IMPACT ASSESMENT OF R&D SUBSIDIES IN SPAIN: SOME PRELIMINARY RESULTS

Transcription

1 European Commission Joint Research Centre - Institute for Prospective Technological Studies Knowledge for Growth Economics of Industrial Research & Innovation (IRI) IMPACT ASSESMENT OF R&D SUBSIDIES IN SPAIN: SOME PRELIMINARY RESULTS Elena Huergo GRIPICO-Universidad Complutense de Madrid Contributed paper to be presented at the 4 th European Conference on Corporate R&D and Innovation CONCORDi-2013, September , Seville (Spain) Conference title Financing R&D and innovation for corporate growth in the EU: Strategies, drivers and barriers Topic Public policies, policy means and financing facilitators File name: Huergo-CONCORDi Author: Elena Huergo Authors' contact: ehuergo@ccee.ucm.es Status: Draft Last updated: Organisation: GRIPICO- Universidad Complutense de Madrid Page 1 of 24

2 Abstract The objective of this paper is twofold: to analyze the determinants of Spanish firms applications to the public agency for low-interest credits to finance their R&D projects and to study the selection of the subsidy rates granted by this agency. The empirical specification builds on the model developed by Takalo, Tanayama and Toivanen (2013). The results obtained for a sample of Spanish firms which includes more than 1,900 applications during the period suggest that young firms, exporters, large firms, companies that belong to a high or medium-tech industry, and especially firms with previous experience in similar programs have a higher probability of applying for a credit. As for the agency s decision, the age of the company, its exporting character or being a SME are not relevant, everything else equal. The most positive effect on the subsidy rate is related to the technological challenge of the proposal, providing evidence that the agency tries to stimulate the best projects. Independent firms and firms with previous experience in the system are also granted higher subsidy rates. Key words: D04, G38, H25, H81, L59, O38. JEL classification: Low-interest credits, public aid for R&D, application, subsidy rate. CONCORDi-2013 Page 2 of 24

3 TABLE OF CONTENTS 1 - Introduction CDTI s role in Spanish innovation policy The data The analysis The application stage The agency s decision Conclusions ANNEXES References Annex References Annex 1 LIST OF ANNEXES CONCORDi-2013 Page 3 of 24

4 1 - Introduction 1 In recent years, the growth in public R&D budgets has underlined evaluation as an essential part of the design of technological policy. As a consequence, the number of empirical papers that try to evaluate the net impact of R&D subsidies on firm behavior and output has grown dramatically, with a wide variety of countries analyzed and methodologies employed (Wallsten, 2000; Busom, 2000; Duguet, 2004; González, Jaumandreu and Pazó, 2005; Czarnitzki and Licht, 2005; OECD, 2006; Clausen, 2008; Takalo, Tanayama and Toivanen, 2013). In this assessment it is crucial to take into account that the receipt of an award probably depends on the same firm characteristics that determine a firm s technological strategy. That s why besides those studies that specifically examine the determinants of participation in public R&D programmes, most empirical papers regarding impact analysis also study these determinants as a previous step to control for potential endogeneity (Busom, 2000; González, Jaumandreu and Pazó, 2005; Czarnitzki and Licht, 2005). One limitation for doing this is that the information in public databases does not allow to distinguish between rejected applicants and non-applicants. This is an important restriction because it prevents us from separating the variables that influence a firm s self-selection from those related to public agencies selection rules, which can be designed to stimulate specific target groups such as national champions (picking-thewinners strategy), SMEs (with higher financial constraints), or firms in sectors with larger spillovers, among others. One exception to this restriction is the paper by Takalo, Tanayama and Toivanen (2013). 2 To evaluate the impact of R&D subsidies in Finland, these authors develop a model (hereafter the TTT model) that explains the functioning of a R&D subsidy program in terms of two different actions: the actions of firms applying for subsidies and those of the agency deciding on the subsidies. Using the rich R&D project-level data from Tekes (the National Technology Agency of Finland), they separately estimate the determinants of an application cost function and of the Tekes decision rule in terms of the subsidy rate. Regarding the application cost, companies with larger boards and export activity have lower application costs. In addition, the size of the firm and the number of past applications reveal a non-linear, decreasing effect. The results concerning the Tekes decision rule suggest that the more challenging is a project technically, the higher is its subsidy rate. Another significant determinant is firm size, with a positive effect, although SMEs are granted a higher subsidy, everything else equal. However, the number of previous applications has no effect. In this research, and other related papers (see Beyer, Czarnitzki and Toivanen, 2013a y b; Czarnitzki and Toivanen, 2013a y b; Pacher and Mohnen, 2013a y b; Takalo and Toivanen, 2013a y b), we aim to contribute to this literature by studying how firms decide whether or not to apply for an R&D subsidy in different European countries and how public agencies of different European countries decide on R&D subsidy rates they 1 This paper is the joint result of two discussion papers: Huergo (2013a,b). 2 Other exceptions are Barajas and Huergo (2010) and Huergo and Trenado (2010) for Spanish firms applying for public aid to finance cooperative or individual R&D projects. CONCORDi-2013 Page 4 of 24

5 grant to their national firms. In particular, this paper builds on the TTT model with the objective of providing new insights about the determinants of Spanish firms application to public system of low-interest credits for R&D projects carried out by the Center for the Development of Industrial Technology (CDTI) during the period In addition, the paper studies the determinants of the R&D subsidy rates granted by this Spanish public organization that, under the Ministry of Economy and Competitiveness, grants financial help of its own to companies and facilitates access to that of third-party funds for the execution of both national and international research and development projects. Nowadays, this agency is the main non-business source of funding for firms R&D projects in Spain. The study of the determinants of the participation in the CDTI s system is a first step in quantifying its effect on Spanish business R&D expenditures. In addition, the methodology employed for the analysis will allow for international comparisons. The empirical analysis rests on the complementary information of two databases: the CDTI database, which includes information about more than 1,900 proposals (accepted or rejected) that were received by the public agency between 2002 and 2005; and the SABI database, which contains the officially registered financial accounts of up to one million Spanish firms. The rest of the paper is structured as follows: Section 2 presents CDTI s role and objectives. Section 3 is devoted to a presentation of the data. Section 4 shows the results of the empirical analysis, and Section 5 offers concluding remarks. 2 - CDTI s role in Spanish innovation policy As many studies point out, public instruments to support business R&D are quite diverse and their design tends to pretend to a balance between the objectives of administrations and the costs of political action. A first form of intervention is the public procurement of innovative goods and services, which by its tractor effect results in improving the competitiveness and the internationalization of the providers. 3 A second group of instruments comprises all actions affecting the environment in which innovative companies operate, but that do not entail any financial aid to them. This group includes the design of patent systems to protect intellectual property rights, the regulatory policies to monitor the competitive behavior in markets, the administrative procedures for the creation of new companies (especially, those technology-based) or the creation of diffusion channels for new technologies. However, the type of public intervention more focused on reducing the gap between the social and the private return of R&D investments is the one that involves financial support, which can be direct, mainly through subsidies or low-interest loans, or indirect, mainly through R&D tax credits. 3 In Spain, almost half of the economy is managed from the public side: public procurement represents 13% of the GDP and public spending around 40% (see Arana, 2011). CONCORDi-2013 Page 5 of 24

6 According to OECD (2010), countries differ in the use and the intensity of this financial aid. While United States place the emphasis on direct support, Canada and Japan focus mainly on indirect aid. The combination of direct and indirect support in each country depends on the final objective of Governments. In general, tax credits are mostly used to encourage short-term applied research, while direct subsidies are more targeted to foster long-term research (OECD, 2010). In the specific case of Spain, the R&D fiscal treatment has a mixed design that combines the deduction by volume and by increase, and that is applicable to most of the costs generated by R&D investments, including wages, raw materials, R&D hired externally and assets amortization. However, it is the tax authority who recognizes such expenses as eligible costs. The proportion of deductible R&D expenditure has increased over time, going from 10% in the early 1980s to 50% over the past years. Nowadays, the Spanish tax system is considered as one of the most generous among OECD countries in terms of the tax subsidy rate (OECD, 2012). Furthermore, in Spain 20% of business R&D expenditures (BERD) are funded with public support, mainly through direct aid (17-18% of the BERD in , Heijs, 2011). In particular, the data from the Panel of Technological Innovation (PITEC) shows that around 30% of Spanish innovative companies had access to public subsidies to finance their R&D expenditures in recent years. As can be seen in Figure 1, these subsidies came principally from the Autonomous and Local Administrations and the Central Government, although a non-negligible percentage of innovative firms had also access to funds of the European Union (EU). 4 Figure 1. Access to public subsidies to finance R&D projects in Spain between 2007 and 2010 (% of innovative firms) (*) Access to Autonomous & Local Administrations funding Access to Central Government funding Access to EU funding Source: PITEC database and own elaboration. (*) The categories are not mutually exclusive. The firms can have access to more than one type of aid. 4 The Spanish National Institute of Statistics constructs this database on the basis of the annual Spanish responses to the Community Innovation Survey. CONCORDi-2013 Page 6 of 24

7 Notice that these percentages do not include reimbursable loans obtained from public administrations or other sources to finance R&D projects. Following the design of the Spanish Community Innovation Survey, in the PITEC database reimbursable credits are considered as part of firms own funds. This is important because most of the direct R&D support of the Spanish Central Government is channeled through the Center for Industrial technological development (see Figure 2), and the main CDTI s instrument to support R&D projects consists of credits at a preferential interest rate (below the market rate). Figure 2: Non-business sources of funding of business R&D spending in Spain. Year 2011 (thousands of euros) Source: CDTI (2011) CDTI s main objective is to contribute to improving the technological level of the Spanish companies by means of implementing the following activities: Financial and economic-technical assessment of R&D projects implemented by companies. Managing and fostering Spanish participation in international technological cooperation programmes. Fostering international business technology transfer and support services for technological innovation. Supporting the setting up and consolidating technological companies. In particular, CDTI provides different types of support depending on the types of project, that have evolved along time (see Table 1) 5 : individual R&D projects, cooperative R&D projects involving several companies, projects for the creation and consolidation of 5 The list of instruments included in Table 1 is not exhaustive. Other CDTI instruments are, for example, CENIT projects (funding from 2006 to 2010 by means of subsidies of up to 50%, of large scale and major scientific-technical consortium projects aimed at planned research in future technology areas and with potential international projection) or INNODEMANDA action (financing instrument for supporting technological offers to the innovative public purchase process) CONCORDi-2013 Page 7 of 24

8 technology based companies,... And, in most cases CDTI offers a permanent call for multi-year aid, so applicant companies can develop their projects without the temporal restriction of a deadline for the opening or the closing of a call. Table 1: CDTI funding of Business R&D projects Project typology Up until July 2008: Maximum funding coverage Interest rate Amortization Nonreimbursable part Concerted Industrial Research Projects 60% 0% 10 years Up to 20% Technological Innovation Projects 60% 0% 10 years - Technological Development Projects 60% 0% 10 years - From July 2008: R&D Projects 75% 0% 10 years 15% Integrated and Technology Cooperation Projects between SMEs Banking line for Financing Technology Innovation Year 2011: R&D Projects 85% (75% + 10 pp of credit) Integrated and Technology Cooperation Projects between SMEs (Funded by the Technology Fund) Banking line for Financing Technology Innovation INNPRONTA (Funding for big integrated projects of industrial research, strategic dimension and high budget) FEDER INNTERCONECTA (Grants for integrated projects of experimental development. Co-funded by Technology Fund) Source: CDTI (several years) 75% 0% 10 years Up to 33% 75% Euribor -0,5 Up to 7 years, with a 2-year grace period 0% 10 years 15%-33% (SMEs) 75% 0% 10 years Up to 33% 75% Euribor -0,5 Up to 7 years, with a 2-year grace period 75% 0% 10 years Up to 40% (SMEs), up to 35% (Large) According to Community framework Subsidy - - In 2011, CDTI spent 1, million euros in financing 1,744 business R&D&I projects, of which: million euros were earmarked for 983 individual R&D projects; million euros for 191 cooperation projects; million euros for 7 INNPRONTA CONCORDi-2013 Page 8 of 24

9 projects; 6 5 million euros for 50 technology promotion projects; and million euros for 84 Neotec Initiative related projects for the creation and consolidation of technologybased companies (CDTI, 2011). 3 - The data The data used in the project comes from two different sources: the CDTI database and the SABI database. The CDTI database contains information on around proposals to the three types of project managed by this agency between 2002 and 2005: Technological Development Projects (TDP), Technological Innovation Projects (TIP) and Concerted Industrial Research Projects (CIRP). Through these programmes, the CDTI provided financing with a zero interest rate and a period of repayment of up to 10 years. The quantity of the credit could reach 60% of the total budget. The allocation of a proposal to one of the three programmes is determined by the character of the R&D that is embodied in each project. Both the TDP and the TIP have an applied character and are developed by firms, with or without the collaboration of technological centers. The difference between them is that the former implies the creation or the significant improvement of a process, good, or service, and the latter consists of the incorporation and the active adaptation of rising technologies within the firm or new markets. In the case of CIRP, the main objective is to finance initiatives of pre-competitive research: those projects in which the results are not directly marketable or that entail high technical risk. These proposals are presented by industrial firms in collaboration with universities, public research organizations, and/or Spanish centers of innovation and technology. Between the years 2002 and 2005, 2,648 Spanish firms presented 4,132 proposals to be supported by the CDTI through a TDP, TIP, or CIRP. Of these, 2,416 applications (1,719 firms) were ultimately approved. Companies with CIRP account for more than 70% of applications. Although we have pointed out some differences between TDP, TIP and CIRP, the common features are more remarkable from the point of view of the firms participation, so the proposals corresponding to these programs are jointly analyzed in the following sections. The second database used in this study is the SABI database, which contains the company accounts of more than 1,000,000 Spanish firms from 1995 to From these, a control sample has been selected taking into account the availability of data about the relevant variables for each firm. The sample includes all companies with more than 200 employees. In addition, firms employing between 10 and 200 employees are selected by a random sampling scheme for each NACE class (two-digit) level, and represent around 4% with respect to the Spanish Central Companies Directory (CCD), which compiles all Spanish companies and their local units located in the country in a single information system. This makes the control sample representative of the Spanish economy. 6 INNPRONTA is a programme for the financing of large integrated industrial research projects, of a largescale and strategic nature, which serve to develop new technologies in forward-looking technological areas with economic and commercial prospects at the international level. These projects are financed through a mixed mechanism of credit and subsidy. CONCORDi-2013 Page 9 of 24

10 Although the CDTI database contains data since 1995, the sample used in the empirical analysis refers only to the period 2002 to 2005, given that the information about the technical evaluation of the proposals is only available from 2002 on. However, the information from 1992 is employed to construct the variable that reflect the experience in proposals and funded projects. The final sample consists of 69,672 firm-year observations, 1,930 (1,489) of which correspond to applications (awarded projects) to the CDTI for an interest-free credit between 2002 and As can be seen in Figure 3, during this period the CDTI s budget for funding business R&D projects increased substantially. In parallel, the percentage of applications also increased, probably because firms internalized the public information about the raise in CDTI s budget. Figure 3: CDTI funding of R&D projects between 2002 and % 500 MEur 3.5% 3.0% 2.5% 2.0% 1.5% 2.2% 2.7% 3.2% 3.0% 400 MEur 300 MEur 200 MEur 1.0% Applications (% in sample) Direct funding of projects (Million ) 100 MEur Source: CDTI database and own elaboration. Table 2 displays summary statistics of the explanatory variables available in the database the will be used in the application equation. The average (log) age of firms in the whole sample is 21.5 (3.07) years; the average (log) number of employees is 25 (3.21), and the average sales per employee is 165,000. Only 0.5% of firms are independent, 7 19% have applied previously for a CDTI s loan, and 29% are exporters. Additionally, given that Spanish firms are smaller on average than the European firms (European Commission, 2003), it is not surprising to see a very high percentage of SMEs in the sample. 8 Descriptive statistics for industry and region dummies are reported in Tables A.1 and A.2 of the Appendix. 7 A firm is considered as independent in the database if it has not shareholder recorded with more than 25% direct or total ownership. 8 According to the official EU definition, an SME should have less than 250 employees and either sales less than 40 million euros or a balance sheet less than 27 million euros. CONCORDi-2013 Page 10 of 24

11 Table 2: Descriptive statistics Total Nonapplicants Applicants t-test a [p-value] Log of age (years) [0.000] (0.858) (0.859) (0.786) Log of employment (nº employees) [0.000] (1.947) (1.948) (1.657) Sales (million ) / employee [0.000] (1.515) (1.535) (0.287) Independent (yes/no) [0.000] (0.070) (0.067) (0.153) Previous applicant (yes/no) [0.000] (0.394) (0.384) (0.490) Exporter (yes/no) [0.000] (0.455) (0.450) (0.485) SME (yes/no) [0.000] (0.440) (0.438) (0.489) Number of observations 69,672 67,742 1,930 Source: CDTI and SABI databases, and own elaboration. Numbers reported are sample means (standard deviations). a : t-test=two-sample difference of means test. The difference of means tests confirm that the average of all variables in the table are different between applicants and non-applicants. Specifically, the presence of firms with previous experience in CDTI s system, exporters and independent firms is higher among applicants, which also appear to be younger than non-applicants. 9 As for to the characteristics of R&D projects, we can take advantage of the data from the technical evaluation that is performed by CDTI, which allows for a detailed description of the proposals. For the selection of projects, CDTI specifies the following general criteria: a) Degree of fulfilment of the information required; b) Scientific-technical quality of the proposal and degree of innovation; c) The firm s technical and financial capability to undertake the project; d) The firm s ability to exploit the results; e) Potential market of the developments to be achieved; and f) Fulfilment of encouraging effect. 10 Specifically, the evaluation consists of a rating analysis of more than 50 items that can be classified in a low, medium, or high category. To summarize this information, following the specification used in the TTT model, the evaluation of these items has been grouped into 2 global indicators that are related to the CDTI screening: the commercial/technological risk and the technological challenge of the proposal. Both grading variables consist of grades on a Likert scale of 3 (0, 1, 2 = low, medium, high). 9 The firm s age is computed as the difference between the current year and the constituent year reported by the firm. 10 In order to ensure the encouraging effect, CDTI only funds projects that have not been started by the date of presentation of the proposal. CONCORDi-2013 Page 11 of 24

12 As can be seen in Table 3, both dimensions achieve higher levels for awarded projects, although the difference between successful and rejected proposals is significantly higher in terms of the technical challenge than in terms of their risk. Table 3: Characteristics of applicants Successful Applicants applicants Rejected applicants t-test a [p-value] Risk (low, medium, high = 0, 1, 2) [0.027] (0.623) (0.605) (0.680) Technical challenge (low, medium, high=0, 1, 2) [0.000] (0.635) (0.536) (0.708) Log of age (years) [0.445] (0.786) (0.788) (0.780) Log of employment (nº employees) [0.074] (1.655) (1.648) (1.673) Sales (thousand ) / employee [0.000] (0.287) (0.253) (0.372) Independent (yes/no) [0.016] (0.153) (0.164) (0.107) Previous applicant (yes/no) [0.002] (0.490) (0.485) (0.499) Exporter (yes/no) [0.929] (0.485) (0.485) (0.485) SME (yes/no) [0.004] (0.489) (0.492) (0.473) Number of observations 1,925 1, Source: CDTI and SABI databases, and own elaboration. Numbers reported are means (standard deviations). Table 3 also displays summary statistics of the rest of explanatory variables available in the database the will be used in the subsidy rate equation. The average (log) age of applicants is 18.9 (2.94) years; the average (log) number of employees is 64 (4.16), and the average sales per employee is 66, % of firms are independent, % have applied previously for a CDTI s loan, and 62.3% are exporters. Additionally, most firms in the sample are SMEs, which is not surprising given that Spanish firms are smaller on average than the European firms (European Commission, 2003). 12 Descriptive statistics for industry and region dummies are reported in Tables A.3 and A.4 of the Appendix. 11 A firm is considered as independent in the database if it has not shareholder recorded with more than 25% direct or total ownership. 12 According to the official EU definition, an SME should have less than 250 employees and either sales less than 40 million euros or a balance sheet less than 27 million euros. CONCORDi-2013 Page 12 of 24

13 The difference of means tests also show that the presence of independent firms and previous applicants to CDTI s system is higher among award recipients, while SMEs are more frequent among rejected proponents. 13 However, the differences in the age and the exporter character between successful and rejected applicants and non-applicants seem to be unimportant. 4 - The analysis 4.1 The application stage For the empirical analysis, the application equation is estimated as a Probit model; that is, it takes the form: * 1 if yi f ( xi ui ) 0 y i, 0 otherwise * where y i is a latent dependent variable, x i are the determinants of the firm s decision to apply, is the vector of coefficients and u i ~N(0, 2 * ). Firm i applies for the credit if y i is positive. As determinants of the firm s decision to apply, besides the variables in Table 2, the specification includes a set of dummy variables to account for differences in regional support and for firms operating in high and mid-high technology sectors. 14 In addition, by including a set of dummy variables that are related to the year of application, the model considers the effect of changes in the business cycle and the fact that firms take into account the yearly public information about budget of the agency and the pattern of recent awards. Some quantitative variables are included in squared terms to allow for non-linear relationships. The results of the estimation are summarized in Table 4. Most variables are statistically significant, although marginal effects are small, probably due to the absence of variables that are more related to the technological profile of the companies. Past experience of firms in CDTI s proposals has a positive effect on the probability of applying to the program again, as well as the fact of being an independent or an exporting. Firms in international markets could have lower costs in preparing the applications, perhaps because they are relatively more experienced in dealing with government bureaucracy than are non-exporting firms (Takalo, Tanayama and Toivanen 2008). The log of age has a negative impact that indicates that younger companies are more willing to apply. This is consistent with previous empirical evidence suggesting that market failures especially affect cash-constrained small and start-up firms in R&Dintensive industries (Hall, 2002). However, the coefficients of the log of employment and its square indicate a positive but non-linear impact of size on the probability of applying. 13 The firm s age is computed as the difference between the current year and the constituent year reported by the firm. 14 See the correspondence with NACE-Rev.1 in Table A.5 of the Appendix. CONCORDi-2013 Page 13 of 24

14 Table 4: Application equation Coefficients Std. Err. dy/dx Std. Err. Log of age *** (0.014) *** (0.0005) Log of employment *** (0.024) *** (0.0008) Log of employment squared *** (0.003) *** (0.0001) Sales / employee *** (0.061) *** (0.0021) Sales / employee squared *** (0.000) *** ( ) SME *** (0.029) *** (0.0012) Independent *** (0.092) *** (0.0067) Previous applicant *** (0.025) *** (0.0025) Exporter *** (0.026) *** (0.0012) Year *** (0.032) *** (0.0013) Year *** (0.032) *** (0.0013) Year *** (0.032) *** (0.0013) Dummies for high and mid-high technology sectors Chemicals and chemical products *** (0.043) *** (0.0031) Machinery and equipment n.e.c *** (0.040) *** (0.0032) Office machinery and computers *** (0.179) ** (0.0139) Electrical machinery and apparatus n.e.c *** (0.064) *** (0.0041) Radio, television and communication equipment *** (0.076) *** (0.0061) Medical, precision and optical instruments *** (0.079) *** (0.0065) Motor vehicles, trailers and semi-trailers *** (0.062) *** (0.0048) Other transport equipment ** (0.106) ** (0.0064) Post and telecommunications ** (0.109) ** (0.0060) Computer and related activities *** (0.051) *** (0.0030) Research and development *** (0.101) *** (0.0063) Geographical dummies Andalucía ** (0.054) ** (0.0015) Aragón (0.064) (0.0021) Cataluña (0.035) (0.0013) Castilla La Mancha *** (0.100) *** (0.0015) Castilla y León (0.068) (0.0020) Galicia *** (0.096) *** (0.0010) Madrid *** (0.042) *** (0.0011) Navarra *** (0.051) *** (0.0058) País Vasco ** (0.046) ** (0.0019) Valencia *** (0.047) *** (0.0013) Pseudo R No. Observations 69,672 Notes: Dummy variable excluded for the year Marginal effects (dy/dx) are computed at sample means. For dummy variables, the marginal effect corresponds to the change from 0 to 1. Coefficients or marginal effects significant at: 1%***, 5%**, 10%*. At the same time being a SME decreases this probability. Although smaller firms could have more incentives to apply for public aid than larger firms due to credit constrains, the results suggest that in this case application costs are relatively higher for small firms. However, we have to be cautious with this interpretation, as our control sample is CONCORDi-2013 Page 14 of 24

15 biased towards large firms, which are chosen on the basis of a census, while firms employing between 10 and 200 employees are selected by a random sampling scheme. Sales per employee have a U-shaped effect on the application probability. Following TTT (2013), this could indicate that firms with high value current production have also higher opportunity costs of applying. However, in the case of Spanish firms, this opportunity costs seem to exhaust at certain point. The marginal effects of the time dummies reflect that probability of applying increases with respect to the base year (2002) in 2003, 2004 and 2005, confirming what we have observed through the descriptive analysis. These figures are consistent with the generous budget increases that the CDTI benefited from during this period (especially since 2004). As expected, firms that carry out high or mid-high technological activities are also more prone to become applicants. With respect to the geographical dummies, firms located in Navarra and País Vasco show higher probabilities of submitting an application, while firms in Andalucía, Castilla La Mancha, Galicia, Madrid and Valencia are less likely to apply. This heterogeneity can be revealing the different role of local and regional agencies in the Spanish technological policy. 4.2 The agency s decision For the empirical analysis, the subsidy rate equation is estimated as a two-limit Tobit model; that is, it takes the form: * * si di if si xi ui (0, s ) * si s di if si xi ui s, * 0 if si xi ui 0 where * s i is a latent dependent variable, x i are the determinants of the agency s award decision, is the vector of coefficients and u i ~N(0, 2 ). d i is an indicator function that takes the value one if a firm applies for a loan and is zero otherwise. The dependent variable in this equation is the subsidy rate granted by CDTI, which is defined as the ratio of the loan to the total R&D cost of the project. It takes the value zero if the optimal (latent) subsidy rate is non-positive, takes the value of the maximum allowed subsidy rate ( s 0.6 ) if the optimal (latent) subsidy rate is s or larger, and it is equal to the optimal subsidy rate between these two (censoring) values. Table 5 reports the evidence obtained from the estimation of this model. As for the grading variables, the results are consistent with the ones of TTT (2013), although with some interesting differences. The risk negatively affects the subsidy rate, while the technical challenge of the proposal has the opposite sign. Notice that, while the average of both variables in the descriptive analysis is higher among awarded applicants, it seems that once we control for the technical challenge of the proposal, the agency prefers those projects with lower risk. In fact, in terms of the magnitude of marginal effects, the grading regarding the technical challenge is much more important than the grading relative to the risk of the proposal. CONCORDi-2013 Page 15 of 24

16 Table 5: Subsidy rate equation Coefficients Std. Err. Risk *** (0.018) Technical challenge *** (0.019) Log of age (0.015) Log of employment *** (0.009) Sales / employee *** (0.036) SME (0.025) Independent *** (0.072) Previous applicant *** (0.023) Exporter (0.025) Year *** (0.030) Year *** (0.032) Year *** (0.033) Dummies for high and mid-high technology sectors Chemicals and chemical products *** (0.038) Machinery and equipment n.e.c (0.034) Office machinery and computers (0.152) Electrical machinery and apparatus n.e.c *** (0.055) Radio, television and communication equipment (0.065) Medical, precision and optical instruments (0.065) Motor vehicles, trailers and semi-trailers (0.052) Other transport equipment (0.094) Post and telecommunications (0.103) Computer and related activities (0.051) Research and development (0.091) Geographical dummies Andalucía ** (0.056) Aragón ** (0.067) Cataluña (0.036) Castilla La Mancha (0.104) Castilla y León (0.065) Galicia (0.115) Madrid (0.044) Navarra *** (0.042) País Vasco ** (0.045) Valencia (0.047) Pseudo R No. observations 1,925 Notes: Reported numbers are coefficients. Dummy variable excluded for the year Coefficients significant at: 1%***, 5%**, 10%*. The age of the company has no impact on the subsidy rate. As firms in declining or traditional industries tend to be older, it seems that the public agency is not especially worried about the technological upgrading of firms in these industries. Consistently with what we observed in the descriptive analysis, the exporting character of the company is not relevant for the agency s decision, everything else equal, while independent firms and firms with previous experience as applicants in CDTI s system are granted higher subsidy rates. CONCORDi-2013 Page 16 of 24

17 Regarding the firm size, the effect is negative in terms of the log of employment. However, being a SME does not affect the agency s decision. In this sense, notice that, unlike the Finnish Agency, CDTI does not allow a different level of maximum subsidy for SMEs. The marginal effects of the time dummies reflect that the subsidy rate increased over time, confirming also what we observed through the descriptive analysis. This evolution is consistent with the outstanding increase in CDTI s budget during this period, suggesting that the agency s budget constraint could have a binding effect at least in the initial years of the period. It is also noteworthy that, when introducing the two grading dimensions of the proposal into the specification, firms in high or mid-high technological sectors are not in general expected to receive a higher subsidy rate. The two exceptions correspond to proposals of firms that produce Chemical and chemical products and Electrical machinery and apparatus, that are granted subsidy rates respectively above and below the average. With respect to the geographical dummies, firms located in Aragón and País Vasco are granted a higher subsidy rate, while the effect is the contrary in firms located in Andalucía and Navarra. The location in other Autonomous Communities does not appear to have any effect. 5 - Conclusions The objective of this paper is to analyze the determinants of Spanish firms applications to the public agency for low-interest credits to finance their R&D projects and the selection of the subsidy rates granted this public agency: the Center for the Development of Industrial Technology, the main non-business source of funding for firms R&D projects in Spain. The empirical specification builds on the model developed by Takalo, Tanayama and Toivanen (2013) that explain the functioning of a R&D subsidy program in terms of two different decisions: the firm s decision to apply for a subsidy and the agency s selection. For the analysis, two complementary statistical sources are used: the CDTI database, which includes information about all the proposals (accepted or rejected) that were received by the public agency since its creation in 1977, and the SABI database, which contains the officially registered financial accounts of up to one million Spanish firms between 1995 and The total dataset contains information on about 70,000 observations and 1,900 proposals, 1,489 (77%) of which correspond to awarded projects between 2002 and The results suggest that young firms, exporters, large firms, companies that belong to a high or medium-tech industry, and especially firms with previous experience in similar programs have a higher probability of applying for a credit. As for the agency s decision, the age of the company, its exporting character or being a SME are not relevant, everything else equal. The most positive effect on the subsidy rate is related to the technological challenge of the proposal, providing evidence that the agency tries to stimulate the best projects. Independent firms and firms with previous experience in the system are also granted a higher subsidy rate. CONCORDi-2013 Page 17 of 24

18 ANNEXES References Arana, L. (2011). La estrategia estatal de innovación, e2i. El compromiso para acelerar el cambio de modelo económico. Papeles de Economía Española, 127, pp Barajas, A. and E. Huergo (2010). International R&D Cooperation within the EU Framework Programme: Empirical Evidence for Spanish Firms. Economics of Innovation and New Technology, 19(1-2), Beyer, M., Czarnitzki, D. and O. Toivanen (2013). Applying for R&D subsidies: Evidence from Germany, SIMPATIC Discussion Paper. Blanes, J. V. and I. Busom (2004). Who participates in R&D subsidy programs? The case of Spanish manufacturing firms. Research policy, 33(10), Busom, I. (2000). An empirical evaluation of the effects of R&D subsidies, Economics of Innovation and New Technology, 9(2), CDTI (several years). Annual Report. Center for the Development of Industrial Technology. Madrid. Clausen, T. H. (2007). Access (not) granted: What kinds of firms participate in technology programs? TIK Working paper on Innovation Studies No Oslo. Clausen, T. H. (2008). Do subsidies have positive impacts on R&D and innovations activities at the firm level? TIK Working papers on Innovation Studies No Oslo. Czarnitzki, D. and G. Licht (2005). Additionality of Public R&D grants in a transition economy: the case of Eastern Germany. The Economics of Transition, 14(1), Czarnitzki, D and O. Toivanen (2013a). Applying for R&D subsidies: Evidence from Flanders, SIMPATIC Discussion Paper. Czarnitzki, D and O. Toivanen (2013b). R&D subsidy rate decisions: Evidence from Flanders, SIMPATIC Discussion Paper. Duguet, E. (2004). Are R&D subsidies a substitute or a complement to privately funded R&D? Evidence from France using propensity score methods for non-experimental data, Revue d Economie Politique, 114(2), European Commission (2003). SMEs in Europe The European Observatory for SMEs. Luxembourg: Office for Official Publications of the European Communities. González, X., Jaumandreu, J. and C. Pazó (2005). Barriers to innovation and subsidy effectiveness. The Rand Journal of Economics, 36(4), Hall, B. (2002). The financing of research and development. Oxford Review of Economic Policy, 18, Heijs, J. (2011): Mini Country Report/SPAIN, Thematic Report 2011 under Specific Contract for the Integration of INNO Policy Trend Chart with ERAWATCH ( ), Huergo, E. (2013a). R&D subsidy rate decisions: Evidence from Spain, SIMPATIC Discussion Paper. Huergo, E. (2013b). Applying for R&D subsidies: Evidence from Spain, SIMPATIC Discussion Paper. CONCORDi-2013 Page 18 of 24

19 Huergo, E. and M. Trenado (2010). The Application for and the Awarding of Lowinterest Credits to Finance R&D Projects. Review of Industrial Organization, 37, OECD (2006). Government R&D funding and company behaviour. Measuring behavioural additionality. OECD report. OECD (2010). Measuring Innovation: A New Perspective, OECD Publishing. OECD (2011). OECD Science, Technology and Industry Scoreboard 2011, OECD Publishing. OECD (2012). OECD Science, Technology and Industry Outlook 2012, OECD Publishing. Pacher, S. and P. Mohnen (2013a). Applying for R&D subsidies: Evidence from The Netherlands, SIMPATIC Discussion Paper. Pacher, S. and P. Mohnen (2013b). R&D subsidy rate decisions: Evidence from The Netherlands, SIMPATIC Discussion Paper. Takalo, T., Tanayama, T. and O. Toivanen (2008). Evaluating innovation policy: a structural treatment effect model of R&D subsidies. Bank of Finland Research, Discussion Papers 7/2008. Takalo, T., Tanayama, T. and O. Toivanen (2013). Estimating the benefits of targeted R&D subsidies. The Review of Economics and Statistics, 95(1), Takalo, T. and O. Toivanen (2013a). Applying for R&D subsidies: Evidence from Finland, SIMPATIC Discussion Paper. Takalo, T. and O. Toivanen (2013b). R&D subsidy rate decisions: Evidence from Finland, SIMPATIC Discussion Paper. Wallsten, S. J. (2000). The Effects of Government-Industry R&D Programs on Private R&D: The Case of the Small Business Innovation Research Program. The Rand Journal of Economics, 31(1), CONCORDi-2013 Page 19 of 24

20 Annex 1 Table A.1: Descriptive statistics for industry dummies. Whole sample Total Nonapplicants Applicants t-test a [p-value] Chemicals and chemical products [0.000] (0.001) (0.001) (0.007) Machinery and equipment n.e.c (0.001) (0.001) (0.007) [0.000] Office machinery and computers [0.053] (0.0002) (0.0002) (0.002) Electrical machinery and apparatus n.e.c [0.000] (0.001) (0.001) (0.196) Radio, television and comm. equipment [0.000] (0.0003) (0.0003) (0.004) Medical, precision and optical instr [0.000] (0.0004) (0.0003) (0.004) Motor vehicles, trailers and semi-trailers [0.000] (0.0005) (0.0004) (0.005) Other transport equipment [0.016] (0.0003) (0.0003) (0.003) Post and telecommunications [0.632] (0.0004) (0.0004) (0.002) Computer and related activities [0.000] (0.001) (0.001) (0.005) Research and development [0.000] (0.0003) (0.0003) (0.003) Number of observations 69,672 67,742 1,930 Source: CDTI and SABI databases, and own elaboration. Notes: Industry dummies only for high and mid-high technology industries. Numbers reported are sample means (standard deviations). a : t-test=two-sample difference of means test. CONCORDi-2013 Page 20 of 24

21 Table A.2: Descriptive statistics for regional dummies. Whole sample Total Nonapplicants Applicants t-test a [p-value] Andalucía [0.000] (0.001) (0.001) (0.005) Aragón [0.227] (0.001) (0.001) (0.004) Cataluña [0.000] (0.002) (0.002) (0.011) Castilla La Mancha [0.000] (0.002) (0.001) (0.002) Castilla y León [0.116] (0.001) (0.001) (0.004) Galicia [0.000] (0.001) (0.001) (0.002) Madrid [0.000] (0.001) (0.001) (0.007) Navarra [0.000] (0.001) (0.001) (0.007) País Vasco [0.000] (0.001) (0.001) (0.007) Valencia [0.000] (0.001) (0.001) (0.006) Number of observations 69,672 67,742 1,930 Source: CDTI and SABI databases, and own elaboration. Notes: Regional dummies only for Autonomous Communities with highest internal R&D expenditures in the business sector. Numbers reported are sample means (standard deviations). a : t-test=two-sample difference of means test. CONCORDi-2013 Page 21 of 24

22 Table A.3: Descriptive statistics for industry dummies. Sample of applicants Applicants Successful applicants Rejected applicants t-test a [p-value] Chemicals and chemical products [0.000] (0.007) (0.008) (0.011) Machinery and equipment n.e.c [0.090] (0.008) (0.008) (0.017) Office machinery and computers [0.227] (0.002) (0.002) (0.006) Electrical machinery and apparatus n.e.c [0.023] (0.005) (0.005) (0.012) Radio, television and comm. equipment [0.079] (0.004) (0.005) (0.006) Medical, precision and optical instr [0.003] (0.004) (0.005) (0.005) Motor vehicles, trailers and semi-trailers [0.009] (0.005) (0.005) (0.013) Other transport equipment [0.276] (0.003) (0.006) (0.006) Post and telecommunications [0.419] (0.002) (0.003) (0.004) Computer and related activities [0.021] (0.005) (0.006) (0.009) Research and development [0.658] (0.003) (0.003) (0.005) Number of observations 1,925 1, Source: CDTI and SABI databases, and own elaboration. Notes: Industry dummies only for high and mid-high technology industries. Numbers reported are sample means (standard deviations). a : t-test=two-sample difference of means test. CONCORDi-2013 Page 22 of 24

23 Table A.2: Descriptive statistics for regional dummies. Sample of applicants Applicants Successful applicants Rejected applicants t-test a [p-value] Andalucía [0.740] (0.005) (0.005) (0.010) Aragón [0.044] (0.004) (0.005) (0.007) Cataluña [0.789] (0.011) (0.012) (0.022) Castilla La Mancha [0.767] (0.002) (0.003) (0.005) Castilla y León [0.822] (0.004) (0.001) (0.002) Galicia [0.543] (0.002) (0.002) (0.005) Madrid [0.422] (0.007) (0.009) (0.005) Navarra [0.062] (0.007) (0.008) (0.017) País Vasco [0.506] (0.007) (0.008) (0.015) Valencia [0.649] (0.006) (0.007) (0.013) Number of observations 1,925 1, Source: CDTI and SABI databases, and own elaboration. Notes: Regional dummies only for Autonomous Communities with highest internal R&D expenditures in the business sector. Numbers reported are sample means (standard deviations). a : t-test=two-sample difference of means test. CONCORDi-2013 Page 23 of 24

24 Table A.5: Classification of high and mid-high technology sectors NACE-Rev.1 High and mid-high technology sectors 24 Chemicals and chemical products 29 Machinery and equipment n.e.c. 30 Office machinery and computers 31 Electrical machinery and apparatus n.e.c. 32 Radio, television and communication equipment 33 Medical, precision and optical instruments 34 Motor vehicles, trailers and semi-trailers 35 Other transport equipment 64 Post and telecommunications 72 Computer and related activities 73 Research and development CONCORDi-2013 Page 24 of 24