Page 30 - IDEA Studie 07 2023 TACR

ARE SUBSIDIES TO BUSINESS R&D EFFECTIVE? REGRESSION
DISCONTINUITY EVIDENCE FROM THE TA CR ALFA PROGRAMME IDEA 2023
Innovation output additionality and economic performance
Table 3 provides the results for innovation output additionality in terms of patenting, and economic performance in terms of employment, sales, and labour productivity. Across the board, regardless of firm size and timing of the effects, the results indicate that there is little to say with regards to the impact of the subsidies in these respects. In fact, most of the coefficients prove to be actually negative, although they are estimated too imprecisely to be statistically significant, which signals that the recipients did not manage to convert the subsidy into something of significant benefit for their development that would not have materialized without the subsidy.
A pessimistic interpretation of this (non)finding is that the additional projects realized by the SMEs thanks to the subsidy were of relatively low quality and, thus, have low economic returns. A more nuanced interpretation, however, is that the additional projects have low private returns, because otherwise they would be undertaken even without the subsidy, but their social returns are greater due to spillover effects, which we do not observe. In this respect, it needs to be noted that what matters most is not necessarily the direct impact of the subsidies on their recipients, but knowledge spillovers to unsup- ported firms that the newly stimulated activities generate (see Bloom et al., 2013, Lychagin et al., 2016, and Zacchia, 2020 for recent evidence on the magnitude of, and mechanisms for, such spillovers).
Most importantly, however, the typically small amount of a subsidy relative to the size of the recipient, and the limited size of our sample, mean that we cannot rule out the possible existence of some impacts on economic performance, especially in the SMEs, albeit too small in magnitude to be measurable with the data at hand. The standard error on the estimated impact on sales is 0.1, which implies that the true effect would have to be greater than 0.165 to be detected at a 10% significance level (a t-value of 1.65). For a subsidised SME with average (median) sales, this would require an increase in sales by about 39 (24) million CZK. But, given that the average annual subsidy to SMEs supported in the ALFA programme is about 1 million CZK, this would require an annual rate of return to the additional R&D expenditures generated by the subsidies on the order of 100s of %. It is not realistic to expect rates of return of this magnitude.
28