Page 13 - IDEA Study 8 2017 Direct subsidies and R&D output in firms

As far as the implementation of this strategy is concerned, the estimation depends on the choice of method for propensity score matching. For this paper, we rely on a matching method that does not explicitly assume a functional form for ATT estimation: propensity score nearest neighbour matching. More specifically, we use propensity score nearest- neighbour matching with three neighbours (NN 3) as the default estimator. As a robustness check, we also estimate nearest-neighbour matching with one (NN 1) and five neighbours (NN 5) as well as NN 3 with a caliper value of 0.01 and kernel matching with the biweight kernel. Where there are tied observations, the match is with all ties. Firms with propensity scores that are not on the common support are dropped. Also firms with propensity scores greater than 0.9 are eliminated from the sample, as suggested by Crump et al (2009). In empirical tools, which are widely used to estimated propensity scores, such as psmatch2 Stata command, standard errors do not take into account that propensity scores are estimated. As a result, researchers have widely used bootstrap to estimate standard errors non-parametrically. Abadie & Imbens (2008) showed that as a general case, bootstrapping is not valid for matching estimators. Abadie & Imbens (2015) suggested a method for estimating analytical standard errors, which was implemented in Stata command teffects psmatch, and which is therefore used as the preferred estimator in this paper (for more details on this estimator see StataCorp 2015). Description of the R&D subsidy programmes The purpose of this section is to provide a brief overview of the direct R&D support programmes, the output additionality effects of which are the focus of our interest in this study. The IMPULS and TIP programmes were administered by the MIT and provided funding during the periods 2004-2010 and 2009-2016, respectively. In the meantime, the TA CR was established in 2009 and launched a portfolio of new programmes including ALFA, which funded projects during the period 2011–2017. All three programmes were organized in annual calls for proposals, primarily targeted business enterprises although research organizations were also eligible, and allowed for proposals from both individual entities and consortia of several partners. All three programmes were financed exclusively from the national budget. The main objectives of IMPULS were to boost the performance of industrial enterprises, particularly small and medium firms, increase competitiveness and modernize 11