How to Finance Innovation Persistently? A Panel Data Study on Exporting Firms in Sweden

Transcription

1 European Commission Joint Research Centre - Institute for Prospective Technological Studies Knowledge for Growth Economics of Industrial Research & Innovation (IRI) How to Finance Innovation Persistently? Authors Hans Lööf 1, Pardis Nabavi 2 1) 2) Centre of Excellence for Science and Innovation Studies, Royal Institute of Technology, Stockholm Centre of Excellence for Science and Innovation Studies, Royal Institute of Technology, Stockholm 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 < CONFERENCE TOPIC > R&D and innovation: Sources and constraints at company level Industrial dynamics & the role of R&D and innovation for Europe's competitiveness New avenues for policy and for management practices File name: < How to Finance Innovation Persistently_CONCORDi-2013 > Author: <Hans Lööf> contact t.b. in charge for correspondence Authors' contact: <hans.loof@indek.kth.se; pardis.nabavi@indek.kth.se> Status: <Draft> Last updated: < > Organisation: < Royal Institute of Technology, Stockholm > Page 1 of 24

2 Preliminary version, please do not cite. Abstract This paper examines variation in patenting activity across the business cycle among exporting firms in Sweden. We consider the importance of both persistent presence on the export markets and equity supply. Our data cover the period which includes two economic booms, and two recessions as well. During the economic downturn related to the burst of the IT-bubble, the applications dropped with over 40 per cent. Our data show that the recession periods had almost none effect on the patenting activity among persistent exporters, while the econometric analyses also revealed that high-equity firms among the whole group of exporters were able to maintain a smooth patenting profile over the business cycle. Key words: Exporting, Innovation, Financing constraints, Firm level, Panel data JEL classification: C16, F14, G32, O31 CONCORDi-2013 Page 2 of 12

3 TABLE OF CONTENTS 1 - Introduction Data description Empirical Approach Results Conclusion ANNEXES References Table section CONCORDi-2013 Page 3 of 12

4 1 - Introduction The paper considers the relationships between exporters' innovative activity and economic fluctuations across a volatile time period. The period we focus on, starts with the economic boom of the late 1990s, and it continues with the downturn related to the burst of the IT-bubble in the early 2000s, and then the growth period after the ITdebacle, and it ends with the financial crisis and the two following stagnation years. Recent research discusses the importance of sustained recurrence of the innovationengagement. But while the literature on innovative activity shows the advantages with innovation as persistent and stable activity, firms access to finance is typically volatile and highly affected by both cash-flow and supply of equity. Is this also the case for the selected group of firms that are exporters? The existing literature on the performance of exporting firms has largely overlooked the issue of financing innovation. Thanks to the accessibility to extensive micro data provided by Statistics Sweden over the period , and the OECD/EPO PATSTAT data, this study aims to full this gap by testing whether improved overall access to equity capital for different categories of exporters might allow firms to maintain a smooth innovation profile over time. A growing empirical literature using firm or plant level data documents that exporters are more productive than other firms due to factors such as the direct effect of exporting on future productivity (learning-by-exporting) and self-selection of the productive firms into the export market. The selection process might also be associated with innovation efforts. The theoretical prediction is that innovative firms, with attractive products, efficient production technology, and high productivity become exporters (Grossman and Helpman 1995). Using micro data, some recent works provide empirical evidence that innovation and exports are interrelated (Cassiman, Golvko and Martinez-Ros, 2010; Lileeva and Trefler, 2010; Aw et al., 2007; Aw et al, 2011). CONCORDi-2013 Page 4 of 12

5 However, in a comprehensive study that includes all of Sweden's exporting companies within the manufacturing industry, with at least 10 employees, Lööf and Nabavi (2013) show that most of these companies cannot be considered as frequent innovators. Not more than 10 per cent of the exporting Swedish firms applied for patent(s) during at least two years between 1997 and 2008, and only 20 per cent renewed at least 4 export products during this 12-year period. Lööf and Nabavi estimate that the level of productivity is 3-12 per cent higher for persistent innovative exporters compared to other exporters, ceteris paribus, depending on how innovation is defined and where the innovator is located. The annual long-run growth rate is higher for persistent innovative exporters. In a previous study, Martinsson and Lööf (2013) find that patent applications among Swedish firms dropped substantially during the economic downturn, following the burst of the IT-bubble in the beginning of the 2000s, but that the downturn had little effect on the patenting of high-equity firms. They show that firms with the best supply of equity, other things equal, can maintain their patenting strategy over time, whereas firms with less equity experience drops in the number of patent applications when internal equity wanes. This paper takes the analysis further by considering the export market as well as persistent presence on the export market and the long history of a firm s innovation engagement. Business cycle effects matter in the context of access to finance because, during booms when overall demand is high, firms that otherwise are constrained financially can fund their operations and innovate due to the cash-flow generation caused by the high demand. Firms that are capable of being persistent innovators develop internal capabilities, which enable them to benefit from knowledge spillovers; they also tend to be less sensitive to adverse macroeconomic shocks. (Geroski et al., 1997; Cohen and Levinthal, 1990 Similar to Martinsson and Lööf (2013) we separate our sample into two groups depending on their average equity-ratio over the time period. Firms in the top quartile CONCORDi-2013 Page 5 of 12

6 are referred to as high-equity firms, while firms in quartiles 1-3 are referred as the rest of the firms. We test the importance of equity-financing among exporters formally by adopting the pecking order approach behind investment-cash flow sensitivity analyses first introduced by Fazzari et al. (1988). The prediction from this theory is that firms with good access to external finance (equity) are not sensitive to variation in cash-flow over time in their innovation activities. In contrast, firms with less equity are displaying sensitivity of investment to cash-flow. There are a number of studies documenting financial effects on firm-level innovation. Investments in innovation are difficult to finance externally (Hall (2002) and Hall and Lerner (2010)) for plenty of reasons including, i) large information asymmetries between firms and investors, ii) returns to innovative projects are often skewed and highly variable, and iii) little collateral value. While any type of financing is plagued by moral hazard and adverse selection, the financing of innovation is probably the most vulnerable to these problems (Arrow (1962)) since the information that needs to be conveyed is hard to communicate to outsiders. Large information asymmetries makes it difficult to raise any kind of external finance, but, especially, ii) and iii) impede the use of debt financing. Equity holders share in the upside returns and thus skewness and variability need not be a problem; in contrast, the very nature of the debt contract is not well suited for investments with high uncertainty such as long-term innovative investments since creditors share only in the low returns associated with failure. Furthermore, collateral is not relevant for equity finance whereas banks almost always require risky firms to post collateral to obtain debt finance (e.g., Berger and Udell (1990)). These reasons imply that innovation is dependent on equity finance. Another strand of the literature on innovation and economic volatility emphasizes persistency of innovation, and refers to the learning and maintenance effects of being a recurrent innovator. Griliches (1995), Geroski, van Reenen and Walters (1997), Hall (2007), and others suggest that persistent innovation efforts over the business cycle increase the stock of knowledge assets in the firm through technical solutions as well as other business routines and novel product varieties. In addition, with persistent R&D efforts a firm builds up skills, procedures and routines of how to carry out innovation CONCORDi-2013 Page 6 of 12

7 activities (Nelson and Winter (1982); Teece (2007)). This creates a self-enforcing effect implying that profitable firms and/or firms with access to external finance are able to better preserve their innovation activities. A firm's, and an industry's, or even a nation's capacity to progress technologically underpins its long-run economic performance (Cohen, 2010). Recent research has shown important long-run growth implications of persistent innovation (Dosi and Nelson (2010)), and its interrelationship with presence on global markets. Therefore, the results might have implications for policies designed to encourage innovation and growth. If the innovativeness of exporting firms correlates systematically with the equity intensity across economic fluctuations, there might be arguments for policies designed to stimulate equity-financing of innovation activities. Our data show that the recession periods had almost none effect on the patenting activity among persistent exporters, while the econometric analyses also revealed that high-equity firms among the whole group of exporters were able to maintain a smooth patenting profile over the business cycle. The rest of this paper is organized as follows. Section 2 presents the data, define the variables and report descriptive statistics. Section 3 reveals the empirical approach. The results are reported in section 4 and section 5 present some conclusion. 2 - Data description The data set is composed from several different sources. First, audited register information from the annual accounts of all firms in Sweden provided by Statistic Sweden (SCB). Second, information on the educational background of all the employees in the firms observed. Third, trade statistics for all manufacturing firms in Sweden over the same period. The final data source is patent statistics from the EPO Worldwide Statistical Database (PATSTAT) supplemented by national data from the Swedish Patent Office. In the merging process we have managed to match 75% of the patent applications in PATSTAT. The using data includes information on all CONCORDi-2013 Page 7 of 12

8 manufacturing firms with at least 10 employees in average over the observed period, and at least one year of exporting. We have made some restriction for construction the using data set. First we excluded all non-manufacturing firms, since the trade data only covers exports of goods. Second, since the trade data has some quality problems for the very small firms, therefore exclude firms with average number of workers below 10 during the sample period. Third, in order to be included onto the sample, a firm must have exported during at least one year over the period we consider. Forth, we also exclude firms with obvious erroneous observations. Following Brown et al. (2009), Fazzari et al. (1988) and Scellato (2007), all firms with negative sums of cash-flow or total assets during the sample period are dropped. We also eliminate implausible values such as negative sales and equity figures, etc 1. Following the sample construction we end up with an unbalanced panel of about 8,104 unique firms for the period About 15% of the exporter firms applied at least once for a patent during the sample period. Focusing on persistent exporters who have exported during the whole period, we observe 2,444 unique firms. Among persistent exporters, 30% of the firms applied for a patent at least 1 year during this 15-year period. In a previous study, Martinsson and Lööf (2013) showed that equity supply is important for firms in order to maintain a consistent patenting activity over time. But the authors did not address the issues why equity supply differs across firms. Neither were they able to control for persistency in innovation engagement. This paper hypothesizes that the importance of equity is closely related to persistency in exporting as well as persistency in innovation. We separate our sample into two groups based on their average equity-ratio over the time period, firms in the top quartile are referred to as high-equity firms, the rest is denoted other firms. Moreover, we also split the sample based on their export activities. Those firms who have exported during all years are persistent exporters and we report the descriptive statistic for them separately. In figure 1 it is clear that the patent profile 1 All the financial variables are winsorized at 1% level. CONCORDi-2013 Page 8 of 12

9 during the recession only dropped among non-persistent exporters. The persistent exporters display some annual variation, but do not share the development of all firms. Figure 1 Number of patent applications across different types of exporters Figure 2 presents the total number of patent applications for high-equity firms and all the other firms. It can be seen that the high-equity firms were able to maintain their patent profile also during the recession. Regarding the end of the period considered in the study, we see a dramatic decline if the patent applications. However, this is largely explained by shortcomings in the PATSTAT to cover the most recent periods. The reason for including year 2010 and 2011 into the analysis is that we believe that there is no systematic pattern in the problem with the PATSTAT across firms and sectors.. CONCORDi-2013 Page 9 of 12

10 Figure 2 Number of patent applications across equity-ratio groups Tables 1 and 2 present descriptive statistics for all exporting firms and all persistent exporters respectively, dividing each into two groups; all firms and only firms with at least one year of patent applications. Each group is also dividing into high-equity firms and other firms. [Insert table 1 about here] We measure profitability by Cash-flow which is defined as after-tax income plus depreciation and amortization divided by the beginning of the period total assets. Sales is the firms sale divided by the beginning of the period total assets. Firms Long-term debt is also divided by the beginning of the period total assets. Equity is total equity divided by the beginning of the period total assets. We measure employment as log values of total employees (number of employees is reported in the descriptive statistic table). Human capital is the fraction of employees with at least 3 years of university degree. We also have ownership indicator, which distinguishes between non-affiliate, members of a domestic group, a domestic multinational group and a foreign multinational group. CONCORDi-2013 Page 10 of 12

11 Following OECD-suggested classifications, we separate manufacturing firms into four broad sectors based on R&D and human capital intensity: high technology, mediumhigh technology, medium-low technology, low technology. The difference between all firms and patenting firms is mainly in terms of firm size and human capital. Larger fractions of patenting firms belong to high-tech manufacturing sectors and are owned by foreign multinational firms. This is consistent both for all exporting firms and persistent exporters. Concerning exporters with different equity-ratios, high-equity firms are on average smaller, and use less long-term debt. The high-equity group also have slightly higher human capital and higher share of high-tech firms. A similar pattern can be found among persistent exporters. 3 - Empirical Approach We use two categories of models. The first is a binary outcome model estimating the determinants of the probability of being a persistent exporter and a persistent innovative firm, respectively. The outcome model applied is probit estimator, specified as Y it 0 1PI i 2EQit 3LTDit 4Eit 5HCit it it i t (1) where PI is an indicator for being a persistent innovator and this measure is assumed to measure the firms accumulated knowledge stock, EQ and LTD are equity capital and long term debt, both normalized by total assets in period t-1, E is total employment, HC measure share of employees with at least 3 years of university education, it is the specific sector effect of which firm i belongs to in time t, it is the specific effect of the ownership structure of firm i in time t, i is firm specific fixed effect and t is a time specific effect. Our second category of models belongs to the count-data family and we consider both Poisson and negative binomial. The dependent variable in both models is number of patent applications, and we specify the model as CONCORDi-2013 Page 11 of 12

12 Q How to Finance Innovation Persistently? it 0 1CFit 2CFit 1 3Sit 4 Sit 5LTDit 6Eit 7HCit it it i t (2) where CF is cash-flow, S is sales, normalized by total assets in period t-1. The main difference between the two count data models is that while the negative binomial accounts for over dispersion (equality of mean and variance of Q) and therefore also presence of unobserved firm heterogeneity, the variance of Q it in the Poisson model is equal to the mean. This has the consequence that the unobservable heterogeneity i is ruled out. The motivation for using both models is that the distribution of Q it varies across both time and sectors and none of them is superior to the other over the considered period. However, we consider the negative binomial as the preferred model. 4 - Results Table 3 shows the estimates for high-equity firms and other firms obtained from the probit model. The aim of this analysis is to examine financial factors as determinants of the persistency behaviour. Columns 1 and 2 report results for the probability to be a persistent innovator and columns 3 and 4 report results for the probability to be a persistent exporter. The table shows estimates for both high equity exporters and for exporters with less equity capital. [Insert table 3 about here] Persistent exporting has positive impact on the probability of sustained innovation, while firms persistently innovating are more likely to stay on the export market. The estimated coefficients are highly significant for both high-equity firms and other firms. Concerning the financial variables, the table reports that the impact of equity capital on the likelihood of being a persistent exporter is positive among the firms irrespectively of the equity group (high or not high). The equity variable enters with different importance for high equity firm and for other firms. It seems that when the equity has reached a particular threshold the variation has no impact on the likelihood of persistent exporting or innovation. While long term debt correlates positively with persistent innovation for both CONCORDi-2013 Page 12 of 12

13 high equity firms and other firms, it has a significant and positive impact on persistency of exporting only among high equity firms. Tables 4 and 5 report the regression results for exporting firms and persistently firms, respectively obtained from count-data estimation of the negative binomial and Poisson models. The preferred model is the one accounting for over dispersion (negative binomial) and the results are shown in Table 4A and Table 5A. The Poisson estimates are given in Table 4B and 5B. [Insert table 4A and 4B about here] The structure of tables is as follow. The left part of the table provides estimates for high equity firms, while the right part shows corresponding estimates for non-high equity firms (other firms). Columns (i) and (iv) report the estimates for the whole period for the two categories of firms. Columns (ii) and (v) display results for the nonrecession periods and Columns (iii) and (vi) report regression results for the recession periods and [Insert table 5A and 5B about here] Starting with the results for all exporting firms displayed in Table 4, we first consider the dummy variable for persistent innovation (patent applications five years or more). Blundell et al. (1995) shows that not controlling for persistent innovation would cause a potential omitted variable bias in an econometric approach such as ours. The estimates are highly significant and sizable across all six columns, confirming the literature on internal spillovers between innovation processes over time. Our main interest is the cash-flow variable and based on the pecking-order theory, we expect that high-equity firms are less sensitive to variation in cash flow than other firms. This prediction is confirmed in Table 4. The estimated relationship between cash-flow and patent applications is non-significant over the business cycle among high equity firms in both the negative binomial and Poisson regressions. In contrast and as we CONCORDi-2013 Page 13 of 12

14 expected, exporters with non-high equity are sensitive to variations in cash-flow. The preferred negative binomial estimates show significant estimates only for the recession period, while the elasticity of patent application with respect to cash flow is highly significant over the whole period in the Poisson model. The control variables in Table 4 enter with some mixed results across the business cycle for the two equity groups. The most robust results are shown for human capital and employment (positive and highly significant). Figure 1 indicated that the drop in patent applications in economic downturns is restricted to non-persistent exporters, which accounts for about 1/3 of all exporters in our data. The patenting profile among persistent exporters is almost constant over the business cycle. Table 5 explores this issue econometrically and our main interest is whether there is any difference between firms with abundant access to equity financing and other firm. Table 5A reports results from the over dispersion model. The cash-flow estimates for both high equity firms and for other firms is non-significant in the boom periods as well as in the recession periods. Thus, the results suggest that firms with ability to defend the oversee markets over time are also able to find financial means for the innovation activities from other sources than the current cash-flow. Since there is no difference between high-equity firms and other firms, one possible explanation for the result might be that foreign customers or collaboration partners may be actively involved in the innovation processes. The Poisson estimates, however, present deviating results. Table 5B shows no positive association between cash-flow and patenting among high-equity firms. But the estimates for both boom and recession results are positive and significant concerning non high equity firms. The diagnostics indicate that the Poisson may be the inappropriate choice of model, and our main conclusion from Table 5 is therefore based on the over dispersion results. 5 - Conclusion This paper has examined the link between finance and innovation among exporting manufacturing firms in Sweden. We used a data set with firm level observation on about 8,100 unique firms over The frequency of patent application was highly CONCORDi-2013 Page 14 of 12

15 volatile over the business cycle studied and dropped by more than 40 per cent between 1998 and The data showed large difference in the patent profile over time between persistent and temporary exporters. While the persistent exporters were able to smooth the patenting activity through time, non-persistent exporter accounted for the entire volatility in patenting. In the econometric analysis we addressed the issues of both persistency in exporting and persistency of innovating as well as the importance of equity capital and cash-flow. Applying a count-data approach, the regression results show some interesting differences both between all exporting firms and persistent exporters, and between high equity-firms and other firms. Confirming the predictions from the pecking order model, the study found that high-equity firms are not sensitive to cash-flow in the patent applications, irrespectively of whether we considered only persistent exporters or all exporters. Regarding non-high equity firms the result suggests that firms persistently engaged in exports are able to finance their innovation activities also during economic downturns, while temporary exporters with limited access to equity supply have to cut down their innovation activities in recessions when the cash-flow decreases. The concluding finding is that business cycle effects matter in the context of access to finance because, during booms when overall demand is high, firms that otherwise are constrained financially can fund their operations and innovate due to the cash-flow generation caused by the high demand. However, firms that are capable of being persistent innovators and persistently present on export markets tend to be less sensitive to adverse macroeconomic shocks. This is the case for both high equity exporters and for other exporters. CONCORDi-2013 Page 15 of 12

16 ANNEXES References Arrow, K., Economic Welfare and the Allocation of Resources for Invention. In: R. Nelson (ed.) The Rate and Direction of Inventive Activity: Economic and Social Factors, New York: National Bureau of Economic Research, Inc. Aw, B. Y, Roberts, M. J., and Winston, T., (2007). Export Market Participation, Investments in R&D and Worker Training, and the Evolution of Firm Productivity, The World Economy, 30, Aw, B. Y, Roberts, M. J., and D. Y. Xu (2011), R&D investment, exporting, and Productivity Dynamics, American Economic Review. American Economic Association, 101:4, Berger, A. N., Udell, G. F. (1990). Collateral, loan quality, and bank risk, Journal of Monetary Economics, Vol. 25, pp Blundell, R., Griffith, R., Van Reenen, J, "Dynamic Count Data Models of Technological Innovation," Economic Journal, Royal Economic Society, vol. 105(429), pages , March. Brown, J. R., Fazzari, S. M., Petersen, B. C. (2009). Financing innovation and growth: cash- Finance, Vol. 64, pp flow, external equity, and the 1990s R&D boom, Journal of Cassiman, B., & Golovko, E., & Martínez-Ros, E., (2010). "Innovation, exports and productivity," International Journal of Industrial Organization, Elsevier, 28:4, Dosi, G., Nelson, R.R., Technical Change and Industrial Dynamics as Evolutionary Processes. In B. Hall and N. Rosenberg (eds.) The Economics of Innovation, Elsevier. Cohen, W.M., Fifty Years of Empirical Studies of Innovative Activity and Performance. In B. Hall and N. Rosenberg (eds.) The Economics of Innovation, Elsevier. Fazzari, S. M., Hubbard, R. G., Petersen, B. C. (1988). Financing constraints and corporate investment, Brookings Papers on Economic Activity, Vol. 1, pp CONCORDi-2013 Page 16 of 12

17 Geroski, P. A., Van Reenen, J., Walters, C. F. (1997). How persistently do firms innovate?, Research Policy, Vol. 26, pp Grossman, G. M., & Helpman, E., (1995). "Trade Wars and Trade Talks," Journal of Political Economy, 103:4, Hall, B. H. (2002). The financing of research and development, Oxford Review of Economic Policy, Vol. 18, pp Hall, B. H. (2002). "Measuring the Returns to R&D: The Depreciation Problem," NBER Working Papers 13473, National Bureau of Economic Research, Inc. Hall B. H., Lerner, J. (2010). The financing of R&D and innovation, Handbook of the Economics of Innovation, Elsevier-North Holland. Lileeva, A., & Trefler, D., (2010). "Improved Access to Foreign Markets Raises Plant- Level Productivity... for Some Plants," The Quarterly Journal of Economics, MIT Press, 125:3, Lööf, H., Nabavi, P., Learning and Productivity of Swedish Exporting Firms: The importance of Innovation Efforts and the Geography of Innovation, working paper series, CESIS - Centre of Excellence for Science and Innovation Studies 296. Martinsson, G., Lööf, H. (2013) Financial Factors and Patents. in In Andreas Pyka and Esben Sloth Andersen (eds.) Innovation, Organization, Sustainability and Crises. Springer. Nelson R., Winter, S.G., An Evolutionary Theory of Economic Change. Cambridge, Mass: The Belknap of Harvard University Press. Scellato, G. (2007). Patents, firm size and financial constraints: an empirical analysis for a panel of Italian manufacturing firms, Cambridge Journal of Economics, 31, Teece, D. J. (2007), Explicating dynamic capabilities: the nature and microfoundations of (sustainable) enterprise performance. Strat. Mgmt. J., 28: CONCORDi-2013 Page 17 of 12

18 Table section Table 1: Summary statistics, Exporting manufacturing firms in Sweden All firms: 8,104 Patenting firms: 1,211 Equity-ratio groups: Other firms High Other firms High Patent applications Nr export year Cash-flow Mean Median Sales Mean Median Long-term debt Mean Median Equity Mean Median Employment Mean Median Human capital Mean Median High-tech sector Foreign MNE Domestic MNE Domestic UNIE Domestic NAE Notes Other firms and high-equity are divisions based on equity-ratio. We calculate average equity-ratio over the sample period and the bottom 75% are in the other firms-group, and the top 25% are in the highequity group. Cash flow, sales, long-term debt and equity are normalized by beginning of the period total assets. High-tech sector is a dummy indicating 1 if the firm operates in a high-tech sector based on OECD:s classification Human capital is share of employees with at least 3 years of education as a fraction of total employment. Foreign MNE is a dummy variable indicating if it is a multinational enterprise. CONCORDi-2013 Page 18 of 12

19 Table 2: Summary statistics, Persistent Exporting manufacturing firms in Sweden All firms: 2,444 Patenting firms: 725 Equity-ratio groups: Other firms High Equity Other firms High Equity Patent applications Cash-flow Mean Median Sales Mean Median Long-term debt Mean Median Equity Mean Median Employment Mean Median Human capital Mean Median High-tech sector Foreign MNE Domestic MNE Domestic UNIE Domestic NAE Notes Other firms and high-equity are divisions based on equity-ratio. We calculate average equity-ratio over the sample period and the bottom 75% are in the other firms-group, and the top 25% are in the highequity group. Cash flow, sales, long-term debt and equity are normalized by beginning of the period total assets. High-tech sector is a dummy indicating 1 if the firm operates in a high-tech sector based on OECD:s classification Human capital is share of employees with at least 3 years of education as a fraction of total employment. Foreign MNE is a dummy variable indicating if it is a multinational enterprise. CONCORDi-2013 Page 19 of 12

20 Table 3: Propensity to be persistent exporter or Innovators Probit estimates. Dependent variable Persistent innovator Persistent exporter Sample High equity firms Other firms High equity firms Other firms Persistent Innovator *** 0.779*** (0.042) (0.027) Persistent Exporter 0.669*** 0.778*** (0.046) (0.031) Equity *** 0.244*** 0.434*** (0.047) (0.033) (0.081) (0.072) Long-term debt 0.153** 0.224*** 0.485*** (0.062) (0.025) (0.102) (0.058) Employment 0.247*** 0.214*** 0.442*** 0.380*** (0.011) (0.006) (0.017) (0.009) Human capital 0.217** *** 2.333*** 2.661*** (0.088) (0.069) (0.118) (0.100) HT 0.191*** *** 0.916*** 0.521*** (0.038) (0.026) (0.056) (0.046) HTM 0.199*** 0.179*** 0.642*** 0.733*** (0.025) (0.014) (0.048) (0.032) LTM 0.101*** *** 0.566*** 0.538*** (0.023) (0.014) (0.051) (0.035) LT ref ref Ref Ref Domestic MNE 0.043* 0.128*** *** (0.023) (0.014) (0.055) (0.044) Foreign MNE 0.351*** 0.319*** 0.393*** 0.288*** (0.028) (0.017) (0.047) (0.035) Independent 0.265*** 0.195*** 0.245*** 0.113*** (0.035) (0.021) (0.056) (0.041) Domestic UNIE Ref Ref Ref Ref Observations 18,531 54,787 18,531 54,787 Unique firms 1,874 6,323 1,874 6,323 Notes Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 Equity long-term debt and equity are normalized by beginning of the period total assets. Human capital is number of employees with at least 3 years of education as a fraction of total CONCORDi-2013 Page 20 of 12

21 Table 4A: Patent applications , all exporting firms. Negative binomial model Whole period High equity firms Recession Whole period Other firms Nonrecession Nonrecession Recession Persistent innov *** 3.239*** 2.880*** 3.293*** 3.567*** 3.377*** (0.104) (0.119) (0.142) (0.086) (0.090) (0.102) Cash-Flow ** ** (0.182) (0.264) (0.265) (0.118) (0.151) (0.203) Cash-Flow, lag 0.356** ** (0.181) (0.254) (0.277) (0.115) (0.151) (0.188) Sales ** *** *** *** ** (0.073) (0.097) (0.108) (0.043) (0.056) (0.069) Sales growth * (0.071) (0.096) (0.110) (0.044) (0.059) (0.068) Long-term debt 0.403** 0.413* *** *** (0.186) (0.233) (0.331) (0.111) (0.146) (0.174) Employment 0.315*** 0.324*** 0.395*** 0.166*** 0.172*** 0.218*** (0.041) (0.050) (0.059) (0.021) (0.027) (0.029) Human capital 2.351*** 2.269*** 2.110*** 1.888*** 2.330*** 2.165*** (0.312) (0.366) (0.456) (0.222) (0.281) (0.305) HT 0.349*** 0.537*** *** 0.675*** 0.870*** (0.133) (0.164) (0.192) (0.100) (0.127) (0.141) HTM 0.371*** 0.462*** 0.259* 0.605*** 0.562*** 0.692*** (0.107) (0.142) (0.150) (0.073) (0.093) (0.106) LTM 0.437*** 0.635*** 0.311* 0.231*** 0.239** 0.266** (0.120) (0.154) (0.169) (0.082) (0.104) (0.120) Domestic MNE ** (0.129) (0.191) (0.183) (0.095) (0.138) (0.135) Foreign MNE 0.477*** 0.553*** 0.429*** 0.370*** 0.501*** 0.331*** (0.107) (0.176) (0.137) (0.077) (0.125) (0.099) Independent 0.448*** 0.391** 0.681*** 0.407*** 0.529*** 0.358*** (0.116) (0.188) (0.158) (0.080) (0.129) (0.109) Domestic UNE Ref Ref Ref Ref Ref Ref Observations 16,105 8,818 7,287 46,288 25,386 20,902 Unique firms 1,732 1,644 1,632 5,776 5,387 5,308 Notes: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 Cash flow, sales, long-term debt and equity are normalized by beginning of the period total assets. Human capital is share of employees with at least 3 years of education as a fraction of total employment. CONCORDi-2013 Page 21 of 12

22 Table 4B: Patent applications , all exporting firms. Poisson model Whole period High equity firms Recession Whole period Other firms Nonrecession Nonrecession Recession Persistent innov *** 3.279*** 3.178*** 4.505*** 4.749*** 3.701*** (0.255) (0.235) (0.253) (0.184) (0.184) (0.173) Cash-Flow *** 0.478*** 0.451*** (0.114) (0.157) (0.190) (0.044) (0.055) (0.098) Cash-Flow, lag 0.192* *** *** * (0.112) (0.150) (0.198) (0.040) (0.051) (0.085) Sales *** (0.059) (0.075) (0.098) (0.025) (0.035) (0.048) Sales growth *** *** (0.048) (0.062) (0.084) (0.023) (0.031) (0.041) Long-term debt 0.521*** 0.622*** *** 0.449*** 0.404*** (0.115) (0.142) (0.247) (0.060) (0.076) (0.114) Employment 0.616*** 0.596*** 0.608*** 0.380*** 0.313*** 0.611*** (0.054) (0.064) (0.073) (0.012) (0.015) (0.027) Human capital 3.923*** 4.841*** 3.212*** 4.402*** 3.592*** 6.553*** (0.302) (0.368) (0.466) (0.104) (0.141) (0.195) HT *** *** 0.404*** 0.282** 0.944*** (0.127) (0.148) (0.211) (0.097) (0.114) (0.174) HTM ** *** 0.181*** 0.487*** (0.099) (0.120) (0.148) (0.052) (0.068) (0.082) LTM * ** (0.147) (0.165) (0.204) (0.077) (0.090) (0.118) Domestic MNE * *** (0.115) (0.164) (0.157) (0.076) (0.106) (0.119) Foreign MNE *** 0.247*** *** (0.085) (0.137) (0.114) (0.035) (0.080) (0.048) Independent 0.202** 0.347** 0.310*** 0.277*** 0.214*** 0.310*** (0.085) (0.150) (0.120) (0.026) (0.076) (0.038) Domestic UNE Ref Ref Ref Ref Ref Ref Observations 16,105 8,818 7,287 46,288 25,386 20,902 Unique firms 1,732 1,644 1,632 5,776 5,387 5,308 Notes: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 Cash flow, sales, long-term debt and equity are normalized by beginning of the period total assets. Human capital is share of employees with at least 3 years of education as a fraction of total employment. CONCORDi-2013 Page 22 of 12

23 Table 5A: Patent applications , persistent exporting firms. Negative binomial model Whole period High equity firms Recession Whole period Other firms Nonrecession Nonrecession Recession Persistent innov *** 3.190*** 2.673*** 3.291*** 3.444*** 3.330*** (0.123) (0.146) (0.176) (0.088) (0.103) (0.117) Cash-Flow * (0.234) (0.323) (0.353) (0.136) (0.178) (0.231) Cash flow, lag * (0.239) (0.316) (0.376) (0.132) (0.177) (0.214) Sales * *** *** (0.098) (0.131) (0.149) (0.054) (0.071) (0.085) Sales growth 0.171* (0.097) (0.132) (0.145) (0.053) (0.074) (0.082) Long-term debt (0.254) (0.315) (0.431) (0.130) (0.172) (0.204) Employment 0.252*** 0.299*** 0.298*** 0.163*** 0.172*** 0.226*** (0.052) (0.061) (0.077) (0.025) (0.032) (0.035) Human capital 1.956*** 2.138*** 1.611** 1.576*** 2.154*** 1.797*** (0.417) (0.488) (0.639) (0.284) (0.361) (0.405) HT ** *** 0.507*** 0.704*** (0.174) (0.213) (0.259) (0.118) (0.152) (0.170) HTM *** 0.481*** 0.459*** (0.138) (0.180) (0.197) (0.090) (0.117) (0.129) LTM 0.342** 0.520*** *** 0.288** 0.310** (0.152) (0.192) (0.217) (0.100) (0.128) (0.145) Domestic MNE (0.164) (0.231) (0.239) (0.117) (0.172) (0.165) Foreign MNE 0.335** ** 0.205** ** (0.136) (0.204) (0.179) (0.089) (0.151) (0.112) Independent *** 0.232** *** (0.151) (0.223) (0.209) (0.093) (0.156) (0.122) Domestic UNE Ref Ref Ref Ref Ref Ref Observations 6,895 3,748 3,147 20,479 11,156 9,323 Unique firms ,811 1,782 1,784 Notes: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 Cash flow, sales, long-term debt and equity are normalized by beginning of the period total assets. Human capital is share of employees with at least 3 years of education as a fraction of total employment. CONCORDi-2013 Page 23 of 12

24 Table 5B: Patent applications , persistent exporting firms. Poisson model Whole period High equity firms Recession Whole period Other firms Nonrecession Nonrecession Recession Persistent innov *** 3.509*** 3.262*** 3.292*** 3.641*** 3.441*** (0.243) (0.194) (0.306) (0.170) (0.168) (0.161) Cash-Flow * * 0.212*** 0.183** 0.285** (0.142) (0.187) (0.257) (0.059) (0.078) (0.114) Cash flow, lag *** (0.141) (0.187) (0.255) (0.059) (0.085) (0.101) Sales *** *** (0.074) (0.095) (0.136) (0.033) (0.046) (0.059) Sales growth *** *** (0.061) (0.082) (0.108) (0.028) (0.039) (0.047) Long-term debt 0.422*** 0.486*** 0.692** 0.670*** 0.793*** 0.567*** (0.154) (0.188) (0.319) (0.063) (0.083) (0.120) Employment 0.444*** 0.480*** 0.440*** 0.840*** 0.619*** 0.802*** (0.064) (0.068) (0.095) (0.028) (0.034) (0.039) Human capital 2.110*** 2.734*** *** 4.136*** 5.928*** (0.471) (0.516) (0.791) (0.143) (0.196) (0.252) HT *** *** 0.168* *** (0.154) (0.181) (0.256) (0.102) (0.118) (0.202) HTM ** (0.123) (0.147) (0.203) (0.059) (0.080) (0.083) LTM *** * (0.200) (0.201) (0.272) (0.083) (0.099) (0.132) Domestic MNE ** *** (0.142) (0.195) (0.205) (0.097) (0.138) (0.137) Foreign MNE *** *** (0.108) (0.153) (0.154) (0.038) (0.088) (0.050) Independent *** (0.113) (0.170) (0.167) (0.031) (0.083) (0.041) Domestic UNE Ref Ref Ref Ref Ref Ref Observations 6,895 3,748 3,147 20,479 11,156 9,323 Unique firms ,811 1,782 1,784 Notes: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 Cash flow, sales, long-term debt and equity are normalized by beginning of the period total assets. Human capital is share of employees with at least 3 years of education as a fraction of total employment. CONCORDi-2013 Page 24 of 12