R&D investments, Financial Constraints and Export

Transcription

1 European Commission Joint Research Centre - Institute for Prospective Technological Studies Knowledge for Growth Economics of Industrial Research & (IRI) R&D investments, Financial Constraints and Export Carlo Altomonte 1, Simona Gamba 4, Maria Luisa Mancusi 2, Andrea Vezzulli 3 1) Bocconi University 2) Department of Economics, Catholic University (Milan) and CRIOS, Bocconi University 3) UECE-ISEG, Technical University of Lisbon and CRIOS, Bocconi University 4) Catholic University (Milan) Contributed paper to be presented at the 4 th European Conference on Corporate R&D and CONCORDi-2013, September , Seville (Spain) Conference title Financing R&D and innovation for corporate growth in the EU: Strategies, drivers and barriers < CONFERENCE TOPIC > R&D and innovation: Sources and constraints at company level File name: <FILENAME> Author: <AUTHOR NAME> contact t.b. in charge for correspondence Authors' contact: < > Status: <Draft> Last updated: <DATE> Organisation: <AFFILIATION> Page 1 of 24

2 Abstract This paper adds new empirical evidence on the circular relationship between financial constraints (FC), Research and Development investments (R&D) and exports. Our empirical analysis focuses on a large sample of manufacturing firms drawn across four European countries: France (FRA), Germany (GER), Italy (ITA) and Spain (SPA). Our results suggest that FC are a key determinant of both firm s innovation and export activities. Nevertheless, also positive feedbacks going from export and innovation to financial health are present. Policy implications are discussed. Key words: financial constraints, R&D investments, margins of export, innovation. JEL classification: F10, G20, G21, O30. CONCORDi 2013 Page 2 of 24

3 TABLE OF CONTENTS R&D investments, Financial Constraints and Export 1 - Introduction Literature review Data and variables Financial constraints variables and export variables Control variables Empirical findings Econometric analysis... Errore. Il segnalibro non è definito. 4.2 Robustness checks Conclusions and policy implications ANNEXES References Annex 1... Errore. Il segnalibro non è definito. Annex 2... Errore. Il segnalibro non è definito. LIST OF ANNEXES References CONCORDi 2013 Page 3 of 24

4 1 - Introduction European policy makers are belatedly starting to pay attention also to Europe s growth and jobs agenda. A key policy take from this debate is that Europe, and especially its Southern periphery, should foster competitiveness, i.e. the rate of growth of productivity, in order to leave behind the legacy of the financial crisis and avoid a lost decade scenario. [Ref to the ECB CompNet project] To this extent, among others, two sets of policies are often advocated: to encourage internationalization of local firms, with the implicit understanding that internationalization is associated with productivity growth and hence economic growth; and to increase the rate of innovation in the domestic economy, as the latter is a channel through which productivity growth happens [Ref_ Trade and Project run by the Trade Committee of the OECD]. While there is a growing and ample debate in the economic literature on the interconnection between these two policies, and the relevant policy implication [Ref Altomonte et al. 2013], the novelty of the present situation is that both internationalization and innovation take place in a context of tight financial constraints for firms. As such, the potential productivity-enhancing effects of these policies traditionally identified by the literature might fail to materialize, or might materialize under a different set of conditions. More specifically, the literature has identified the links between financial constraints and export as well as between credit rationing and innovation (see the literature review in the next section), but not much has been produced on the simultaneous interaction of all those variables. One of the few exceptions, with a focus mainly on developing countries, is Gorodnichenko and Schnitzer (2010), who claim that for financially healthier firms innovation and exporting are complementary activities. In particular, they find that those activities become substitutes in the presence of high financial constraints, whereas with mild financial restrains they can be both funded by internal or external resources. The main aim of this paper is thus to provide a contribution to a now growing, but still incomplete literature on the relationship among financial constraints (henceforth FC), innovation and export, accounting for their mutual interactions. To this extent, we will capitalize on a unique firm-level representative sample of manufacturing firms drawn across the four largest euro area countries- France (FRA), Germany (GER), Italy (ITA) and Spain (SPA)- as retrieved from the 1st survey on European Firms In a Global Economy (EFIGE). The survey, carried out in 2010, spans a large array of questions with both qualitative and quantitative data on firms characteristics and activities, for a total of around 150 different variables. The latter allow us to go beyond balance sheet information to address several crucial issues related to the link between internationalization, innovation and credit constraints. CONCORDi 2013 Page 4 of 24

5 2 - Literature review This paper builds upon three different streams of literature. The first one focuses on the relationship between FC and innovation. A number of papers have recently shown the negative effect of FC on R&D investment (Aghion et al., 2012; Mancusi and Vezzulli, 2012; Brown et al., 2013) and innovation (Savignac, 2008). One of the main challenges faced by this stream of literature is how to define, measure and test for the presence of FC. The most widely used strategy is to proxy FC using indirect instruments such as investment sensitivity to cash flow (Fazzari et al., 1988), leverage and liquidity ratios (Greenway et al., 2007), dividend payouts, size and age. This approach presents several drawbacks because most of these indicators are at least partially chosen endogenously by firms and because their effectiveness to detect unambiguously the presence of FC has been severely criticized (Kaplan and Zingales, 1997). Several studies try to overcome these limitations by using direct measures of FC coming from survey data (Tiwari et al., 2007; Hajivassiliou and Savignac, 2011; Mancusi and Vezzulli, 2012) or payment incidents (Aghion et al., 2012). In this paper we adopt both an indirect measure of FC, which consists in the index proposed by Whited and Wu (2006), and a direct one, coming from survey data, that aims to detect the presence of bank credit rationing, a problem that was first introduced by the pioneering work of Stiglitz and Weiss (1981). Although, from a theoretical perspective, a negative relationship between FC and R&D has been argued in the previous literature motivated by the presence of information asymmetries (Myers and Majluf, 1984), lack of collateral and high adjustment and sunk costs (Arrow, 1962), the empirical evidence is still ambiguous. Indeed, while some contributions have found that internal sources of funds are more important for R&D than for ordinary investment (Himmelberg and Petersen, 1994; Czarnitzki and Hottenrott, 2009), others have found that R&D investment is sensitive to financial constraints, but not significantly more than ordinary investment (e.g. Mulkay et al., 2001), or it may even not be sensitive to financial constraints (e.g. Bond et al., 2005). The second stream of literature focuses on the relationship between FC (and credit rationing) and firms international activities (most often export). Also in this case the relationship seems to be ambiguous and composed by at least two possible patterns. In the first one the causal relationship moving from FC to export consists in a self-selection mechanism generated by high sunk cost thresholds that prevents firms to participate to international markets (Bellone et al., 2010; Manova, 2011) and in the presence of high variable trading costs hampering the firm s export intensive margin (Manova, 2013). Focusing on credit rationing, Minetti and Zhu (2011) provide evidence that limited access to bank debt has a negative impact on a firm s export. The second theoretical pattern going from the export status to the less binding FC may consist of several determinants. First of all, exporters usually have an easier access to international financial markets, widening the credit supply they can draw from. Secondly, revenues from export are thought to be more stable due to international diversification in sales, thus improving the liquidity status of exporting firms (Greenaway et al., 2007). Finally, from Melitz (2003) we know that firms self-select into international competition, this implying that the exporter status is per se an instrument for quality signalling (based on the expectation of more stable future cash flows) that can lessen the severity of financial constraints through a reduction of information asymmetry (Campa and Shaver, 2002). The third and final stream of relevant literature focuses on the circular link between innovation and exporting. The main theoretical argument is that innovation fosters firm s CONCORDi 2013 Page 5 of 24

6 productivity and therefore promotes export (Melitz, 2003), while learning by exporting, in turn, feeds back into innovation and productivity. In most cases, empirical contributions focus on one of the two sides of the innovation-export relationship. On the one hand Vanbeveren and Vandenbussche (2010) and Cassiman et al. (2011) show that product innovation has a positive impact on the decision to enter a foreign market. On the other hand Damijan et al. (2010), Bustos (2011), Bratti and Felice (2012) study the positive effect of export (generated through learning by exporting ) on innovation. The paper by Melitz and Costantini (2007) represents the first attempt to endogenize innovation and export decision simultaneously. Their dynamic model explains the firm s joint decisions to innovate and export as an adjustment to a trade liberalization shock. Finally, Harris and Moffat (2011) study how R&D, innovation and the decision to sell abroad are interrelated. 3 - Data and variables Our main data source is the 1st survey on European Firms In a Global Economy (EFIGE) carried out in 2010 and mainly concerning year 2008 (with some questions asking information for 2009 and previous years). The survey provides detailed qualitative and quantitative information on several firms characteristics (such as ownership, internal structure, investment, innovation, internationalization, financial structure, market and pricing strategies) 1 and has been integrated with firm s balance sheet data (concerning the years ) using the Amadeus database developed and maintained by Bureau van Dijk. While the original EFIGE survey sample includes manufacturing firms from Austria, Hungary, United Kingdom and the four largest euro area countries (France, Germany, Italy and Spain) and is fully harmonized and comparable across countries, the availability of firm s balance sheet data from the Amadeus database varies significantly amongst these countries. In particular, only for the four largest euro area countries the percentage of surveyed firms with missing balance sheet information are less than 50%. Thus, we restrict our analysis to these countries (11799 manufacturing firms), and in particular to the 6702 observations characterized by the full availability of balance sheet information. 2 In the following analysis we will focus on this final sample. Table 1 reports further details on the number of surveyed firms in each country for different size classes (both for the original and the final sample), while Table 2 presents the number of firms by country and sector. Table 1: Number of firms by country and size class Class size FRA GER ITA SPA Total Employees (10-19) Employees (20-49) See for a detailed description of the EFIGE dataset and of the EFIGE sampling design. 2 Altomonte et al. (2012) discuss in detail the characteristics and the representativeness of the restricted sample matched with accounting data for Italy, France and Spain, finding no major differences with respect to the original survey sample. CONCORDi 2013 Page 6 of 24

7 Employees (50-249) Employees (over 250) Total Table 2: Distribution of firms by country and sector Sector FRA GER ITA SPA Total NACE Total The sampling design has been targeted to firms with at least 10 employees, with an oversampling of larger firms with more than 250 employees. In order to take into account this oversampling and to retrieve the sample s representativeness of the original firm s population, a weighting scheme (where weights are inversely proportional to the variance of an observation) has been set up according to firm s industry and size classes. The weight (weight) for the firms in industry k and size class s has been built as: weight ks =[(Pfirms ks /Pfirms)/(Sfirms ks /Sfirms)]*(Pfimrs/Sfirms) where Pfirms ks is the number of firms in industry k and size class s for the population in a given country (data have been retrieved from Eurostat Structural Business Statistics year 2007); Sfirms ks is the number of firms in industry k and size class s in the EFIGE sample; Pfirms and Sfirms are the total number of firms in the population and sample respectively. The sum of these weights over the firms is equal to the total number of firms in the reference population by country. Firms belonging to the same sector/size cell share the same weight. CONCORDi 2013 Page 7 of 24

8 All following descriptive statistics and regression results have to be considered as weighted, except where otherwise specified. Thus, they have to be interpreted as referring to the whole original population (hencefoth Population). 3.1 Financial constraints variables The EFIGE questionnaire asks the following question aiming at detecting the presence of potential FC in the form of credit rationed by the main bank of reference: "During the last year (2008), did the firm apply for more credit?". Firms answering "Yes, applied for it but was not successful" may be considered as credit rationed (RATIONED=1) and thus financially constrained, whereas firms answering one of the two other possible answers ("Yes, applied for it and was successful" and "No, did not apply for it") may be considered as not financially constrained (RATIONED=0). However this question was asked only to firms recurring to external finance (i.e. funds not generated internally) over the period and willing to increase their borrowing at the same interest rate of their current credit line. Thus, only a small fraction of firms (2189 out of in the original sample) answered to this question. Given the high percentage of missing values for this direct indicator of FC, we will mainly rely on an indirect indicator. In particular, we will use the index proposed by Whited and Wu (2006) 3, henceforth WW, measuring the firm's need of external finance. The WW index is obtained (in the author s original article) as a solution of a constrained maximization problem of a structural investment model and it is defined as follows: WW= CF DIVPOS TLTD LNTA ISG SG (1) The variables included in the formula (computed in our case by taking 2008 as reference year) are: CF: the ratio of cash flow to total assets; DIVPOS: a dummy variable that takes the value of one if the firm pays cash dividends; TLTD: the ratio of the long-term debt to total assets; LNTA: the natural log of total assets; ISG: the firm s three-digit industry sales growth (measured, in our case, at the country level over all firms included in the AMADEUS database); SG: firm sales growth. Since information concerning dividends is available in our sample only for a small fraction of listed firms, we will rely on a proxy for this variable. In particular, following Mancusi and Vezzulli (2010), we will attribute a value of 1 to the variable DIVPOS if the firm s net assets in 2008 were less than the sum of firm s net assets in 2007 plus the firm s profits (or losses) computed before tax, in formula: NET ASSETS t < NET ASSETS t-1 + PROFIT (or LOSS) BEFORE TAX t. 3 This index is a considerable alternative to the well-known Kaplan and Zingales (1997) index. CONCORDi 2013 Page 8 of 24

9 Descriptive statistics for variables included in the index (thus, referring to the year 2008) are provided in Table 3. 4 Table 3: Descriptive statistics for the variables composing the WW index (reference year 2008) Variable Observations Population Mean Std. Dev. Min Max CF DIVPOS TLTD LNTA ISG SG In order to improve the comparability across countries of the measure of financial constraints, we will compute a new index (called WWdiff) obtained by subtracting from each firm s value of WW index the country sample median. Descriptive statistics for both WW and WWdiff are reported in Table 4, while Figure 1 shows their distribution. Notice that firms having a negative value for WWdiff are less constrained than firms having a positive value. Table 4: FC indexes (reference year 2008) Variable Observations Population Mean Std. Dev. Min Max WW WWdiff Notice that, to clean outliers, values from the first and the last percentile have been replaced by the next value counting inwards from the extremes. Moreover, for Germany, given the high number of missing values for the variable CF, these have been substituted by predicted values. CONCORDi 2013 Page 9 of 24

10 Figure 1: FC indexes Kernel densities (reference year 2008) Not surprisingly, the average WWdiff value is smaller for firms declaring in the EFIGE questionnaire not to be financially constrained (RATIONED=0), while it is higher for firms self-assessing to be constrained (RATIONED=1) (see Table 5). The difference in the mean value of the index for constrained and unconstrained firms is statistically different at the 1% level. Table 5: WW index sorted by credit rationing status (reference year 2008) RATIONED =1 if the firm self-assessed to be rationed. RATIONED =0 otherwise. RATIONED Observations WWdiff Mean value for unconstrained firms is significantly different from the mean value for constrained firms at the 1% level of significance. Moreover, the value of the index decreases with the firm size (see Table 6), confirming theoretical literature prediction that bigger firms are less financially constrained. CONCORDi 2013 Page 10 of 24

11 Table 6: WWdiff index sorted by size (reference year 2008) Employees WWdiff Population Observations and more Total and export variables Information on R&D activities and export are directly provided by the EFIGE questionnaire. In particular, we will consider as R&D investing those firms both declaring to have undertaken any R&D activities (in-house, acquired from another firm in the group or acquired from external source) in the period and having in 2008 a positive number of employees involved in R&D activities. Concerning export activities, we created a dummy variable (export) which takes the value of one if the firm sold abroad directly from the home country some or all of its own products/services in 2008 and zero otherwise. From the EFIGE survey data we also created two measures of export and R&D intensities: expint and RDint. While the first one corresponds to the percentage of 2008 annual turnover represented by the export activities, the second one represents the percentage of the total turnover that the firm invested in R&D in the period Table 7 provides descriptive statistics for these variables. 44% of firms in the population invest in R&D, with an average investment, for the whole population, of around 3.5% of the turnover. About 55% of firms instead export and, on average, export represents 18% of annual turnover. Table 7: Descriptive statistics for RD, EXP, RDint, expint Variable Observations Population Mean Std. Dev. Min Max RD EXP RDint expint Notice that the majority of firms investing in R&D were also exporting (68,92%), as well as the majority of exporting firms (55.36%) were also investing in R&D (see Table 8). Table 8: Number of exporting and innovating firms variables EXP=0 EXP=1 Total RD=0 2,085 1,649 3,734 RD= ,045 2,968 Total 3,007 3,695 6,702 5 Notice that outliers (firms declaring to have invested more than 100% of the total turnover) have been dropped. CONCORDi 2013 Page 11 of 24

12 In Table 9, sorting EXP and R&D by size, we highlight how the percentage of exporters and R&D investing firms grows with size. Moreover, Table 10 shows that the percentage of 2008 annual turnover represented by the export activities grows with size, as well as the percentage of the total turnover that the firms invested in R&D in the period The increase, along with firm s size, of the share of exporting or R&D investing firms holds for all countries, as showed in Table 11. Also the importance of export activities grows with firm size in every country, while the percentage of the turnover invested in R&D increases with firm size only in Germany and Italy. Moreover, we can notice that, independently from the firm size, Italy displays the higher share of exporting firms and that the percentage of 2008 annual turnover represented by export activities is higher than in all other countries. Number of employees Table 9: Share of EXP and RD sorted by size Share of EXP=1 Std. Dev. Population Observations (10-19) (20-49) (50-249) (250 and over) Number of employees Share of RD=1 Std. Dev. Population Observations (10-19) (20-49) (50-249) (250 and over) Number of employees Table 10: Mean values of expint and RDint sorted by size expint (mean) Std. Dev. Population Observations (10-19) (20-49) (50-249) (250 and over) Number of employees RDint (mean) Std. Dev. Population Observations (10-19) (20-49) (50-249) (250 and over) CONCORDi 2013 Page 12 of 24

13 Table 11: Means of EXP, RD, expint and RDint sorted by country and size Employees (10-19) Employees (20-49) Share of Share of expint Rdint Share of Share of expint Rdint Country EXP=1 RD=1 (mean) (mean) EXP=1 RD=1 (mean) (mean) FRA GER ITA SPA Employees (50-249) Employees (250 and over) Share of Share of expint RDint Share of Share of expint RDint Country EXP=1 RD=1 (mean) (mean) EXP=1 RD=1 (mean) (mean) FRA GER ITA SPA Control variables The causal effect of firms' age, size and total factor productivity on financial constraints have been widely discussed in the literature (see for example Hadloc and Pierce, 2010) and their use as control variables in this context is quite common. Table 12 presents some descriptive statistics for size and age, where the first is computed as the total number of employees in 2008 and the latter is expressed in years. The average firm size differs across countries, with French and German firms that are definitely larger than Spanish and Italian firms. Italian and Spanish firms are not only the smallest, but also the youngest (respectively with an average age of about 29 and 26 years), while German firms are the oldest (with an average age of about 43 years). Countr y Table 12: Descriptive statistics for the control variables Observation Populatio Mean Std. Dev. Min Max Variable s n FRA Employees Age GER Employees Age ITA Employees Age SPA Employees Age To calculate the total factor productivity (TFP), we have run for each sector separately (defined at NACE 2 digit level) the Levinsohn and Petrin (2003) semi-parametric production function estimation algorithm, controlling for country and year fixed-effects. CONCORDi 2013 Page 13 of 24

14 To avoid reverse causality, in the following regressions we will use the total factor productivity of year Table 13 resumes the variables used in the analysis, providing a short description for each of them. Variable WW WWdiff rationed size age tfp_2007 RD EXP RDint expint Table 13: Definition of the variables Definition Whited and Wu (2006) index measuring the firm's need of external finance. To make the measure of financial constraints comparable across countries, we subtracted from the firm's WW index value the median WW value for the firm's country. Dummy variable =1 if in 2008 the firm applied for more credit and was not successful; =0 if it was successful or if did not apply for more credit. Number of employees of the firm. Age of the firm in years. Total factor productivity of year 2007 estimated through the Levinsohn and Petrin (2003) semi-parametric production function algorithm. Dummy variable =1 if: -in the period the firm undertook any R&D activity AND -in 2008 had a positive number of employees involved in R&D activities; =0 otherwise. Dummy variable =1 if the firm sold abroad directly from the home country some or all of its own products/services in 2008; =0 otherwise. Percentage of the total turnover that the firm invested in R&D in the period Percentage of 2008 annual turnover represented by the export activities. 4 - Empirical findings In this section we proceeded with the empirical analysis of the relationship between financial constraints, innovation and export. 4.1 Unconditional regressions Table 14, 15 and 16 show the unconditional regressions having as dependent variable respectively the dummy for being an exporter (EXP), the dummy for investing in R&D (RD) and the WWdiff index. In all regressions tfp_2007, age and size have been included in their logarithmic form. Moreover, country and industry fixed effects have been added. CONCORDi 2013 Page 14 of 24

15 Notice that, in Table 14 and 15 (the ones having a binary dependent variable), results obtained through a probit model have been included in the last column. Table 14: Unconditional regression having EXP as dependent variable OLS Probit (1) (2) (3) (4) (5) VARIABLES EXP EXP EXP EXP EXP WWdiff *** *** *** (0.112) (0.109) (0.125) RD 0.193*** 0.191*** 0.207*** (0.0133) (0.0133) (0.0142) ltfp_ *** 0.112*** *** *** 0.103*** (0.0165) (0.0163) (0.0175) (0.0173) (0.0199) lage *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) lsize *** *** *** *** *** ( ) ( ) ( ) ( ) (0.0100) Constant ** 0.137*** 0.144*** (0.0438) (0.0423) (0.0450) (0.0436) Observations 6,702 6,702 6,702 6,702 6,702 R-squared Country and industry dummies included. Robust standard errors in parenthesis. *** p<0.01, ** p<0.05, * p<0.1 Table 15: Unconditional regression having RD as dependent variable OLS Probit (1) (2) (3) (4) (5) VARIABLES RD RD RD RD RD WWdiff *** * (0.111) (0.108) (0.124) EXP 0.193*** 0.192*** 0.207*** (0.0133) (0.0134) (0.0142) ltfp_ *** ** (0.0164) (0.0163) (0.0173) (0.0171) (0.0196) lage ** ** ( ) ( ) ( ) ( ) ( ) lsize 0.111*** *** *** *** 0.103*** ( ) ( ) ( ) ( ) (0.0102) Constant * (0.0448) (0.0433) (0.0462) (0.0447) Observations 6,702 6,702 6,702 6,702 6,700 R-squared Country and industry dummies included. Robust standard errors in parenthesis. *** p<0.01, ** p<0.05, * p<0.1 Table 16: Unconditional regression having WWdiff as dependent variable OLS CONCORDi 2013 Page 15 of 24

16 (1) (2) (3) (4) VARIABLES WWdiff WWdiff WWdiff WWdiff RD *** * ( ) ( ) EXP *** *** ( ) ( ) ltfp_ *** *** *** *** ( ) ( ) ( ) ( ) lage ( ) ( ) ( ) ( ) lsize *** *** *** *** ( ) ( ) ( ) ( ) Constant 0.110*** 0.110*** 0.109*** 0.110*** ( ) ( ) ( ) ( ) Observations 6,702 6,702 6,702 6,702 R-squared Country and industry dummies included. Robust standard errors in parenthesis. *** p<0.01, ** p<0.05, * p<0.1 The main pairwise unconditional relations explored by the literature on FC and exports and FC and RD seem to hold in our sample. As expected, the less financially constrained is the firm, the higher is its likelihood to export or to invest in R&D activities (see Table 14 and 15). Moreover, as already suggested by descriptive statistics, being an exporter is significantly positively correlated with investing in R&D, and vice-versa. At the same time, being an exporter and investing in R&D are negatively correlated with financial constraints (see Table 16). Firms having a higher total factor productivity, a bigger size or being older have an higher probability to export, while only size has a significant (and positive) effect on the probability to invest in R&D (Table 14 and 15). Bigger or more productive firms are also characterized by a lower need of external finance (Table 16). Results are confirmed when substituting the dummy variables RD and EXP with their intensive margins counterparts, i.e. the percentage of the total turnover that the firm has invested in R&D on average in the previous three years (RDint) and the percentage of the firm total turnover represented by export activities in year 2008 (EXPint) (see Table 17, 18 and 19). These findings seem to suggest the existence of a positive circular relationship between being an exporter, investing in R&D and not being financially constrained. This circular relationship (and the consequent endogeneity) will be addressed in next sub-section. CONCORDi 2013 Page 16 of 24

17 Table 17: Unconditional regression having expint as dependent variable, intensive margin OLS Tobit (1) (2) (3) (4) (5) VARIABLES expint expint expint expint expint Wwdiff *** *** *** (0.0533) (0.0523) (0.0919) RDint 0.479*** 0.469*** 0.729*** (0.0656) (0.0650) (0.0896) ltfp_ *** *** *** *** *** ( ) ( ) ( ) ( ) (0.0150) lage *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) lsize *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) Constant *** *** *** *** *** (0.0214) (0.0213) (0.0225) (0.0223) (0.0380) Observations 6,405 6,403 6,405 6,403 6,403 R-squared Country and industry dummies included. Robust standard errors in parenthesis. *** p<0.01, ** p<0.05, * p<0.1 Table 18: Unconditional regression having RDint as dependent variable, intensive margin OLS Tobit (1) (2) (3) (4) (5) VARIABLES RDint RDint RDint RDint RDint Wwdiff *** ** * (0.0172) (0.0169) (0.0307) expint *** *** *** ( ) ( ) ( ) ltfp_ ( ) ( ) ( ) ( ) ( ) lage ( ) ( ) ( ) ( ) ( ) lsize *** ** ( ) ( ) ( ) ( ) ( ) Constant * ** *** ( ) ( ) ( ) ( ) (0.0124) Observations 6,700 6,403 6,700 6,403 6,403 R-squared Country and industry dummies included. Robust standard errors in parenthesis. *** p<0.01, ** p<0.05, * p<0.1 CONCORDi 2013 Page 17 of 24

18 Table 19: Unconditional regression having WWdiff as dependent variable, intensive margin OLS (1) (2) (3) (4) VARIABLES WWdiff WWdiff WWdiff WWdiff RDint *** ** (0.0110) (0.0112) expint *** *** ( ) ( ) ltfp_ *** *** *** *** ( ) ( ) ( ) ( ) lage ( ) ( ) ( ) ( ) lsize *** *** *** *** ( ) ( ) ( ) ( ) Constant 0.110*** 0.106*** 0.110*** 0.107*** ( ) ( ) ( ) ( ) Observations 6,702 6,405 6,700 6,403 R-squared Country and industry dummies included. Robust standard errors in parenthesis. *** p<0.01, ** p<0.05, * p< SURE regressions In order to analyse in a more comprehensive framework the relationships amongst our variables, we estimate the following econometric model, defined as a system of three equations: WW i = β 10 + β 11 RD i + β 12 EXP i + β 13 ltfp_2007 i + β 14 lage i + β 15 lsize i + u 1i (2a) RD i = β 20 + β 21 WW i + β 22 EXP i + β 23 ltfp_2007 i + β 24 lage i + β 25 lsize i + u 2i (2b) EXP i = β 30 + β 31 WW i + β 32 RD i + β 33 ltfp_2007 i + β 34 lage i + β 35 lsize i + u 3i (2c) We estimate the system of equations using Zellner s seemingly unrelated regression (SURE) technique, by assuming an unstructured variance-covariance matrix of the error terms u. The model is again estimated by including total factor productivity, age and size (all in log) as controls and country and industry fixed effects. The estimation results (for the extensive margin of EXP and RD) are reported in Table 20. CONCORDi 2013 Page 18 of 24

19 Table 20: Simultaneous equations model, SURE. Extensive margins of RD and EXP VARIABLES EXP RD WWdiff EXP 0.368*** *** (0.0118) ( ) RD 0.366*** ** (0.0117) ( ) WWdiff *** ** (0.100) (0.101) ltfp_ *** *** (0.0155) (0.0156) ( ) lage * *** *** ( ) ( ) ( ) lsize *** ( ) ( ) ( ) Constant 0.209*** *** (0.0418) (0.0419) ( ) Observations 6,702 6,702 6,702 R-squared Country and industry dummies included. Standard errors in parenthesis. *** p<0.01, ** p<0.05, * p<0.1 Estimated coefficient have the expected sign, with being an exporter and investing in R&D associated with a decrease in financial constraints and financial constraints associated with a decrease in the probability of being engaged in export or R&D activities. As in unconditional regressions instead investing in R&D is positively related to the probability of exporting, and vice-versa. A higher total factor productivity is positively associated with the probability of export, while it causes a decrease in financial constraints. A higher number of employees have a positive effect on export, while it has a non significant effect on R&D and need of external finance. Finally, a longer experience causes an increase in the probability to invest in R&D and a decrease in financial constraints. Interestingly however it is also associated with a decrease in export propensity. While the sign of the estimated coefficients for financial constraints, export and R&D remain the same when we use the intensive margins of RD and EXP, the sign of control variables coefficients change (see Table 21). Total factor productivity of 2007, size and age now have a significant and negative effect on the percentage of the total turnover that the firm has invested in R&D in the period At the contrary age has a positive effect of the share of 2008 annual turnover represented by the export activities. Coefficients of control variables for equation (3) do not change in significance or sign. CONCORDi 2013 Page 19 of 24

20 Table 21: Simultaneous equations model, SURE. Intensive margins of RD and EXP (1) (2) (3) VARIABLES expint RDint WWdiff expint *** *** ( ) ( ) RDint 0.909*** *** (0.0415) (0.0103) WWdiff *** *** (0.0510) (0.0151) ltfp_ *** *** *** ( ) ( ) ( ) lssize *** ** *** ( ) ( ) ( ) lage *** ** ( ) ( ) ( ) Constant ** *** 0.105*** (0.0212) ( ) ( ) Observations 6,403 6,403 6,403 R-squared Country and industry dummies included. Standard errors in parenthesis. *** p<0.01, ** p<0.05, * p< Robustness checks In this section we replicate the estimation of the three simultaneous equations (2a), (2b) and (2c) restricting our analysis to small and medium sized enterprises (SMEs), i.e. those firms with less than 249 employees, that are likely to be more opaque than large firms and thus more affected by financial constraints and by sunk cost barriers. The estimation results are reported in Table 22. Results confirm the robustness of our previous findings. Table 22: Simultaneous equations model, SURE. SMEs sub-sample (1) (2) (3) VARIABLES EXP RD WWdiff EXP 0.364*** *** (0.0122) ( ) RD 0.361*** ** (0.0122) ( ) WWdiff *** ** (0.105) (0.106) ltfp_ *** *** (0.0162) (0.0163) ( ) lage *** *** (0.0104) (0.0103) ( ) lsize *** * ( ) ( ) ( ) Constant 0.167*** ** 0.116*** (0.0463) (0.0465) ( ) Observations 6,254 6,254 6,254 R-squared CONCORDi 2013 Page 20 of 24

21 Country and industry dummies included. Standard errors in parenthesis. Observations *** p<0.01, ** p<0.05, * p< Conclusions and policy implications We add several contributions to the existing literature. First, to the best of our knowledge, this is the first contribution providing empirical evidence on the mutual causal directions among FC, innovation and exporting activities in a simultaneous framework. Secondly, we analyze the effect of FC and export on both the decision to invest in R&D activities (the R&D participation decision) and the amount of resources devoted to such activities (the intensity of R&D investments). Similarly, we account for both the export extensive margin (e.g.: the decision to go international or not) and the intensive margin (e.g.: the percentage of sales from exporting on total turnover). Finally, we provide a focus on small and medium enterprises (SMEs). Our results have relevant implications for the general debate on firms competitiveness. While the policy community tends to agree that R&D activities, or the presence in foreign markets, or the supply of cheap credit can all be associated to firms competitiveness, these policies tend to be fragmented at the EU level both across countries, as well as across domains: different countries run different policy schemes on internationalization, innovation or access to credit, and all these schemes do not necessarily address the correlations between policy areas that we have detected. The empirical evidence provided on the mutual causal directions among FC, innovation and exporting activities in a simultaneous framework should allow for a more precise design of policy actions, in which appropriate forms of complementarities across domains, or benefits from coordination across countries, can be achieved. CONCORDi 2013 Page 21 of 24

22 ANNEXES References Aghion P., Askenazy P. Berman N., Cette G. and Eymard L. (2012), Credit Constraint and the Cyclicality of R&D Investment: Evidence from France, Journal of the European Economic Association, 10(5), Altomonte, Carlo, Aquilante,Tommaso and Ottaviano, Gianmarco (2012). "The triggers of competitiveness: The EFIGE cross-country report," Blueprints, Bruegel, number 738. Arrow K. (1962), Economic welfare and the allocation of resources for invention. In The Rate and Direction of Inventive Activity: Economic and Social Factors, NBER Chapters, National Bureau of Economic Research, Inc. Bellone F., Musso, P., Nesta L., and Schiavo S. (2010), Financial Constraints and Firm Export Behaviour, World Economy, 33(3), Bratti M. and Felice G. (2011), Are exporters more likely to introduce product innovations?, The World Economy, 35(11), Bustos P. (2011), Trade liberalization, exports, and technology upgrading: Evidence on the impact of Mercosur on Argentinian firms, American Economic Review, 101(1), Bond S., Harhoff D. and Van Reenen J. (2005), Investment, R&D and Financial Constraints in Britain and Germany, Annales d'economie et de Statistique., N.79/80. Campa J. M. and Shaver J. M. (2002), Exporting and capital investment: On the strategic behavior of exporters, IESE Research Papers D/469, IESE Business School. Cassiman B., Golovko E. and Martnez-Ros E. (2010),, exports and productivity, International Journal of Industrial Organization, 28, Cassiman B. And E. Golovko (2011), and internationalization through exports, Journal of International Business Studies, 42, Czarnitzki D. and Hottenrott H. (2011), R&D investment and financing constraints of small and medium sized firms, Small Business Economics, 36(1), Damijan J. P., Kostevc C. and Polanec C. (2010), From to Exporting or Vice Versa?, The World Economy, 33(3), Gorodnichenko, Y. and Schnitzer, M. (2010), Financial constraints and innovation: Why poor countries don t catch up, NBER Working Papers 15792, National Bureau of Economic Research, Inc. CONCORDi 2013 Page 22 of 24

23 Greenaway D., Guariglia A. and Kneller R. (2007), Financial factors and exporting decisions, Journal of International Economics, 73(2), Fazzari S., Hubbard R. G. and Petersen B. C. (1988) Financing constraints and corporate investment, NBER Working Papers 2387, National Bureau of Economic Research, Inc. Hadloc C.J. and Pierce J.R. (2010), " New evidence on measuring financial constraints: Moving beyond the KZ index", Review of Financial Studies,23(5), Hajivassiliou V. and Savignac F. (2011), Novel Approaches to Coherency Conditions in LDV Models with an Application to Interactions between Financing Constraints and a Firms Decision and Ability to Innovate, mimeo. Harris R. and Moffat J. (2011), R&D, and Exporting, SECR Discussion Paper n. 73. Himmelberg C.P. and Petersen B.C. (1994), R&D and Internal Finance: A Panel Study of Small Firms in High-Tech Industries, Review of Economics and Statistics 76, Kaplan S. and Zingales L. (1997), Do Investment-Cash Flow Sensitivities Provide Useful Measures of Financing Constraints?, Quarterly Journal of Economics, 112(1), Levinsohn, J. and Petrin, A. (2003). "Estimating Production Functions Using Inputs to Control for Unobservables", Review of Economic Studies, 70(2), Mancusi M. L. and Vezzulli A. (2010), R&D, innovation and liquidity constraints, KITeS Working Papers 030, KITeS, Centre for Knowledge, Internationalization and Technology Studies, Universita Bocconi, Milano, Italy. Manova K. (2013), Credit constraints, heterogeneous firms, and international trade, Review of Economic Studies, (forthcoming). Manova K., Wei S. J., and Zhang Z. (2011), Firm exports and multinational activity under credit constraints. NBER Working Papers 16905, National Bureau of Economic Research, Inc. Mayer T. and Ottaviano G.I.P., (2007), The Happy Few: new facts on the internationalisation of European firms, Blueprint 3, Bruegel. Melitz M. J., (2003), The impact of trade on intra-industry reallocations and aggregate industry productivity, Econometrica, 71(6), Melitz M. and Costantini J. (2007), The Dynamics of Firm-Level Adjustment to Trade Liberalization. In The Organization of Firms in a Global Economy, E Helpman, Marin, D, and Verdier, T. Cambridge: Harvard University Press Minetti R., Zhu S. C., (2011), Credit constraints and firm export: Microeconomic evidence from Italy, Journal of International Economics, 83, CONCORDi 2013 Page 23 of 24

24 Myers S.C. and Majluf N. (1984), Corporate Financing and Investment Decisions When Firms Have Informations That Investors Do Not Have, Journal of Financial Economics, 13(2), Mulkay B., Hall B.H. and Mairesse J. (2001), Investment and R&D in France and in the United States, in Deutsche Bundesbank (ed.), Investing Today for the World of Tomorrow. Springer Verlag. Savignac F. (2008), Impact of Financial Constraints on : What Can Be Learned from a Direct Measure?, Economics of and New Technology, 17(6), Stiglitz J. and Weiss A. (1981) Credit rationing in markets with imperfect information. American Economic Review, 71: Tiwari A.K., Mohnen P., Palm F.C. and Van der Loeff S.S. (2007), Financial Constraints and R&D Investment: Evidence from CIS, UNU-MERIT Working Paper No. 11. Vanbeveren I. and Vandenbussche H., (2010), Product and process innovation and firms decision to export, Journal of Economic Policy Reform, 13, Whited T.M. and Wu G. (2006), Financial Constraints Risk, Review of Financial Studies, 19(2), CONCORDi 2013 Page 24 of 24