Incorporating innovation subsidies in the CDM framework: empirical evidence from Belgium

ABSTRACT

This paper integrates innovation input and output effects of R&D subsidies into a modified Crépon–Duguet–Mairesse (CDM) model. Our results largely confirm insights of the input additionality literature, i.e. public subsidies complement private R&D investment. In addition, results point to positive output effects of both purely privately funded and subsidy-induced R&D. Furthermore, we do not find evidence of a premium or discount of subsidy-induced R&D in terms of its marginal contribution on new product sales when compared to purely privately financed R&D.

KEYWORDS:

• CDM model
• R&D
• subsidies
• innovation policy

JEL CLASSIFICATION:

• C14
• C30
• O38

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. The maximum subsidy rate is usually limited to 50% of the total project cost in the EU, but there are exceptions for small- and medium-sized firms and for firms in structurally weak regions (so-called Objective 1 regions) where the maximum subsidy rate may exceed 50%.

2. Note that we focus in this paper solely on the effects subsidies may have on the beneficiary firm itself. We do not investigate whether a potential increase in R&D spending generates positive spillover effects to other members of society which is the standard justification for government intervention in the market for R&D.

3. Firms receiving a subsidy might be different from companies that do not receive a subsidy: some firms might be more likely to apply for public funding than others; some firms might consider the administrative burden or the information sharing conditional upon being subsidized as reasons not to apply. In addition, funding agencies typically follow a picking-the winner strategy, i.e. firms that are highly innovative and conduct a lot of R&D might be more likely to get a subsidy. In other words, subsidies become an endogenous variable in any equation on innovation-related activities.

4. The empirical study will be limited to firms that engaged in innovation activities.

5. We show regression results of these ad-hoc approaches in the appendix in Table A3.

6. In a (knowledge) production function, Griliches suggested to look at the effect of different components of R&D by weighting one of the terms (say ) differently than the other, labeled as in this example. The full R&D term can then be decomposed as follows: , where δ corresponds to an output premium or discount of this second R&D term.

7. The IWT administers the R&D subsidy schemes in Flanders. The scope of its existing funding programs is quite broad as it supports a wide range of activities of small as well as large companies, universities, third-level education institutions and other Flemish innovative organizations, individually or collectively. More ample background information on the agency and its activities can be found on the website of the agency, www.iwt.be, as well as in Larosse (2004).

8. Note that we cannot implement panel – econometric approaches controlling for unobserved firm heterogeneity as many firms are only observed once in the final sample.

9. The yearly amount of subsidies is calculated based upon a monthly redistribution of the total subsidy grant. E.g. if a firm starts a subsidized project in April 2012 that ends in December 2013, 9/21 of the total amount will be allotted to 2012 and 12/21 to 2013.

10. The perpetual inventory method calculates the stock of a specific variable (VAR) in time t, let us name this STOCK as follows: STOCK = STOCK + VAR, where δ refers to the applied discount rate or rate of obsolescence.

11. Admittedly, the pharmaceutical industry may be an exception, as a single patent may correspond almost to a product and development phases of drugs after patent filings may be long. However, for most industries patents more than five years old may be almost obsolete.

12. An overview of the industry structure is given in Table A1 in the appendix.

13. The IWTSUBINT variable is thus 0 for all unsubsidized firms by construction.

14. We also computed the Hansen J-test and did not reject the hypothesis that the instruments are exogenous, i.e. the instruments are not only relevant in the first stage but also valid in terms of statistical requirements.

15. Note that R&D spending in our case is all money invested, i.e. including the subsidy. If we would have measured R&D net of the subsidy the relevant test in the example outlined above would be whether the estimated coefficient is larger than the value 1.

16. Strictly speaking we cannot reject the hypothesis of some crowding – out as a t-test reveals that the estimated coefficient of 3.046 is not significantly larger than 2.304. However, this is not the main focus of the paper and results suggest that the subsidy certainly increases private R&D investment by an economically significant factor.