Regional Innovation Policy and Innovative Behaviour: Looking for Additional Effects

Abstract

This paper aims to evaluate the additionality of innovation policy in terms of innovative behaviours at the regional level. Innovative behaviours are identified both within and across firm and regional boundaries. The role of policy is evaluated for a sample of firms in the Italian region of Emilia–Romagna (ER), exploiting an original, survey-based data set. Propensity score matching is applied to investigate the effects of an innovation subsidy. Funded firms are found to be more likely to upgrade their competencies, compared with similar non-subsidized firms. On the other hand, in most cases, innovation cooperation with other business partners within or outside the region is not significantly affected by policy. Ultimately, the investigated innovation policy in the ER region seems to show what might be termed “cognitive capacity additionality”, rather than “network additionality”.

Acknowledgements

The authors are grateful to Paolo Pini for allowing access to the ER data set and to Silvano Bertini for support in information and data gathering. The ER region is acknowledged for providing financial support for the survey. Preliminary versions of this paper were presented at the 2011 EUNIP International Workshop (Florence, Italy), the 2011 EAEPE Conference (Vienna, Austria) and the INGENIO (CSIC-UPV) 2011 Seminar Series (Valencia, Spain). The authors thank the participants in these events for their comments and suggestions as well as two anonymous referees for their helpful comments and suggestions to improve the paper. The usual caveats apply.

Notes

1. In order to maintain focus, analysis of input and output additionality in relation to the policy is not addressed. The policy scheme investigated was designed to increase regional firms’ cooperation with research organizations (e.g. universities and research institutes). The additionality of this behaviour is also excluded in order to focus only on what we term the “indirect” behavioural effects of policy. The effects of a similar policy related to cooperation with research organizations are investigated in Marzucchi et al. (2012).

2. This is an extension of the more standard ideas of “input” additionality and “output” additionality of a policy. In brief, the former refers to the additional resources that targeted firms can be induced to invest in innovation, with respect to non-targeted firms. The latter instead refers to the additional outcomes that targeted firms could be led to have with respect to non-targeted firms (Georghiou & Clarysse, 2006).

3. Falk (2007) distinguishes among the ideas of scope additionality, cognitive capacity additionality, acceleration additionality, challenge additionality, network additionality, follow-up additionality and management additionality.

4. In brief: a high density of SMEs, co-located in specialized local production systems with diffuse social capital (i.e. industrial districts); deep-rooted unionism especially strong in the most industrialized provinces (e.g. Reggio Emilia); and articulated institutional set-up of business and research organizations.

5. Regional Innovation Scoreboard (http://www.proinno-europe.eu/page/regional-innovation-scoreboard) for 2004 and 2006.

6. For an extended illustration of the history of this instrument, see Marzocchi (2009).

7. In defining the group of non-subsidized firms to be used as controls, we did not discriminate between non-successful applicants and non-applicants. Indeed, given that this discrimination is not required by the rationale of the underlying econometric procedure, this allowed us to keep a higher number of observations and, thus, to enrich the information on the counterfactual.

8. As we will see in the next section, these covariates will be used to carry out a probit estimation, which is functional and propaedeutic to the PSM we will undertake. For this reason, this set represents a careful choice among the variables at our disposal. Indeed, it is our theoretically grounded attempt at mediating between two approaches—a conservative (Augurzky & Schmidt, 2001; Bryson et al., 2002) and an inclusive (Rubin & Thomas, 1996) one—which in the literature have been shown to suffer from econometric problems in the estimation of the propensity score. Finally, it should be noted that, although the fact they are measured before the policy implementation reduces endogeneity problems, the lack of a panel data structure in our sample does not allow us to eliminate, completely, issues related to the selection on unobservables (Caliendo & Kopeinig, 2008).

9. Unfortunately, disaggregated data for the two kinds of expenditures were not available. On the other hand, studies have been emerging recently on their complementarity nature in the current open-innovation and demand-led paradigm (Perks et al., 2009).

10. Short-term debt is considered here to be probably more relevant than long-term debt, given the contingent nature of the decision to plan an R&D project and, thus, apply for a subsidy.

11. One of the dummies (GEO1) captures firms based outside the regional borders, but having at least a production unit in the region.

12. One just needs to think about its very common “picking-the-winner” strategy (Cerulli, 2010).

13. In particular, the five nearest neighbours, the caliper and the kernel, for which, see Becker and Ichino (2002); Cameron and Trivedi (2009); Smith and Todd (2005); Caliendo and Kopeinig (2008).

14. This guarantees the presence of suitable counterfactual firms for each treated (Caliendo & Kopeinig, 2008; Smith & Todd, 2005). Following Caliendo and Kopeinig (2008), we also impose the common support condition with a “minima and maxima” comparison. In addition, a 1% “trim” is applied to the five nearest-neighbours matching.

15. Drawing on Caliendo and Kopeinig (2008), we employ a set of tests (Pseudo-R² test, likelihood-ratio test on joint significance and a regression-based t-test on differences in covariate means). These tests largely support the quality of the matching.

16. Given its functional role with respect to the PSM, this is the only aspect of the probit which matters here. On the other hand, the meaning of the coefficients is not the primary interest of this paper. Accordingly, we avoid quantifying the marginal effects of the covariates as well.

17. R&D could equally increase the willingness and capacity of firms to apply for the policy. Unfortunately, we cannot distinguish whether previous engagement in R&D increases awareness of the need to innovate, and thus the interest/propensity to submit projects, rather than the capacity to present more promising and well-planned proposals.

18. On the same occasion, representatives of policy-makers reported that other firms, not necessarily SMEs, resorted to regional funding, being unable to participate in other policy programmes (e.g. because the calls for applications were already closed).