The impact of credit constraints on foreign direct investment: evidence from firm-level data Preliminary draft Please do not quote David Aristei * Chiara Franco Abstract This paper explores the role of financial constraints on the extensive margin of foreign direct investment of manufacturing firms in seven European countries during the crisis period. Using direct credit rationing indicators and controlling for endogeneity issues, we find that difficulties in accessing external finance significantly reduce the probability of expanding abroad through FDI. Furthermore, we find that the impact of financial constraints significantly varies according to FDI motivations, destination areas and types of production activities. Keywords: credit constraints, FDI, multinational firms JEL Classification: F23; G20; C35 * Department of Economics, University of Perugia, Via Pascoli, 20, 06123 Perugia (Italy); email: david.aristei@unipg.it. Corresponding author, University of Pisa, Department of Political Science, Via Serafini, 3, 56126 Pisa (Italy); email: chiara.franco@unipi.it.
1 Introduction The decision of a firm to enter a foreign market involves extra costs and may affect its external financing needs. As a consequence, difficulties in accessing external funds may represent an additional barrier to firm s foreign expansion. This holds also for multinational firms investing abroad through foreign direct investments (FDI) although they are usually larger and more productive than domestic firms. Much of the empirical literature on financing constraints and firms internationalisation patterns focus on explaining the exporting side of the issue (Wagner 2014), while few other studies consider also the importing side and two-way trading (Aristei and Franco, 2014). One of the crucial finding of those papers is that both extensive and intensive margins of trade activities may suffer, although to a different extent, from credit constraints. Instead, the role of financial frictions on firms foreign direct investments (FDI) has received little attention, despite FDI decisions involve higher fixed costs than any other international activity. De Maeseneire and Claeys (2012) provide evidence on the relevance of external financing constraints for FDI projects of SMEs in Belgium. Buch et al. (2014) analyse the extensive FDI margin of German firms and, consistently with their theoretical model, find that both productivity levels and credit constraints significantly reduce the probability of owning affiliates abroad. Desbordes and Wei (2014) recognise that the decline in external finance availability partly explains the drop in FDI flows during the global financial crisis and highlight the role of source countries financial development on extensive and intensive FDI margins. Similarly, Manova et al. (2015) using Chinese export data find that foreign affiliates are less credit constrained than domestic firms due to their access to both internal capital markets and external capital markets in host countries. In this paper, exploiting detailed firm-level data from the EFIGE survey and adopting the same methodological approach we followed when investigating the effects of credit constraints on trade (see Aristei and Franco, 2014), we add to this empirical literature by providing cross-country evidence on the importance of credit access for firms FDI activities during the crisis. To the best of our knowledge, this is the first study to disaggregate the analysis by FDI characteristics, investigating whether the impact of credit rationing varies according to FDI motivations, destination countries and type of activity. In Section 2 we describe the econometric methods used while in section 3 the dataset is presented. Section 4 discuss about the results and Section 5 proposes a robustness analysis. Section 6 concludes.
2 Econometric Methods As stated above, we use the same methodology adopted in the paper by Aristei and Franco (2014) but with respect to the extensive margin of FDI only. Due to lack of data we are not able to explore the intensive margin. As shown in Manova (2013), financing constraints may hinder a firm s ability to face the fixed costs of entering foreign markets. At the same time, rationing probability reflects firm s credit risk and depends on firm and credit market characteristics. Financing constraints might thus be endogenous with respect to firm s international activities, due to simultaneity and omitted variable bias (Minetti and Zhu, 2011). To tackle this issue, we consider a recursive bivariate probit model for the joint analysis of FDI and credit rationing probabilities: (1) where x i and z i are vectors of control variables and and follow a bivariate standard normal distribution with correlation. Exogeneity of R i can be formally tested by verifying the significance of error correlation: when, univariate probit estimates of and are inconsistent. In the empirical analysis, we will mainly focus on the marginal effect of R i, which allows evaluating the average treatment effect (ATE) of financial constraints on the extensive FDI margin. Based on estimates of model (1), the ATE of R i can be computed as: (2) 3 Data We use data from the European Firms in a Global Economy (EFIGE) survey, which provides cross-sectional information on a representative sample of nearly 15000 manufacturing firms (with more than 10 employees) in seven European countries (Austria, France, Germany, Hungary, Italy, Spain and the United Kingdom) over the period 2007-2009. 1 Exploiting the detailed information on cross-border expansion of firms, we firstly define a binary FDI indicator equal to 1 if firm runs at least part of its production activity abroad through direct investment (i.e., through foreign affiliates/controlled firms). The questionnaire also provides a breakdown of FDI destination areas and information on the types of production 1 See Altomonte and Aquilante (2012) for more details on the survey.
activities carried out abroad (finished products; semi-finished products/components; R&D, engineering and design services; other business services). We are also able to distinguish different types of FDI activities; specifically, we define binary indicators of Horizontal FDI (when the outcome of production abroad is sold in the foreign country), Vertical FDI (when it is re-imported in firm s home country to be used in production, sold in the domestic market or re-exported) and Export-platform FDI (when it is sold directly in third countries where the firm does not produce or where other production facilities are located). From Table 1 we notice some cross-country differences in FDI choices. Austria and the UK have the highest shares of FDI-active firms, while when disaggregating by FDI motivations we notice that vertical FDI is the most common form of foreign investment in France, Italy, Spain and the UK. [Table 1 about here] As in Aristei and Franco (2014), we consider two direct binary indicators of credit rationing, capturing different intensities of financial constraints. Strong rationing identifies rationed firms as those that applied for additional credit during the last year, but their application was rejected. Weak rationing equals 1 for those firms which would have liked to obtain more credit at the market interest rate, but they either did not apply or obtained credit at a higher cost. Figure 1 shows that FDI-active firms have a lower probability of being strongly and weakly rationed. However, differences in observed frequencies with respect to firms not engaged in FDI are statistically insignificant. This may be due to observed and unobserved factors that simultaneously increase (or decrease) rationing and FDI probabilities, which should be taken into account to properly analyse the effect of financing constraints. [Figure 1 about here] Our empirical specification replicates that of Aristei and Franco (2014). Both FDI and rationing equations include firm s age and employees (both included also as squared), turnover, and dummies for part-time employment, quality certification, use of bank debt and increased price margins over costs. We then control for firm s ownership and management, workforce skill composition, R&D and innovation activities over the last three years. Finally,
we include country and sector fixed effects and account for local development using average TFP at the sectoral and regional level. 2 To enhance parameter identifiability, we impose exclusion restrictions and include only in the rationing equation the percentage of total debt held at the firm s main bank, the length of the relationship with this bank and a dummy for collateral requirements, assuming that they directly affect firm s access to credit, while they do not directly influence FDI decisions. 4 Results Table 2 presents marginal effects of the bivariate probit models of extensive FDI margins and credit rationing. Estimated error correlations are positive and significant, suggesting that the unobserved factors affecting FDI and credit constraints are positively correlated and rationing indicators cannot be considered as exogenous. Signs and significance of the coefficients of the control variables in the FDI equation are all in line with expectations. Younger firms, those with lower turnover and more dependent on bank financing have a higher rationing probability, whereas no significant country differences emerge. Instrumental variables are highly significant and increase rationing probability, supporting the validity of our identification strategy. 3 [Table 2 about here] Both strong and weak rationing indicators have a negative and significant impact on FDI decisions. Estimated marginal effects point out that strong rationing reduces the probability of carrying out any FDI activity by 6.57 percentage points, while weak rationing has a much lower impact (2.85%). We further elaborate on the role of financing constraints by disaggregating the analysis by FDI destination areas, motivations and types of production activities. Table 3 presents results distinguishing between single and multiple destinations: while weak rationing affects single destination FDI, strongly rationed firms are hindered when investing in multiple host countries. Conditional on carrying out FDI activities in EU countries, strong rationing also reduces the likelihood of investing outside Europe, especially in China and India, the US and Canada and Latin America. These results, which differ from those obtained by Aristei and 2 Table A1 in the Supplementary Appendix presents complete variable definitions. 3 Instruments exogeneity has been tested by adding the instrumental variables in the FDI equation and testing for their joint significance. Results for strong and weak rationing indicate that instrument exclusion cannot be rejected, with p-values respectively equal to 0.9573 and 0.4829, supporting the hypothesis that the instruments do not directly affect FDI choices.
Franco (2014) for export, suggest that FDI activities imply higher fixed costs that are not common across destination areas. [Table 3 about here] Table 4 shows that the effect of financing constraints significantly varies with respect to FDI motivations. Firms carrying out horizontal FDI are those suffering more both from strong and weak rationing. A possible explanation for this result is that this form of FDI involves shifting the entire production process to the host country thus requiring higher fixed costs than vertical FDI. In the same way, firms investing abroad to produce finished product are more affected by credit rationing than when they produce abroad semi-finished products. [Table 4 about here] Financial constraints do not significantly affect internationalisation of R&D activities. This result supports the hypothesis that performing R&D activities abroad, despite the costs of decentralised research units, implies lower external financing needs than other forms of FDI, due to the fact that investing firms may benefit from the costs already sustained to carry out knowledge intensive activities at home and exploit host country cost advantages. 5. Robustness analysis Table 5 presents results of several robustness checks on the baseline estimates. They follow the same steps and approach used in Aristei and Franco (2014). Firstly, we see whether there is a change in results due to the choice of the set of identifying variables: we therefore include in all specifications additional instruments encompassing overall riskiness and dependence on external financing of each sector. As in Aristei and Franco (2014), using data from the Bureau van Dijk Amadeus database, we compute earnings volatility over 2006-2008 (measured as the sector-region average standard deviation of EBITDA) for firms with more than 10 employees. We then match this variable with our data using EFIGE sectoral and regional identifiers. The reason for which we choose this variable lays in the fact that firms financial constraints can be different following the differences in industry/regional-specific riskiness as this may reverberate on lenders risk taking behaviour as well as credit policies (Laeven and Levine, 2009). We further add two measures accounting for financial dependence at the sectoral and
regional level: the first indicator is a subjective measure obtained by averaging firm s selfassessment of the external financing dependence of its industry (measured on an ordinal scale ranging from 1 ( not dependent at all ) to 5 ( extremely dependent )) by region and sector. One of the drawback of this measure is that it may catch only firm s perception on this industrylevel feature: to address this weakness, we also built an objective measure based on average firms debt ratio at the sector-region level (computed on Amadeus data). As pointed out in several studies (Kroszner et al., 2007; Dell Ariccia, Detragiache, and Rajan (2008)) during a financial crisis period, the sectors relying more on external financing are more vulnerable and can experience reductions in growth rates. Therefore being part of such sectors can increase the extent to which those firms are exposed to rationing probability. As shown in panels a1) and a2), results obtained with these additional instruments largely confirm the evidence obtained in the baseline specifications, supporting the robustness of our identification strategy. 4 In panel b) we include controls for TFP and capital intensity that we do not include in the benchmark models because of high number of missing values. Despite the estimation sample significantly reduces, dropping from 14590 to 7194 firms, the effect of strong rationing remains significant and even increases in absolute terms, confirming the relevance of both real and financial constraints to firms foreign expansion, while the effect of weak rationing turns out to be statistically insignificant. [Table 5 about here] Because of the fact that the inclusion of the share of bank debt over total debt (Bank financing) as a proxy for firms financial conditions can be a determinant of a lower marginal effect of rationing, in panel c) we show the results obtained by re-estimating extensive margins equations by excluding the bank financing control. We find results that are consistent with baseline estimates. We rerun estimates including dummy variables indicating whether the firm strongly relies on export credit (panel d)) and whether it has received any export incentive (panel e)), in order to control for the possibility that the effects of rationing on foreign investment decisions may be altered when international activities heavily relies on external support. Results obtained confirm baseline estimates. 4 The additional instruments proved to be significant in the credit rationing equation of all the bivariate probit models and non significant in explaining firms FDI decisions.
Further additional robustness checks include the re-estimation of benchmark models including a variable measuring whether firm s turnover and/or workforce have decreased during the last year. In this way we take into account that the negative coefficient of rationing may capture the sharp decrease in sales and employment over the crisis period. Robust results are found also in this case. Finally, we check whether results are affected by the underrepresentation of Austria and Hungary inside the EFIGE dataset (respectively 443 and 488 observations): empirical findings are not significantly affected by the country composition of the sample. 6. Conclusions This paper provides empirical evidence on the role of access to finance for FDI activities of European manufacturing firms. Our findings point out that credit rationing significantly lowers the overall extensive FDI margin. Furthermore, the impact of financial constraints significantly varies according to destination areas, FDI motivations and types of production activities. Specifically, investing in multiple host countries, carrying out horizontal FDI activities and producing finished products abroad, are affected to a greater extent by external financing difficulties, and in particular by credit denial, than other forms of cross-border expansion. References Altomonte, C., Aquilante, T., 2012. The EU-EFIGE/Bruegel-Unicredit dataset. Bruegel Working paper, 2012/13. Aristei D., Franco, C., 2014, The role of credit constraints on firms exporting and importing activities, Industrial and Corporate Change 23(6), 1493 1522. Buch, C.M., Kesternich, I., Lipponer, A., Schnitzer, M., 2014. Financial constraints and foreign direct investment: firm-level evidence. Review of World Economics 150 (2), 393-420. Dell'Ariccia, G., Detragiache, E. and Rajan, R. (2008). The real effect of banking crises. Journal of Financial Intermediation 17(1), 89-112. De Maeseneire, W., Claeys, T., 2012. SMEs, foreign direct investment and financial constraints: The case of Belgium, International Business Review 21(3), 408 424 Desbordes, R., Wei, S.-J., 2014. Credit conditions and foreign direct investment during the global financial crisis. Policy Research Working Paper Series 7063, The World Bank. Kroszner, R.S., Laeven, L. and Klingebiel, D. (2007). Banking crises, financial dependence, and growth. Journal of Financial Economics 84(1), 187 228.
Laeven, L. and Levine, R. (2009). Bank governance, regulation and risk taking. Journal of Financial Economics 93, 259 275 Manova, K., 2013. Credit constraints, heterogeneous firms, and international trade. Review of Economic Studies 80 (2), 711 744. Manova, K., Wei, S.-J., Zhang, Z., 2015. Firm exports and multinational activity under credit constraints. Review of Economics and Statistics 97 (3), 574 588. Minetti, R., Zhu, S.C., 2011. Credit constraints and firm export: Microeconomic evidence from Italy. Journal of International Economics 83 (2), 109-125. Wagner, J., 2014. Credit constraints and exports: A survey of empirical studies using firm level data. Industrial and Corporate Change 23 (6), 1477-1492.
Tables Table 1 Extensive margins of foreign direct investment Percentage distribution of FDI-active firms: Country N. of firms FDI (any type) Horizontal FDI Vertical FDI Export-platform FDI AUT 419 7.16 4.51 4.29 3.61 FRA 2965 3.88 1.88 3.53 2.25 GER 2965 5.67 4.97 4.57 3.34 HUN 487 1.98 1.02 0.82 0.41 ITA 3009 2.45 1.59 2.18 1.09 SPA 2832 2.74 1.55 1.80 1.06 UK 2013 6.60 3.53 4.98 3.87 Total 14590 4.00 2.66 3.27 2.21 Notes: percentage frequencies are computed using sample weights.
Table 2 Extensive margin of FDI and credit rationing: marginal effects (1) (2) Strong Rationing FDI Weak Rationing FDI Age -0.0003*** 0.0001-0.0004*** 0.0001 (0.0001) (0.0001) (0.0001) (0.0001) Employees 0.0002*** 0.0001* 0.0001 0.0001* (0.0001) (0.0000) (0.0001) (0.0000) R&D Workforce 0.0047 0.0196*** 0.0058 0.0187*** (0.0048) (0.0050) (0.0074) (0.0045) High Skill Workforce -0.0009 0.0100** 0.0076 0.0104*** (0.0038) (0.0040) (0.0063) (0.0040) Labour Flexibility 0.0029 0.0114*** 0.0045 0.0110*** (0.0030) (0.0038) (0.0049) (0.0036) Individual First Shareh -0.0072 0.0063-0.0000 0.0068 (0.0048) (0.0047) (0.0071) (0.0045) Foreign First Shareh -0.0059 0.0149*** -0.0215** 0.0137*** (0.0081) (0.0056) (0.0099) (0.0053) Group -0.0008 0.0230*** -0.0065 0.0224*** (0.0058) (0.0050) (0.0088) (0.0047) Centralised Decisions 0.0008-0.0093** -0.0063-0.0095*** (0.0035) (0.0036) (0.0057) (0.0036) Family CEO -0.0041 0.0076-0.0043 0.0077* (0.0035) (0.0046) (0.0061) (0.0046) Innovation 0.0045 0.0124*** 0.0202*** 0.0124*** (0.0036) (0.0034) (0.0060) (0.0034) R&D Investment Share 0.0004 0.0006*** 0.0005 0.0005*** (0.0002) (0.0002) (0.0004) (0.0002) Turnover -0.0051** 0.0188*** 0.0005 0.0186*** (0.0023) (0.0020) (0.0034) (0.0020) Increased Margins 0.0029 0.0067-0.0110 0.0055 (0.0073) (0.0056) (0.0120) (0.0054) Quality Certified 0.0057* 0.0108*** 0.0028 0.0102*** (0.0035) (0.0040) (0.0058) (0.0037) Bank Financing 0.0004*** 0.0001* 0.0006*** 0.0001* (0.0000) (0.0001) (0.0001) (0.0001) Mean TFP -0.0291** 0.0144 0.0074 0.0185 (0.0139) (0.0124) (0.0180) (0.0117) Main Bank Share 0.0004*** 0.0015*** (0.0001) (0.0001) Main Bank Length 0.0005*** 0.0017*** (0.0001) (0.0002) Collateral 0.0617*** 0.0919** (0.0159) (0.0409) Austria -0.0127-0.0023-0.0396** -0.0021 (0.0168) (0.0042) (0.0174) (0.0042) France -0.0040-0.0062-0.0395*** -0.0063 (0.0082) (0.0067) (0.0114) (0.0065) Hungary -0.0055-0.0112-0.0500*** -0.0121 (0.0122) (0.0139) (0.0142) (0.0138) Italy 0.0346*** -0.0112 0.0253** -0.0144** (0.0085) (0.0068) (0.0100) (0.0059) Spain 0.0335*** -0.0018 0.0344*** -0.0053 (0.0096) (0.0078) (0.0097) (0.0069) UK -0.0570*** 0.0042-0.0043 0.0078 (0.0090) (0.0057) (0.0095) (0.0052) Strong rationing -0.0657*** (0.0236) Weak Rationing -0.0285* (0.0155) 0.5492*** 0.3086*** (0.1397) (0.1099) Number of firms 14590 14590 Log-Likelihood -3935.79-5743.22 Notes: Table reports average marginal effects. For Age and Employees, reported marginal effects take into account that both the variables are also entered with a quadratic term. Robust standard errors, clustered at the regional level, are reported in parentheses below the estimates. All estimates are obtained using sample weights and include (unreported) sectoral controls. ***, ** and * denote significance at 1, 5 and 10 percent levels, respectively.
Table 3 FDI destinations: marginal effects of credit rationing FDI Destinations: Strong Rationing Weak Rationing Single Destination -0.0123-0.0190** (0.0165) (0.0088) Multiple Destinations -0.0788*** -0.0010 (0.0280) (0.0164) EU -0.0565** -0.0083 (0.0272) (0.0130) Outside EU -0.0532** -0.0229* (0.0250) (0.0136) Other European (Non EU) Countries -0.0393-0.0084 (0.0254) (0.0051) China & India -0.0477** -0.0032 (0.0230) (0.0074) Other Asian Countries -0.0179* -0.0034 (0.0096) (0.0037) USA & Canada -0.0371*** -0.0087* (0.0134) (0.0049) Latin America -0.0516*** -0.0034 (0.0078) (0.0045) Notes: Table reports average marginal effects of strong and weak rationing indicators. Robust standard errors, clustered at the regional level, are reported in parentheses below the estimates. ***, ** and * denote significance at 1, 5 and 10 percent levels, respectively.
Table 4 FDI motivations and types of production activities: marginal effects of credit rationing Strong Rationing Weak Rationing FDI motivations Horizontal FDI -0.0502*** -0.0177* (0.0241) (0.0094) Vertical FDI -0.0319** -0.0149 (0.0163) (0.0138) Export-platform FDI -0.0303*** -0.0119 (0.0087) (0.0089) Types of production activities Finished products -0.0648** -0.0364** (0.0273) (0.0141) Semi-finished products/components -0.0341* -0.0100 (0.0199) (0.0096) R&D, engineering and design services -0.0082-0.0076* (0.0104) (0.0045) Other business services -0.0022-0.0011 (0.0028) (0.0085) Notes: Table reports average marginal effects of strong and weak rationing indicators. Robust standard errors, clustered at the regional level, are reported in parentheses below the estimates. ***, ** and * denote significance at 1, 5 and 10 percent levels, respectively. Table 5 Robustness analysis a) Including additional instruments a1) EBITDA volatility at the regional-sector level and self-assessed sectoral financial dependence a2) EBITDA volatility at the regional-sector level and objective sectoral financial dependence Strong Rationing -0.0641*** Strong Rationing -0.0635*** (0.0225) (0.0237) Weak rationing -0.0284* Weak rationing -0.0278* (0.0158) (0.0157) b) Controlling for TFP and capital intensity c) Excluding bank financing control Strong Rationing -0.1051*** Strong Rationing -0.0599*** (0.0323) (0.0231)
Weak rationing -0.0237 Weak rationing -0.0276* (0.0459) (0.0157) d) Controlling for significantly relying on export credit e) Controlling for having received export incentives Strong Rationing -0.0650*** Strong Rationing -0.0640*** (0.0238) (0.0232) Weak rationing -0.0276* Weak rationing -0.0286* (0.0162) (0.0158) f) Controlling for turnover and/or workforce decrease g) Excluding Austria and Hungary Strong Rationing -0.0650*** Strong Rationing -0.0622*** (0.0223) (0.0241) Weak rationing -0.0290* Weak rationing -0.0266* (0.0156) (0.0162) Notes: Robust standard errors, clustered at the regional level, are reported in parentheses below the estimates. Estimates are obtained using sample weights. All regressions, except c), include the same controls used in the baseline models. In panel a), we include as additional instruments the sector-region average standard deviation of EBITDA for firms with more than 10 employees (computed on Amadeus data) and a self-assessed (a1)) and an objective (a2)) measure of sectoral financial dependence. In panels d), e) and f) additional dummies are included to control for relying on export credit, for having received export incentives and for turnover and workforce decrease in the last year, respectively. Sample size reduces to 7194 and 13684 observations for estimations reported in panels b) and g), respectively. ***, ** and * denote significance at 1, 5 and 10 percent levels, respectively.
Figures Figure 1 Percentage distribution of strongly and weakly rationed firms (conditional on applying for or willing to increase credit) by FDI status
Appendix Table A1 Control variables: definitions and descriptive statistics Variable Definition Mean Std. Dev. Age Years since firm s establishment 34.126 30.573 Employees Total number of employees 51.229 80.757 R&D Workforce High Skill Workforce Equals 1 if the share of R&D employees is higher than the corresponding national average; 0 otherwise Equals 1 if the share of graduate employees is higher than the corresponding national average; 0 otherwise 0.109 0.311 0.278 0.448 Labour Flexibility Equals 1 if firm uses part-time employment or fixed-term contracts; 0 otherwise 0.593 0.491 Individual First Shareh Equals 1 if the first shareholder is an individual or a group of individuals; 0 otherwise 0.768 0.422 Foreign First Shareh Equals 1 if the first shareholder is foreign; 0 otherwise 0.079 0.27 Group Equals 1 if the firm belongs to any kind of group (national or foreign); 0 otherwise Centralised Decisions Equals 1 if the CEO/owner takes most of the decisions in every area; 0 otherwise 0.193 0.395 0.698 0.459 Family CEO Equals 1 if the CEO is the individual (or a member of the family) who owns/controls the firm; 0 otherwise 0.64 0.48 Innovation Equals 1 if the firm has carried out any product or process innovation; 0 otherwise 0.642 0.479 R&D Investment Share R&D investment as a percentage of total turnover 3.452 7.663 Turnover Turnover classes, from 1 ( less than 1 million euro ) to 7 ( more than 250 million euro ) 2.816 1.252 Increased Margins Equals 1 if the size of price margin over costs has increased during the last year; 0 otherwise 0.063 0.244 Quality Certified Equals 1 if firm has any form of quality certification; 0 otherwise 0.571 0.495 Bank Financing Percentage of short and medium-long bank debt over total debt 15.123 29.442 Mean TFP Average TFP at the sectoral and regional level -0.025 0.204 Notes: descriptive statistics are computed using sample weights.
Table A2 Extensive margins of FDI and credit rationing: univariate probit results (1) (2) Strong rationing Weak rationing Credit rationing 0.0081 0.0071 (0.0094) (0.0053) Age 0.0001 0.0001 (0.0001) (0.0001) Employees 0.0001 0.0001* (0.0000) (0.0000) R&D Workforce 0.0183*** 0.0183*** (0.0043) (0.0044) High Skill Workforce 0.0101*** 0.0099*** (0.0038) (0.0038) Labour Flexibility 0.0104*** 0.0104*** (0.0035) (0.0035) Individual First Shareh 0.0068 0.0066 (0.0044) (0.0044) Foreign First Shareh 0.0144*** 0.0144*** (0.0052) (0.0052) Group 0.0224*** 0.0225*** (0.0046) (0.0046) Centralised Decisions -0.0090*** -0.0089*** (0.0034) (0.0034) Family CEO 0.0074* 0.0074*
(0.0044) (0.0044) Innovation 0.0113*** 0.0112*** (0.0031) (0.0031) R&D Investment Share 0.0005** 0.0005*** (0.0002) (0.0002) Turnover 0.0183*** 0.0182*** (0.0020) (0.0019) Increased Margins 0.0062 0.0062 (0.0053) (0.0053) Quality Certified 0.0098*** 0.0098*** (0.0037) (0.0037) Bank Financing 0.0000 0.0000 (0.0000) (0.0000) Mean TFP 0.0189 0.0187 (0.0115) (0.0117) Austria -0.0015-0.0014 (0.0040) (0.0040) France -0.0049-0.0047 (0.0064) (0.0064) Hungary -0.0105-0.0106 (0.0132) (0.0132) Italy -0.0150** -0.0148** (0.0059) (0.0059) Spain -0.0068-0.0068
(0.0070) (0.0069) UK 0.0073 0.0071 (0.0052) (0.0051) Number of firms 14590 14590 Log-likelihood -1891.37-1890.90 Notes: Table reports average marginal effects. For Age and Employees, reported marginal effects take into account that both the variables are also entered with a quadratic term. Robust standard errors, clustered at the regional level, are reported in parentheses below the estimates. All estimates are obtained using sample weights and include (unreported) sectoral controls. ***, ** and * denote significance at 1, 5 and 10 percent levels, respectively.