Page 26 - IDEA Study 8 2017 Direct subsidies and R&D output in firms

model of propensity score matching and a conditional difference-in-differences model, as suggested by Blundell and Costa Dias (2000) and applied, for example, by Görg and Strobl (2007). Essentially, this method explores not only the differences between treated and untreated firms but also the differences before and after participating in the programme. The dependent variable in these estimates is the difference in the firm’s probability to apply for IP protection between the year before funding from the programme call began and the respective year afterwards, i.e. the difference (t) - (t-1); (t+1) - (t-1); and (t+2) - (t-1). Table 5 gives the updated results. Everything remains as before, except that the conditional difference-in-differences model is used instead of the conventional estimation. There are two main differences compared to the previous results. First and foremost, none of the estimated effects on international IP protection are statistically significant at the conventional levels, not even weakly as before, thus in statistical terms we cannot reliably rule out the possibility that the programmes had zero impact in this respect. Second, the effect of IMPULS subsidies on Czech IP protection is not statistically significant, which suggests that this programme did not make a tangible difference. Meanwhile the results for TIP and ALFA as regards Czech IP protection remain similar, albeit the statistical significance of the latter has decreased by a notch.12 12 Note that the negative figures for Czech IP protection in both treated and untreated firms that are estimated using the difference-in-differences framework in the third year (t+2) of ALFA primarily testify to the fact that, as explained above, the data in the latter period are incomplete due to the publishing delay. 24