significant (Table 5, percentiles 50th to 90th). For young firms (those with less than five years for the 50 th and 75 th percentiles of growth rate and below ten years for the 90 th percentile), a stronger association can be determined between nongreen technologies and firm growth vs. green technologies and growth. On the contrary, for more mature firms (i.e. those with more than twenty years for the 50 th and 75 th percentiles of growth rate and above thirty years for the 90 th percentile), green technologies exert a higher effect on firm growth compared to non-green technologies ( Figure A3 in the Appendix provides a graphical representation).
These quantile-specific effects further confirm the choice of a quantile approach as the most suitable to identifying the different effects of the interplay between green technology and age on firm growth.

Conclusions
In this paper, we examined the capacity of green technologies to sustain firm growth, building upon the idea that a firm's capacity to grow is closely linked to its ability to master technological knowledge and capture the value of the innovation (Mansfield, 1962;Scherer, 1965). While an extensive body of industrial organisation and innovation literature has addressed the growth impact of technology (e.g. Audretsch et al., 2014), only a few studies have examined the relationship between green technologies and firm growth. Our contribution is novel for two reasons. First, we assessed whether green technologies, compared to non-green technologies, affect the growth of firms with different growth paces (e.g. struggling or rapidly growing). Second, we considered whether green-based growth is affected by a firm's age. We adopted a novel econometric approach, combining panel fixed effects with quantile regression estimations. We thus simultaneously controlled for unobserved heterogeneity (which is likely to affect firm growth) and for the heterogeneity of the growth process, along the distribution of growth rates.
Our analysis of a large sample of Italian firms between 2000 and 2008 confirms the vital role of green and non-green technologies in fostering firm growth, as measured by the growth of employment. This is the author's post-print copy of the article published on Small Business Economics https://doi.org/10.1007/s11187-017-9939-6 Moreover, the results indicate a 'win-win' situation as green technologies exert superior effects on growth than non-green ones. The possibility to enter green markets, to decrease production costs because of greater resource efficiency (e.g. reduced material and energy use) and to reinvest the relative extra returns from eco-innovating can justify this result. However, our analysis shows that the superior effect of green technologies does not extend to the extreme percentiles of the growth rate distribution.
The second contribution of the paper pertains to the moderating effects of age: the green-growth path is mainly taken by mature firms (age higher than 20 or 30 years), with the exception of the slowgrowing ones. Hence, more mature companies seem to be better equipped to transform green technology into growth. Although further research is required, we contend that greater experience, fewer financial constraints and exemption from issues related to the liability of newness (e.g. Freeman et al., 1993) -a set of aspects that are particularly relevant in the Italian context -allow older firms to engage successfully in complex and uncertain technological projects, such as environmentally-related ones. These results are partially balanced by the positive effects on young companies of non-green technologies, which trigger short-term firm growth (for the central quantiles), possibly because of their less complex and costly nature.
These results hold important implications both for management and for policy. Extracting value from green technology and transforming it into higher growth is not a 'one-size-fits-all' strategy. On the one hand, struggling firms might not find it viable to engage in more complex and costly green technological projects. On the other hand, for the elite group of fast growing companies, a green orientation might not add to their portfolios of already outperforming, and possibly unique -compared to their competitors -technological capabilities. As said, our results suggest that the process of greenled growth is a complex and costly one: only older companies are sufficiently broad shouldered to pursue a growth path based on environmental technology.
Building on our evidence, we also believe that our results have relevant implications for policy makers.
If their short-run objective is to maximise the social impact of public resources in supporting the transition towards green forms of production, the main beneficiary group should be made of relatively established firms, rather than start-ups. This aspect should be considered when implementing policies favouring innovative start-ups (e.g. Mason and Brown, 2013;European Commission, 2014).
In conclusion, this is a first attempt at providing empirical evidence for the relation between firm growth and green technology. From a policy implications perspective, future research should investigate the mechanisms that make growth particularly problematic for young companies. Further research should also go beyond patenting firms: patent data, although the most diffused source of This is the author's post-print copy of the article published on Small Business Economics https: //doi.org/10.1007/s11187-017-9939-6 information for defining continuous firm-level innovation variables (e.g. Gagliardi et al., 2016), does not capture all the innovations introduced by firms ( Griliches, 1990). This is the author's post-print copy of the article published on Small Business Economics https://doi.org/10.1007/s11187-017-9939-6   This is the author's post-print copy of the article published on Small Business Economics https://doi.org/10.1007/s11187-017-9939-6

A1 -Data
Three data sources have been used to create our dataset: 1) The first is the ASIA database of the Italian National Statistical Office (ISTAT), which contains information on the structural characteristics of the population of Italian companies. From this source we retrieved information related to the industrial sector, the number of employees and the date of birth for the population of Italian business firms over the period of 2000-2011. Due to the data availability of the other relevant data sources (see Table 1), we restricted the period of interest to 2000-2008. Moreover, building upon other firm-level studies (Geroski et al., 2010;Mata and Portugal, 2002;Coad and Rao, 2011), we considered a firm to have ceased operation if absent from the records for three consecutive years.
2) Our second source of data refers to balance sheet information, obtained from the Bureau van Dijk AIDA database for the period of 2000-2008 and is used to retrieve the: investments in tangible assets, investments in intangible assets and market shares.
3) Finally, following previous studies in the analysis of eco-innovations (see, among the others, Lotti & Marin, 2013;Marin, 2014), we relied on the Worldwide Patent Statistical Database (PATSTAT) to retrieve patent data information for the names of the assignees, filing dates and technological classes.

A2 -Methodology
In general, the distribution of growth rates at the firm level is recognised to be highly non-linear and considerably heavy-tailed. Thus a quantile approach could provide more robust and efficient alternatives to OLS when the error term is non-normal, and in the presence of outliers and fat-tails.
Our empirical setting is not an exemption from the pattern outlined above. The distribution of growth rates (defined in Section 3.3) reveals this non-normality nature, showing a kurtosis well above the accepted threshold value of three (our null hypothesis of a kurtosis equal to zero is rejected at the 99% confidence level).
Further support for this pattern is provided by Figure A1 and Figure A2. Figure A1 illustrates that growth rates for our sample of companies is exponentially distributed (Stanley et al., 1996), therefore showing a tent-shape distribution (Bottazzi and Secchi, 2003). Figure A2, plotting the quantiles of the growth variable against the quantiles of a normal distribution, shows that growth rates deviate from a normal distribution both at the upper and lower tails.
The quantile approach is then motivated in order to overcome the distributional issues discussed above.

[FIGURE A2 ABOUT HERE]
This is the author's post-print copy of the article published on Small Business Economics https://doi.org/10.1007/s11187-017-9939-6