Green start-ups and local knowledge spillovers from clean and dirty technologies

Although interesting, it is beyond the scope of the current work to systematically test for the differences that could arise from the choice of classification. (We selected t) The WIPO IPC green inventory was selected since it is currently a frequently used and well established classification of green technologies. OECD has also developed the OECD Indicator of Environmental Technologies (OECD, 2011), based on the International Patent Classification (IPC), which features seven environmental areas, i.e. (a) general environmental management, (b) energy generation from renewable and non-fossil sources, (c) combustion technologies with mitigation potential, (d) technologies specific to climate change mitigation, (e) technologies with potential or indirect contributions to emission mitigation, (f) emission abatement and fuel efficiency in transportation, and (g) energy efficiency in buildings and lighting. At the same time, the European Patent Office (EPO) is working on completing its own classification system (ECLA) to assign each patent a green tag, depending on the environmental aim of each patent. So far, EPO has allowed technologies to be tagged according to their adaptation or mitigation to climate changes (Y02), in terms of buildings (Y02B), energy (Y02E), transportation (Y02T) and the capturing, storage sequestration and disposal of GHG (Y02C). Costantini et al. (2013b) have recently pointed out the shortcomings of classification methods based on efforts to collect the IPCs that are potentially related to green technologies in one place. Focusing on the biofuel sector, they have shown that the WIPO Green Inventory is likely to overestimate the number of patents that have been assigned due to the fact that IPCs are not specifically designed to identify this narrow and very specific domain. Clinical analyses, based on a keyword search and validations by experts, are likely to yield finer grained classifications. Nonetheless, the WIPO Green Inventory was chosen for this work, due to the wide scope of this analysis, which encompasses many different kinds of green technologies.

The Data are available to the public on the http://​startup.​registroimprese.​it/​ web-site. The data used in this paper were updated to May 2015.

However, we do not deny that local markets are not homogenous with respect to size, and that this can introduce some biases in our results. For this reason, as we specify below, we introduced the employment level in the province among the control variables.

The limits of patent statistics as indicators of technological activities are well known. The main drawbacks can be summarized as: their sector-specificity, the existence of non-patentable innovations and the fact that they are not the only protecting tool. Moreover, the propensity to patent tends to vary over time as a function of the cost of patenting, and it is more likely to feature large firms (Pavitt 1985; Griliches 1990). Nevertheless, previous studies highlighted the usefulness of patents as measures of production of new knowledge. Such studies show that patents represent very reliable proxies for knowledge and innovation, compared to analyses that draw upon surveys that have directly investigated the dynamics of process and product innovation (Acs et al. 2002). Apart from the debate on patents as an output rather than an input of innovation activities, empirical analyses have shown that patents and R&D are dominated by a contemporaneous relationship, thus providing further support to the use of patents as a good proxy of technological activities (Hall et al. 1986).

4-digit technological classes have been used in the calculation.

It should be stressed that in order to compensate for the intrinsic volatility of patenting behavior, each patent application has been made to last five years in order to reduce the noise induced by changes in the technological strategy.

The following Nace Rev. 2 sectors have been classified by ISTAT as ‘dynamic’: CE-Chemicals; CF-Pharmaceuticals; CI-Computers and electronic and optical products; CJ-Electric apparatus; CL-Transport; M – Professional, scientific and technical activities; R – Arts, entertainment, recreation; S – Other service activities.

It is worth noting that we have also checked for the existence of autocorrelation in the dependent variables, by means of Wooldridge’s test for autocorrelation in panel data. In all cases, (we obtained) statistics of between 1.4 and 1.5 were obtained, which do/did not allow us to reject the null hypothesis of no first-order autocorrelation.

Additional regressions have been run separately, in which GT_STOCK and NOGT_STOCK have been included. The results confirm the patterns shown in Table 5.

It is worth recalling that related and unrelated knowledge variety/ are not opposites, but orthogonal in their meaning (Frenken et al. 2007; Castaldi et al. 2015). In principle, a NUTS 3 region can be characterized by both high RKV and UKV. These would be regions that are diversified into unrelated technological categories, while also being diversified into many specific classes in each of these categories. . It is also worth stressing that empirically related and unrelated variety tend to correlate positively (see Table 3; see also Frenken et al. 2007; Quatraro 2010; Boschma et al. 2012; Hartog et al. 2012).