https://orcid.org/0000-0001-8857-8314
![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
ABSTRACT
This paper decomposes the Spanish aggregate R&D distribution to disentangle the contributions of R&D public financing, gazelle firms, and financial constraints. Applying the Chernozhukov, Fernández-Val and Melly (2013) distribution regression approach, we estimate the contributions of these components at each point of the distribution. The analysis is carried out for two periods, pre-crisis 2004–2008 and post-crisis 2009–2014. We thereby introduce a comparative perspective that allows us to consider possible business cycle effects. Our findings show that the main explanatory factors of the significant post-crisis drop in Spanish aggregate R&D are changes in the public financing scheme and the decreased contribution of gazelles. Our results provide a rigorous analysis of Spanish R&D, hint at a possible transmission channel for reduced business dynamism, and offer interesting insights for policymaking.
Acknowledgements
We thank Alex Coad, José García-Quevedo, Clemens Domnick, Francesco Rentocchini, Alexander Tübke, James Gavigan, Elisabeth Nindl, Ramón Compaño, Lorenzo Napolitano, and other participants of the 26th SPRU PhD forum (Brighton, 14–15 May 2020), of the Workshop on Industrial and Public Economics (Reus, 4–5 February 2021), of the XXIII Applied Economics Meeting (online, 3–4 June 2021), EU-SPRI conference (online, 9–11 June 2021) and of the EC-JRC Seminar (Seville, 20 June 2022) for useful feedback.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 Over the last decade a wide number of investigations has highlighted the importance of multi-level studies and proposed estimations based on micro (or sectoral) data, but which can be quantified in relation to the most common aggregate measures (Gabaix Citation2011; Acemoglu et al. Citation2012; Di Giovanni and Levchenko Citation2010; Di Giovanni et al. Citation2014; Carvalho and Gabaix Citation2013; Foerster et al. Citation2011). These studies tend to rely on either input-output matrices or quite long-term micro data.
2 The access to data from the Panel de Innovación Tecnológica (PITEC) promoted a proliferation of works on various aspects related to the drivers and effects of R&D among Spanish companies. Particularly, they covered issues such as the drivers of innovation, firm growth and high-growth firms, R&D and cooperation strategies, barriers to innovation, and the influence of the business cycle (López-García et al. Citation2013; Costa-Campi et al. Citation2014; García-Quevedo et al. Citation2018), among other topics.
3 This group of firms will be formally defined in Section 2, but we anticipate that its definition derives from Birch (Citation1987) and all the subsequent works on high-growth firms and young innovative companies.
4 This question, however, is outside the scope of this paper. Our aim is to disentangle the effect that the public financing scheme has on each part of the R&D distribution.
5 Indeed, one could argue that the inclusion of these companies is almost tautological, given their definition. Nevertheless, we argue that YICs, together with HGFs, are extremely important actors in the economy and offering results that hold for a larger group of firms is also a way to offer more powerful and easier to implement policy suggestions. Given the erratic nature of firms, and of dynamic ones especially, this argument is even more prominent. For the sake of robustness, we repeated the estimations of the paper also without YICs, and the statistically significant results are extremely correlated. Results are available upon request to the authors.
6 See Table A-1 for a description of the representativeness of the total national investment in internal R&D.
7 Particularly, we remove firms declaring 1€ of sales while investing large amounts on R&D and firms declaring R&D intensities higher than 100%.
8 Sectors follow the CNAE-2009 classification, which can be directly linked to the more detailed NACE classification.
9 Typical applications of the original method in the industrial literature regard the decomposition of productivity (i.e. Fariñas and Ruano Citation2004).
10 Kitagawa (Citation1955) was the first to propose this type of decomposition.
11 Despite its advantages, the approach has one main limitation: counterfactuals may not be enough for causal insights. Particularly, the present analysis cannot be considered out of the influence of possible confounding and/or selection on variables, which prevent us from interpreting the results as causal. Thus, the results are commented in an associative manner.
12 It is possible to obtain the same decomposition for the contractionary period by substituting the appropriate time indexes.
13 The results of these decompositions can be order independent. We run the estimations for each possible order of the decomposition. Table A-4 shows the results remain unchanged.
14 Firm-level public financing data are available only for 2003 and 2005. For the baseline version, we use 2003 public financing data but, in the robustness checks, we explore the sensitivity of the findings to alternative choices. All relevant results show correlations higher than 90%, thus we can safely conclude that this does not affect our findings.