Page 24 - IDEA Study 8 2017 Direct subsidies and R&D output in firms

Propensity score matching First, we need to find a group comparable to the programme participants. For this purpose, we use the firms’ observable characteristics to estimate a subsidy programme participation equation, from which we derive a propensity score that is at the core of the matching procedure. The results of the participation equation largely reproduce, in a multivariate framework, the patterns that were detected above (see Appendix Table A2 for details). All three programmes ceteris paribus favoured larger firms that were more solvent, applied for IP protection and had participated in subsidized R&D projects in the recent past. Some of the programmes were prone to support more productive firms, while conflicting evidence emerged on the age of firms. Differences across legal forms, industries and time trends are also accounted for. Overall, these variables are confirmed to be the salient features that explain a firm’s probability to participate in the subsidy programmes. Second, we perform traditional propensity score matching using the NN 3 estimator. Table 4 provides the baseline results. The effects of programme participation on the propensity to apply for IP protection are derived separately for protection in the Czech and international contexts, for the individual subsidy programmes and for the first (t), second (t+1) and third (t+2) years of funding from the respective programme calls. The estimated mean of the “treated” programme participants is in the fourth column and the estimated mean of their matched non-participating counterparts is in the fifth column. At the heart of our interest is the magnitude and statistical significance of the difference between these two figures reported next that indicate the effect of programme participation. N denotes the total number of observations. 11 Our main finding is that once we have constructed a comparable control group, the differences in propensity to apply for IP protection between the treated and untreated firms decrease noticeably compared to the “unconditional” differences presented above. The results indicate that the subsidy programmes had highly statistically significant additionality effects on IP protection in the Czech context, at least as far as the second (t+1) and most notably third (t+2) years of funding are concerned, but largely inconclusive effects on IP protection in the international 11 The diagnostics of the matching procedure, which turn out to be satisfactory, are presented in Appendix Tables A3 and A4 and Figures A1-A5. 22