Innovation Plays A Crucial Role In Firms Economics Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.


It is widely acknowledged that innovation plays a crucial role in firms' survival and growth. Traditionally, innovation has been considered one of the major sources of growth and competitive advantage, both for nations and firms. From the "creative destruction" of Schumpeter, when the key player was represented by the entrepreneur who, with his action, could sustain the long-term economic growth (Schumpeter, 1942), to our recent R&D laboratories, research centers, etc., innovation has always played an extremely relevant role for growth. While at the beginning of economic theory, the role of firms was considered pivotal, nowadays firms' involvement is a key element for innovation. As stated above, the issue is extremely relevant not only for firms, but also at a macro-level, thus considering nations. In fact, "the capability to innovate and to bring innovation successfully to market will be a crucial determinant of the global competitiveness of nations over the coming decade. There is growing awareness among policymakers that innovative activity is the main driver of economic progress and well-being as well as a potential factor in meeting global challenges" (OECD, 2007, p.3). Moreover, innovation is demonstrated to be critical for survival and growth of firms, even more in the high competitive global environment that firms are now compelled to face. Given the importance of the subject, a whole understanding of the factors that consistently influence innovation is extremely important.

Historically, certain factors have always been considered to positively impact on firms' innovation: for instance, size is regarded as a source of innovation (e.g. Ebling & Janz, 1999; Ganotakis & Love, 2011) because greater firms are supposed to own both the financial and knowledge resources necessary to invest in innovation. Among the features that might positively influence innovation, exporting has been recently brought to attention. While traditionally, exporting has only been considered an international expansion strategy, even if the most used one, lately, the positive effects that exports exert on innovation have been repeatedly investigated by scholars. Apart from the relationship with innovation, one of the reasons behind the great attention to exports lies in the great diffusion of these activities among firms. In fact, as a consequence of the changing dynamics of the environment, also the small and medium enterprises (SMEs) are realizing that they can no longer believe international competition will not affect them because they are small or solely focused on the domestic market. Therefore, SMEs are increasing the exporting activities to survive, and hopefully grow, in this challenging environment. As an example, in the data that will be described shortly, more than 60% of the firms in the Spanish manufacturing sector export. All that, has brought to an increased attention to exporting from scholars.

Furthermore, as anticipated above, nowadays it is widely shared the view according to which exports have positive effects on innovation. Several researches have demonstrated that exporters have higher innovative productivity compared to non-exporters. At the same time, studies have also revealed that innovation exert positive influence on exporting. The more a firm innovates, the higher the export intensity (e.g. Kongomanila & Takahashi, 2009; Roper & Love, 2002) and the likelihood that firm exports (e.g. Bleaney & Wakelin, 2002; Caldera, 2010). Therefore, while before exporting and innovation were considered alternative strategies for growth, now they are believed to reinforce each other. Indeed, research have unveiled that the returns in terms of growth of one activity is higher if the firm also engages in the other (Golovko & Valentini, 2011; Ito & Lechevalier, 2010). Innovation may drive to an improvement of the existing products, sold in the domestic market, thus increasing sales, but also, it may lead firms to enter in foreign market with new or modified products. Conversely, operating abroad, exporting firms may come to know diverse and improved knowledge that, once incorporated, might be used in the domestic market to produce better innovation. These concepts are practically embedded in research minds. Yet, despite the great amount of studies aimed at investigating this relationship, scholars' attention has not been focused on the mechanism underlying learning by exporting. Although it is widely acknowledged that export exerts positive influence on innovation, purportedly driven by information exchange from foreign market, through export intermediaries or directly from customers (Salomon & Shaver, 2005a), few studies have tried to understand how firms learn from international trade. Namely, some hypotheses have been advanced but rather unsuccessfully tested. Just to mention the main one, MacGarvie (2006) tried to verify whether informal communication with foreign counterparts is the channel through which knowledge is obtained, thus exerting positive influence on innovation. Since little evidence was found in favor of the hypothesis, how exporters learn from international trade remains unknown. Clearly, disentangling this relationship, thus understanding how firms, but also policymakers, can foster innovation, might be extremely relevant for the development both of firms and of countries.

In this study, I hypothesize that exporting firms improve their innovative performances through formal technology transfer agreement. I try to verify whether the improved knowledge that firms can obtain through formal agreements fosters the number of product innovation that the firm obtains and the number of patents that the firm registers. Therefore, a balanced panel of Spanish manufacturing firms, composed by 10,205 firm-years observations, is used to measure the impacts of the purchase of licenses, technology and technical assistance abroad on product innovation and patent registration counts. The results confirm that, 1 year after the stipulation of a formal technology agreement, firms significantly improve their patent registration; furthermore, if considered with further lag, the purchase of license, technology and technical assistance abroad appears to influence patent registration, although in this case, the evidence is significantly less strong. On the contrary, product innovation does not seem to be affected by the variable of interest, independently of the lags considered.

In the next section, literature linked to the issue is reviewed, followed by the hypotheses of the research. In sections 3, 4, 5 and 6 data are briefly described and descriptive statistics are provided. In the subsequent sections, the method employed is presented, followed by the results of the empirical analysis and additional analyses performed. The final section concludes.

Literature review

The relationship between export, innovation and productivity has been remarkably interesting for scholars. During last decades, several researches have been made to disentangle the complex relationship among these variables. The aim of this study is to go into more depth in understanding how learning by exporting takes place, but, before doing it, an overview on the previous literature's contributions is provided.

Self-selection and determinants of export

Until recently, most of the literature has mainly focused on the link between productivity and exporting. First studies tested the hypothesis that most productive firms self-select into export markets. Self-selection is now embedded into scholars' minds. Research showed that only the most productive firms engage in export activities because they are able to face the high sunk costs of entry and the fiercer competition of international markets (Bernard & Jensen, 1999; Ganotakis & Love, 2011). Hence, a great part of the positive correlation between productivity and exporting is explained by self-selection. Not only future exporters show higher productivity and lower variable costs prior to exporting (Clerides, Lach & Tybout, 1998), but also firms with higher productivity have higher returns on exporting in general, and on exporting to advanced countries in particular (Trofimenko, 2008). Moreover, the greater the productivity plant, the greater the marginal benefits of both exporting and R&D activities, thus increasing the effect of self-selection (Aw, Roberts & Xu, 2010).

It is important to note that studies regarding determinants of exporting have revealed that several variables, apart from innovation and productivity, influence exporting. Although size is considered to exert positive influence on exporting, opinions of the empirical literature are not unanimous. Ebling and Janz (1999), for instance, found that firm size does not affect exports straightaway, but only indirectly, through innovation; investigating French firms in biotechnology, Pla-Barber and Alegre (2007) revealed that size does not impact on export performance; these findings are explained by the features of the industry, where economies of scale and production efficiency are not particularly relevant compared to the competence to develop cutting-edge technologies. As for the presence of foreign equity, instead, and as expected by most of the literature, Wignaraja (2008), investigating a sample of Asian firms, revealed that the presence of foreign equity has a strong positive effect on export performance. Several factors resulted positively correlated to export such as education and well qualified managers (Braymen, Briggs & Boulware, 2010; Higón & Driffield, 2011), responsive marketing organization and customer focused practices (Calantone, Cavusgil, Schmidt & Shin, 2004) and labor skills (Alvarez, 2007). Besides the variables already illustrated, innovation is considered to be correlated to export, although the relationship between these two variables is quite more complex.

The role of innovation and the relationship with export

Innovation is recognized as one of the major sources of heterogeneity among firms, although the relationship between export and innovation is still investigated by researchers. An innovation is defined as "the implementation of a new or significantly improved product (good or service), or process, a new marketing method, or a new organizational method in business practices, workplace organization or external relations" (Eurostat & OECD, 2005, p.46). Traditionally, empirical literature regarded R&D and innovation as interchangeable, thus considering the relationship between R&D and exporting rather than between innovation and exporting (Aw et al., 2010; Barrios, Görg & Strobl, 2003; Braymen et al., 2010). The importance of innovation output has been recently brought into focus by researchers that began to use different variables to measure innovation, such as product and process innovation (e.g. Caldera, 2010; Cassiman, Golovko & Martinez-Ros, 2010; Damijan, Kostevc & Polanec, 2010; Higón & Dreffield, 2011; Van Beveren & Vandenbussche, 2009) and patent application (Salomon & Jin, 2008, 2010; Salomon & Shaver, 2005a,b).

Conventionally, two different models have emerged to describe the relationship between innovation and exporting: on one side, product-cycles models claim that innovation is one of the driving forces of industrialized countries' exports (Vernon, 1966; Krugman, 1979). The core idea is that products follow cycles and developed countries innovate to export to less developed countries. Once the developing countries are able to replicate the innovation by themselves, the initial exporters are forced to innovate again in order to continue exporting. On the other side, endogenous growth models (Grossman & Helpman, 1991a; b) assert that export can be an impulse for innovation for two distinct reasons: first, higher competition in foreign markets leads exporters to invest in R&D to improve products and processes in order to be competitive; furthermore, export helps firms in gaining access to more developed knowledge and technologies, thus stimulating the productivity.

Therefore, product-cycles models claim that innovation has an impact on exporting and, in the wake of these theories, scholars have tried to empirically prove the effects of innovation on export.

The effects of innovation on export

Large part of the literature agrees on the positive effects of innovation on export. It is more likely that a firm engages in exporting when it has previously undertaken innovative activities (Ganotakis & Love, 2011; Kongomanila & Takahashi, 2009; Halpern & MurakÓ§zy, 2012). Basile (2001), examining over 4,000 Italian manufacturing firms, reaches the same conclusion, adding interesting features on the impacts of exchange rate fluctuations on export. Treating innovation as exogenous, the author claims product innovation strategies have strong positive effects on export only after a devaluation of the currency. In fact, during a devaluation of the currency, non-innovative firms are able to enter the foreign market. Finally, not only innovation affects export decision, but it also increases export duration (Chen, 2012).

When considering different industries, the effect of innovation on exports can be quite different. A number of studies (e.g. Barrios et al., 2003; Bleaney & Wakelin, 2002; Caldera, 2010; Lachenmaier & Woessmann, 2004; Nguyen, Pham, Nguyen & Nguyen, 2008; Roper & Love, 2002) have demonstrated that innovation exerts a positive influence on exporting in manufacturing sectors. Exploring this link in different countries and periods, the authors agreed on the positive effect of innovation on export decision, claiming that the likelihood that firms are exporters is higher if they have had more innovations. Whereas literature position is on the same side in manufacturing sector, in other industries studies' results are rather conflicting. Specifically, considering R&D investments as a proxy for innovation of new firms, Braymen et al. (2002) revealed that the relationship between R&D and exports is not significant for new firms that provide only services to their customers. On the other hand, Ebling and Janz (1999), examining more than 1,000 German firms operating in 7 different business oriented service sectors, found that innovation strongly influences export activities. Similarly, investigating the effect of technology and innovation on export in a science-based industry, Pla-Barber and Alegre (2007) unveiled strong positive impacts on export performance.

As stated above, the term innovation applies both to product and process innovation. Concerning the influence of different types of innovation on export performance, scholars' opinions are rather conflicting. Several researchers agree on the importance of product innovation in explaining firms' export decision (e.g. Cassiman & Golovko, 2007; Cassiman et al., 2010; Higón & Driffield, 2011) and intensity (Roper & Love, 2002). Another part of the literature believes that also process innovation is an important determinant for export decision and performance, albeit its contribution is lower than product innovation's (Becker & Egger, 2007; Caldera, 2010). These findings are explained by the greater importance for export performance of differentiation strategy, to adapt products to foreign countries, compared to process improvement, that leads to cost saving. Therefore, also product adaptation is presented as a positively correlated factor with export profitability, even more when the firm has a responsive marketing organization and customer focused practices (Calantone et al., 2004). In a similar vein the contribution provided by Nguyen et al. (2008) which have investigated 2,739 Vietnamese SMEs in the manufacturing industry and, apart from product innovation and process innovation, have considered a third measure of innovation: product modification, consisting in the improvement of existing products. The authors found that also modification of existing products is important as a determinant of exporting activities.

Although the positive effects of innovation on export have been widely proved, nowadays also the view of the impacts of export on innovation is quite shared.

The effects of export on innovation: learning by exporting

As seen above, the empirical literature has historically focused on the positive impact innovation has on exporting. Recently, questions have been raised on whether export influences innovation. To understand this link, and considering innovation as a measure of learning, Salomon and Shaver (2005a) have elaborated on an unbalanced panel of Spanish manufacturing firms. The authors discovered that exporters increase their product innovations after entering into foreign markets, and in particular, this effect is more pronounced two years subsequent to exporting. The authors justify these findings claiming that it takes time to modify products in order to meet customers' needs. Similarly, the authors found a positive effect of exporting on patent applications, although in this case the effect is stronger with further lags since it takes more time to incorporate the technological knowledge needed to register patents. Several more studies confirmed the positive impact of export on process improvement, claiming that previous experiences as medium-large exporter have positive effects on patent applications (Salomon & Jin, 2008; Salomon & Jin, 2010), on both process improvement and product innovations (Alvarez & Robertson, 2004), and on future productivity growth because firms improve efficiency by stimulating process innovation (Damijan et al., 2010). Furthermore, two elements appeared to be important in this matter: destination of the market and technological standing of the industry to which the firm belongs. With regard to market, Alvarez and Robertson (2004) showed that firms that export to developed countries are more likely to innovate in new products, new tools and reorganization of production. Exporting to developing countries, instead, increases the likelihood of engaging in product design activities. Similarly, Trofimenko (2008), investigating how exporting to more developed countries affects productivity gains of Colombian manufacturing firms, asserts that the quality of the learning environment matters, thus exporting to more developed countries increases future productivity gains.

As for the second issue, Salomon and Jin (2008; 2010) divided the sectors to which Spanish manufacturing firms belonged into technological leaders or laggards; results suggested that technological standing of industries significantly influences the amount of knowledge that flows from the host environment to the firm. In fact, firms belonging to technologically lagging industries increase their patent activity disproportionately relative to firms in technologically advanced sectors (Salomon & Jin, 2008), albeit the benefits gained do not permit them to close the gap with their technological leaders counterparts (Salomon & Jin, 2010).

In line with learning by exporting, an interesting study made by Aw and Batra (1998) has sought to analyze the link between efficiency and firm investments in technology and exports. To estimate the technical efficiency of firms by investments in technology as well as by export, the authors used stochastic production frontier techniques on a large sample of Taiwanese manufacturing firms operating in several different industries. The authors stress the importance of technological heterogeneity among firms to understand the relationship between exports and firm productivity. Essentially, for firms that heavily and formally invest in R&D and technology, usually large, with skilled labor and foreign capital, efficiency is not related to export orientation, while for small and locally owned firms, which do not formally invest in R&D, export is strongly linked to higher future productivity. Therefore, researchers justify these findings with the bias that SMEs have of acquiring foreign technology through subcontracting arrangements with foreign firms, and with the consideration of exporting as a channel through which technology and knowledge are passed from foreign markets and incorporated by exporters.

Part of the literature showed how important is the learning by exporting effect for firms but, while self-selection hypothesis has been proved by huge portion of the literature, the evidence on the learning by exporting has been less clear-cut. A growing body of the empirical work has sought to analyze the impact export has on future productivity. Although domestic firms start exporting when having a strength on the domestic market (Salomon & Shaver, 2005b), there is also a strong evidence in favor of positive effects export has on productivity gains (Van Biesebroeck, 2005). According to De Loecker (2007), firms that enter in foreign markets become on average 8.8% more productive than non-exporters. Investigating the effect of the Free Trade Agreement between Canada and USA, Lileeva and Trefler (2010) confirmed that non-exporters that began to export to USA, because of the improved access to foreign market, showed labor productivity gains. Moreover, not only exports bring a decline in marginal costs and increase efficiency and capacity utilization, but also have positive spill-over effects on the mark-up of the domestic markets (Bughin, 1996).

Until now, in all the studies mentioned, the authors solely examined the relationship between export and innovation within the same firm. However, the impacts that other firms' innovation or export might have, in terms of spillover effects, should not be underestimated.

The role of R&D and export spillovers

While there is a fairly large literature investigating the effects of innovation and R&D on exports, fewer scholars have examined R&D and other firms' export spillovers. The concentration of exporting firms and multinational firms has positive effects on export performance, thus export spillover effects are important for export success (Alvarez, 2007). Moreover, R&D spillovers coming from multinationals of the same sectors have positive impacts on export ratios, both of domestic and of foreign firms, and the effect is higher if the firm export to a developed country (Barrios et al., 2003).

Despite the large amount of research aimed at understanding the relationship between innovation and exporting, few studies have hypothesized a positive interdependence between exports and innovation.

Complementarity of innovation and export

Interactions between innovation and export strategies have always been very interesting; however, studies treated the two strategies as isolated from each other, examining the effects that each has on productivity as if it were independent from those of the other. Few studies shed some new lights on the potential interactions between innovation and export. In particular, the two strategies are revealed to be complementary and reinforcing each other. The marginal contribution to SMEs' sales growth of innovation is higher if export is also in place, and vice versa. While previous research demonstrated that the two activities contribute to firms' growth when undertaken separately, Golovko and Valentini (2011) unveiled the mechanism that may underlie the reinforcing cycle between export and innovation and provide empirical evidence in support for the complementarity. Hence, performance improvement do not only come from the optimization of one activity, but managers should also consider interactions between them since those contribute the most to firms' performance improvement. On the same page Ito and Lechevalier (2010), which found that decisions regarding innovation and export strongly affect performance and survival of firms; the authors confirmed the complementarity between the strategies and highlighted that the best performers in Japanese manufacturing firms were those that exported and innovated.

While a part of the literature has investigated learning by exporting, mainly concluding in favor of it, few contributions have tried to comprehend how the phenomenon takes place.

Mechanism of learning by exporting and research hypotheses

Since exporting is increasingly being recognized as an important source of learning, further steps consist in understanding how the learning takes place. First thoughts occurred to researchers' minds are that, through informal communication with foreign purchasers, a large amount of knowledge flows from the counterpart to the firm. Therefore, knowledge will be incorporated by the firm which would use it to modify products, improve production processes or apply for new patent registration (Salomon & Shaver, 2005a). Although this appeared to be quite likely, attempts in showing it empirically failed. Indeed, MacGarvie (2006), using patent citations as a measure of knowledge diffusion, tried to determine whether exporting increases the likelihood of firms' citations to foreign patents as well as foreign citations to the firms' patents. Surprisingly, the author found that, although exporters gain access to foreign technology and knowledge by communicating with foreign buyers, knowledge diffusion through these channels is not well measured by patent citations, and exporting is not statistically significant. Conversely, importing firms' patents are significantly more likely to be influenced by technology in the exporting country than are the patents of the firms that do not import from that country. Therefore the question regarding how learning by exporting takes place is still unanswered.

Finally, literature agrees that exporters are better compared to non-exporters; on average, they produce more than twice the output of non-exporters, are more productive and pay higher wages (Bernard & Jensen, 1999), partly due to self-selection and partly to learning by exporting. But how does the learning by exporting work? What are the mechanisms through which exporters incorporate new knowledge and new technologies? Aw and Batra (1998) stated that export activity is an important tool through which new or improved technology is passed from foreign purchaser. MacGarvie (2006) has then unsuccessfully tried to verify whether informal communication with foreign buyers allowed the transfer of knowledge and technology from abroad. Therefore, in literature there still is a void regarding the process through which export exerts the positive influence on innovation.

The present study is aimed at filling this existing void. If informal communication is not a channel through which technology and knowledge are transferred from foreign counterparts, it can be hypothesized that formal technology transfer agreements play the important role of channel through which new knowledge and technologies are incorporated by the firm. In that sense, if a firm stipulates a formal technology agreement in a given year, buying license, technical assistance and know-how from abroad, the innovative performance of that firm should improve in the following periods. Hence, the aim of this paper is to corroborate the following hypotheses:

Hypothesis 1: A formal technology transfer agreement in a given period has positive effects on the number of product innovations in the following periods.

Hypothesis 2: A formal technology transfer agreement in a given period has positive effects on the number patents registered in the following periods.

As seen above, several variables are demonstrated to influence innovative productivity. For this reason, other elements are included in the models, in order to control for the positive effects they might have on innovation. In particular, the export and import status, the participation of foreign shareholders in firms' equity, investments in R&D and size will serve as control variables.


The data employed in this study are from a yearly survey conducted by Fundación Empresa Publica with the financial support of the Spanish Ministry of Industry. The Fundación surveys a sample of Spanish manufacturing firms in order to generate an overall picture of the country's manufacturing sector. The dataset is a balanced panel of 10,205 firm-year observations spread on 13 years, from 1990 to 2002. The sample contains 6.3% observations of firms with no more than 10 employees, 62.4% with the number of employees included between 10 and 200 and 31.3% with at least 200 employees. In all, the dataset is composed by 785 firms, spread in 20 sectors; in detail, Table 1 presents the industry breakdown of the sample used.

Table 2 displays the percentage of firm-year observations in each sector that undertook the interesting activities during the sample period. Stipulation of technology transfer agreements is not very common; in fact, on average, only 11.4% of the firms stipulated technology agreements, although in sectors such as chemicals, motor vehicles and other transport equipment, the share exceeds 25%. Clearly, sectors with the lowest share are those where technology is not a key element (e.g. meat and wood).

Table 3 provides a snapshot of the distribution of firm-year observations across years, highlighting the percentage of firms that transferred technology and the status as exporter, innovator and importer. Both export and import activities are undertaken by more than 60% of the observations and show a growing trend across years, confirming the opening of the Spanish economy to the international markets.

As regards innovation, on average, less than 30% of the observations obtained product innovations and less than 10% registered a patent. As stated above, not many firms stipulated technology transfer agreements and, across the years, the percentage of firms has never exceeded 12.2%.

Dependent variables

As previously stated, this study focuses on understanding the impacts that formal technology transfer agreements have on firms' innovation performances. To disentangle this relationship, two dependent variables are used: product innovation counts and patent registration counts. Using two different variables may be helpful in having a more comprehensive picture of the situation as each variable might pick up different aspects of innovation (Salomon & Shaver, 2005a).

Product Innovation counts

The most direct way to evaluate innovation performance is to directly measure it. Data provided by the Fundación Empresa Publica include not only information on innovation inputs (i.e. R&D investments), but also on its outputs. Indeed, the dataset reports a count variable of the number of product innovations realized in a given year. The maximum value that this variable takes is 950, extremely high compared to the mean of 3.15. Albeit this number may seem disproportionate compared to the rest of the dataset, the firm that reported this number has, over time, very comparable numbers to this value [1] . The count variable representing product innovations is labeled "Product Innovations".

Patent Registration counts

Another way to assess innovation performances is through patent registrations. A part of the literature considers patent applications as a measure of innovation, claiming that if a firm decides to face the costly patent application process, it probably believes that it has innovated and it is worthwhile to pursue patent protection (Salomon & Jin, 2008, 2010; Salomon & Shaver, 2005a, b). On the contrary, since the dataset provides information on the number of patent registered in a period, and the variable certainly represents an output of innovative productivity, this study uses a count variable of patent registrations. This variable is labeled "Patent Registrations".

Independent variables

In order to evaluate the veracity of the hypotheses, a number of variables are part of the model. Specifically, apart from the variable of interest representing technology agreements, other variables are also used to control for elements that affect innovation.

Technology agreements

The survey collected information on whether a firm stipulated formal technology transfer agreements, or not, during a year. Technology transfer is the process by which a technology, expertise know-how or facilities developed by one individual, enterprise or organization is transferred to another individual, enterprise or organization (WIPO, p.1). In particular, this study is focused on the agreements that formally transfer technology to firms. Depending on the nature of the technology, different agreements might be used to transfer technology, such as the sale or the assignment of IP rights, a license contract, the purchase of technical assistance, etc. The Fundación Empresa Publica provides data on the purchase of technical assistance and licenses from abroad made by the firms. This is represented by a dummy variable that takes the value 1 if the firm bought technical assistance or license in a given period and, 0 otherwise. This variable is labeled "Technology Agreements".

Since knowledge embodied in technology agreements may take time to be incorporated by the firm and to result in an innovation or a patent, a 1-year lagged variable of the technology agreement status is created.

Control variables

Export is demonstrated to positively influence firms' innovation performances (e.g. Alvarez & Robertson, 2004; Salomon & Jin, 2008; Damijan et al., 2010); therefore, it is important to control whether a firm exports or does not. For this purpose, a dummy variable concerning the export status is used and labeled "Export Status"; "Export Status" takes the value of 1 if exporting activities are reported in the given period, 0 otherwise. Since Salomon and Shaver (2005a) demonstrated that the effects of export on innovation are more pronounced with lags, a lagged variable of export is also used.

Similarly, maintaining an importing relationships with foreign counterparts, may increase knowledge flows and, thus, have positive impacts on innovation. In this sense, the use of a dummy variable, regarding the import status of firms in the given period, helps in controlling for import influences on innovation. This variable is labeled "Import Status".

The presence of foreign shares in firms' equity is positively associated with an increase of the probability of innovation (Alvarez & Robertson, 2004; Basile, 2001); therefore, to control for the presence of the foreign equity, a variable expressed as a percentage of the foreign capital is included in the model. The variable is labeled "Foreign Capital".

Size is considered a factor that positively influences innovative productivity (Ebling & Janz, 1999). Since innovators are generally larger than non-innovators, the natural logarithm of the number of employees is used as a control variable. The use of number of employees, instead of sales, to control size is due to the positive effect innovation might have on sales: namely, a firm that produces many innovations in a given year is also expected to increase sales. Therefore, number of employees is used. This variable is labeled "Size".

Finally, large part of the literature also considers positive effects of R&D inputs on innovation (e.g. Salomon & Shaver, 2005b). To control for these effects, an R&D intensity variable is included in the model. "R&D Intensity" is defined as R&D investments normalized on total sales, expressed as a percentage. Since generally, R&D expenditures are highly correlated with size, the normalization of R&D on sales allows controlling for effects related to size.

Descriptive statistics and correlations

Table 4 and Table 5 present descriptive statistics and correlation matrix for the variables used in the model. First of all, it is important to note that firms that stipulate formal technology agreements are more productive in terms of innovation. In fact, for those firms, both the numbers of product innovations and patent registrations, on average, are higher compared to other firms'. Moreover, examining Table 4, it is clear how, on average, those firms are more oriented to export and import activities, have a larger share of foreign capital, invest more in R&D activities and have more employees.

In figure 1 and figure 2, observations are divided based on whether they stipulated technology agreements or not, in order to compare innovative performances of the groups. On Y axes of the graphs, the average number of product innovations and patent registrations for each year is shown. Firms that stipulated technology transfer agreements constantly outperformed the other group, both in terms of product innovations and patent registrations; however, it is necessary to highlight that, albeit interesting, these graphs do not say much about the direction of the causality of the relationship.

Table 5 displays the correlation matrix for the full sample. Although most of the

correlations are as expected, some deserve particular attention. First of all, the intensity of the correlation between technology agreements and patent registrations is higher than that between technology agreements and product innovations. Secondly, and not surprisingly, innovation and exporting are positively correlated as well as export status with size, R&D expenditures and foreign capital. Furthermore,

the variable representing technology agreements is strongly positively correlated with its 1-year lag, thus implying that if a firm formally transferred technology from abroad in a period, it is likely that it will also do it in the following year; finally, it is worth to mention the negative correlation between foreign capital and product innovation,

probably due to tendency of international firms to develop innovation in the headquarter instead of in subsidiaries. Although interesting, Table 5 results do not say much about the causality of the correlations; therefore, it is necessary to turn to multivariate analysis to better understand the nature of the relationships and demonstrate the hypotheses.

Statistical approach

The approach taken to assess technology agreements' effects on innovation is to regress the measures of innovation on the lagged values of technology agreements. The dependent variables used are both count variables (number of product innovations and number of patent registrations) that can only take nonnegative integer values. Furthermore, many observations are close or equal to 0. In these cases, when data are strongly skewed to the right, the OLS regressions are not appropriate. Therefore, a Poisson estimation procedure is considered; however, in Poisson distribution the mean and the variance are the same, while the distributions of the dependent variables display signs of overdispersion, with standard deviations much greater than means. In case of overdispersion, a Negative Binomial model is more appropriate. Hence, at first, estimation with random effects are run; then, to overcome the problem of the uncorrelation of the unobserved effects with the regressors, fixed effects estimations are also used.

Subsequently, to confirm the results of the Negative Binomial regressions, the dependent variables are transformed in their natural logarithms, in order to perform other estimations. Again, both random and fixed effects models will be used.


Table 6 and 7 present the estimation results of the Negative Binomial regression model. In particular in (a) columns the results of random effects specifications are showed while (b) columns display fixed effects'. All the specifications include year and sector dummies; yet, results are not presented in order to focus on the variables of interest.

Table 6 presents the results from regressing the Product Innovation count variable on Technology Agreements. In columns 1, the model also includes 1-year lagged values both of the variable of interest and of the Export Status. Technology agreements coefficients did not result statistically significant either in column 1(a), or in column 1(b). It is interesting to note that, in the second specification, also export coefficients resulted not significant, although the conviction of a positive effect of export on innovation is well-established in the literature. These unexpected results might be explained either by the low variance of the variables, which is absorbed by fixed effects, or by the multicollinearity between the export variable and its lagged value [2] . In columns 2, the variable of interest and Export Status are only considered in the immediately previous year. Again, the coefficient estimates of Technology Agreements in these specifications are positive but statistically not significant. On the contrary, and as expected, Export Status' coefficients resulted significantly positive in both random and fixed effects models, at the 1% and the 5% significance levels, respectively. In columns 3, the variable of interest is replaced by its 1-year lag, to verify whether, increasing the time interval from the stipulation of the agreement to the product innovation, the variable of interest gains statistical relevance in influencing the number of product innovations. Although this seemed reasonable, because larger amount of time might be needed to incorporate the technology and transform it into a product innovation, the coefficient of the variable is still statistically not different from zero. As regards exporting instead, the 1-year lag of the Export Status variable resulted positively significant in both specifications. To sum up, estimations in Table 6 suggest that the purchase of foreign license and technical assistance from abroad does not influence the number of product innovations that firms obtain, neither in the following year, nor after two years. In line with prior literature, the control variables significantly positively influence innovation performance, except for the foreign capital participation whose coefficient is not significant in any of the specifications. In particular, R&D investments coefficients are positive and statistically significant at the 1% significance level confirming that greater investments in R&D boost innovative productivity. Similarly, size of firms has positive impacts on product innovations.

In Table 7 a different measure of innovation is used: Patents Registrations. Table 7 displays the results of the regression of Patent Registrations on Technology Agreements. As for the previous regression, the coefficients of the variable of interest in columns 1 are not statistically significant in any of the year considered. Thus, when considering both the years before the registration of the patents, formal technology transfer agreements do not influence patent registrations. On the contrary, being an exporter significantly positively influences the number of patents registered, consistently with literature's view. In columns 2, the stipulation of technology agreements is considered solely in the year immediately preceding patent registration. In column 2(a), a random effects regression is performed, estimating a positive and statistically significant coefficient for Technology Agreements at the 10% significance level. To confirm this interesting finding, a fixed effects estimation is run. The coefficient estimate remains positive and significant, thus implying that the formal purchase of technology from abroad directly positively influences the registration of patents in the following period. Moving to columns 3, the variable of interest is substituted by the its 1-year lag, in order to understand whether, considering a longer time interval, technology agreements are still significant in influencing the number of patents registered. In this case, the coefficient of Technology Agreements did not result significant either in the random effects, or in the fixed effects models.

Findings regarding exports are consistent with those of prior studies. Indeed, Export Status in the previous year resulted positive and significant in every specification, confirming the positive influence of exports on patent registrations (Salomon & Jin, 2008). Moreover, the coefficients of random effects specifications resulted larger both in magnitude and significance. Columns 3, solely includes a 1-year lag value of Export Status: while the variable resulted significantly positive in the random effects specification, in the fixed effects' the coefficient is not statistically significant. Similarly, Import Status coefficients appeared positive and significant only in the random effects columns, while in fixed effects, albeit positive, they are not significant. These findings might be explained by the low variance of the variables within the groups, since it can be assumed that exporters and importers tend to maintain the status across years. With respect to the control variables, Size and R&D investments positively influence the registration of patents in every specification at 1% significance level, consistently with large part of the literature. Finally, a brief focus on foreign capital coefficients is needed: according to the estimates, the participation of foreign partners negatively influences the number of patent registered, although the magnitude of the coefficient is close to zero. As anticipated above, this may result from the international organizations' inclination to develop innovation in the headquarters and not in the subsidiaries, thus registering the patents in the country where the headquarter is located.

Altogether, the results presented in Table 6 and Table 7 provide evidence that the formal purchase of technology from abroad in a given year does not have any positive influence on the product innovations of the following year; therefore, there is not statistical evidence in support of the first hypothesis. Conversely, the registration of patents is positively influenced by formal purchase of technology, thus confirming the second hypothesis.

To corroborate the findings of the Negative Binomial estimation model, the dependent variables are transformed into their natural logarithms, in order to perform regressions to verify the same hypotheses. Table 8 and 9 show the results of the estimations. Again, year and sector effects are included in the random effects specifications while, in the fixed effects model, sector effects are excluded because their values do not change during the sample period, thus generating multicollinearity and being omitted by the fixed effects.

The results of the regression of the natural logarithm of product innovation counts on Technology Agreements appear in Table 8. The estimates confirm previous results: the coefficients of the variable of main interest are not statistically significant in any of the specifications, thus corroborating that technology agreements do not influence product innovation. As before, after analyzing impacts of Technology Agreements on Product Innovations, the focus shifts on the other dependent variable: Patent Registrations.

Table 9 shows the results of the estimation of Technology Agreements coefficient considering the natural logarithm of the Patent Registration as dependent variable. Columns 1 allow for technology agreements stipulated one and two years before the registration of the patent. While the coefficients of the lagged variable are not significant in columns 1(a) and 1(b), those of the Technology Agreements are positively significant at 1% and 5% significance level, respectively. In particular, if a firm stipulates a formal technology transfer agreement in a given year, the number of

registered patents in the following year can increase up to 6.8%. This is also confirmed by the following columns, where the coefficients of the dependent variable remain significant and positive also after dropping the lagged variable. In columns 3, a 1-year lagged variable replaces the technology agreements: while in fixed effects model the coefficient of the variable of interest is not significant, in column 3(a) a positive and significant relationship between this variable and patent registrations is found at the 5% significance level. Therefore, patent registration is positively affected by technology agreements not only in the previous year, but also if the firm stipulates the agreement two years before the registration of the patent.

To sum up, a formal technology transfer agreement with a foreign counterpart does not improve innovative productivity of firms in terms of product innovations; this is confirmed by every analyses made, thus implying that the first hypothesis of the study is not verified. On the contrary, if the measure of innovation is changed into the number of patent registrations, the purchase of technical assistance or license from abroad positively and significantly impacts on innovative productivity, hence validating the second hypothesis of this study.

Additional analyses

Several additional analyses are conducted to evaluate the veracity of the findings. So far, a relationship between technology agreements and the registration of patents has been confirmed; however, the direction of the relationship is still not entirely clear. Technology agreements might positively influence the registration of patents but also, previous patents may have an impact on the stipulation of formal technology agreements. Therefore, in order to ensure that the presumed direction of the relationship is the correct one, a dynamic Arellano Bond estimator is used. In this estimation, lagged values of the dependent variables are included as regressors. Including dependent variables allows accounting for the potential serial correlation of errors, as well as for the dynamic component in the firm characteristics (Golovko & Valentini, 2011).

In particular, two estimations are made. In the first case, the dependent variable is represented by "Technology Agreements" in a given period, while the independent variable, besides the control variables, is the count variable of "Patent Registrations" in the previous period. As the estimation allows controlling for dynamic component of firm characteristics, the lagged value of technology agreements is also included in the specification.

The second estimation uses "Patent Registrations" number as a dependent variable, including its lagged value as covariate, and above all, "Technology Agreements" in the previous period as regressor. As stated above, the aim of these estimations is to clarify the direction of the relationship between technology agreements and registrations of patents. The direction of the relationship will be from "Technology Agreements" to "Patent Registrations", if the first variable is a statistically significant regressor when the second is the dependent variable, but not vice-versa. Table 10 presents the results of the estimations. In the first column "Technology Agreements" is the dependent variable. While the lagged value of the dependent variable resulted positively significant, confirming the univariate analysis results, the estimate of the registration of patents resulted not significant. On the contrary, the results displayed in the second column validate the second hypothesis of the study: technology agreements positively influence the registration of patents in the following year. Among the estimates of the control variable, the most interesting are those regarding "Export Status" and "Import Status". Indeed, exporting to foreign countries does not influence the number of patent registered, while, engaging in importing activities positively and significantly impacts on the number of patents registered.

To assess the relevance of the models, Arellano Bond tests for first and second order serial autocorrelation of first-differenced errors are implemented for both estimations. These tests verify whether errors are independently and identically distributed. If that is the case, the first differenced errors are first-order serially correlated; hence the hypothesis of zero autocorrelation in first-differenced errors in the test can be rejected. On the contrary, serial correlation in the first-differenced errors at an order higher than 1 implies that the moment conditions used by the model are not valid. As expected, the Arellano and Bond test for first-order serial autocorrelation for the first and second estimations are found to be negative and significant at the 1% and 5% level respectively, while the tests for second-order autocorrelation are rejected. There is no second-order correlation, which confirms the consistency of the estimators. Therefore, it can be concluded that the conditions are valid.

In nearly all previous estimations, "Patent Registrations" were not affected by "Technology Agreements" stipulated two years before the registration of the patent. At the same time, the variable "Technology Agreements" is positively influenced by its 1-year lagged value. Since it can be assumed that this relationship is valid also for previous periods, 2, 3 and 4-years lagged values of "Technology Agreements" can be considered instrumental variables as they influence the independent variable but not the errors, thus the dependent variable. To control for this dynamic relationship, a third Arellano Bond estimation, whose results are shown in Table 11, is made; as evident from Table 11, also controlling for the dynamic relationship, the results of the second column of Table 10 are confirmed. Indeed, "Technology Agreements" coefficient is still positive and significant at the 1% significance level, albeit lower in magnitude than previous specification. Again, to assess the relevance of the specification, Arellano Bond tests for first and second order serial autocorrelation of first-differenced errors are implemented, confirming the validity of the model.

To confirm the results of the previous estimations, namely that technology agreements only influence registration of patents, a logit model is used. Indeed, the aim is to demonstrate that the stipulation of technology agreements in a given year increases the odds of registering patents in the following year, but it does not affect product innovations. For this purpose, a dummy variable concerning the achievement of at least a product innovation is used and labeled "Product Innovator"; this variable takes the value of 1 if product innovations are obtained in the given period and, 0 otherwise. Similarly, a dummy variable regarding the registration of at least one patent is used and labeled "Patent Registration". The model includes "Technology Agreements" as independent variable together with the control variables previously used. The results of the estimations are presented in Table 11. Only random effects estimations are run including year and sector effects in the specifications. The first column presents the estimation where "Product Innovator" serves as dependent variable. In line with previous results, the coefficient of "Technology Agreements" did not result statistically significant. To the contrary, the independent variable's coefficient is positive and significant when the dependent variable is "Patent Registration". In particular, if the firm has stipulated a technology transfer agreement in the previous period, the log-odds of patent registration increases by 0.372. In other words, the odds of registering a patent if the firm has stipulated technology agreements are 1.45 [3] times higher than if the firm has not. Furthermore, "Export Status", "Import Status", "R&D Intensity" and "Size" significantly positively affect the likelihood of registering a patent, while the effects on patent registration of the participation of foreign shareholders to firms' equity are negative, albeit low in magnitude.

Discussion and conclusion

The aim of this study has been to delve deeper into the matter of the relationship between innovation and export. Even though a great a part of researchers agrees on the positive influence innovation has on export, and vice-versa, there has been little evidence on the mechanism underpinning learning by exporting. Albeit scholars have investigated the phenomenon, claiming that firms gain knowledge through informal relationship with foreign counterparts, the evidence in support of this hypothesis is rather scarce (MacGarvie, 2006). In this research it has been argued that formal agreements regarding the purchase of technology, license and technical assistance positively influence innovative productivity of firms. Therefore, the stipulation of a formal agreement serves as channel through which firms gain knowledge. Using product innovations and patent registrations as measures of innovative productivity, the impacts that technology agreements have on innovation have been explored. The empirical approach is to regress the number of product innovations and patent registrations on technology agreements using a unique balanced panel of Spanish manufacturing firms provided by the Fundación Empresa Publica for the period 1990-2002.

The empirically analysis provided evidence partially consistent with the hypotheses. In fact, it has been found that firms do not increase their product innovations if they have formally transferred technology neither in the previous year, nor with further lag. Explanations. On the contrary, patent registrations appeared to be positively influenced by technology agreements in the previous year. In addition, the stipulation of a technology agreement also has positive impacts if considered with further lag, although in this case evidence is not as strong as for 1-year lag. Explanations.

Furthermore, this study gives also evidence in support of learning by exporting. Unlike Higón and Driffield (2011), which found that exporting does not influence product innovation, this research's results confirm positive impacts of exports on innovation. Indeed, being an exporter increases both the number of product innovations and patent registrations, also with two-year lags. Clearly, entering in a foreign market increase the likelihood of being in contact with improved knowledge, methods, technologies, thus positively influencing innovation. Moreover, in line with Alvarez (2007), import resulted greatly important for innovation. In fact, both the measures of innovation considered turned out to be positively affected if the firm imported. MacGarvie (2006) argues that, after beginning to import, firms register an increase in the knowledge available, and, assuming that this will result in an increase of innovation, these findings are consistent with this study's. Finally, it is interesting to underline the role played by the presence of foreign shareholders in firms' equity. According to this research estimation, if a firm is a subsidiary of an international company, the likelihood that it will register a patent decreases. These findings might be explained by the fact that international companies tend to developed knowledge, and thus register patents, in the country where the headquarter is located. Therefore, subsidiaries are inclined to play a secondary role in the matter of innovation.

Although additional analyses are performed to confirm these findings, they should be carefully interpreted because of some caveats that may warrant attention. First of all, the empirical analysis relies on data regarding solely a country, Spain. Thus, it does not take into consideration external factors, such as institutional, economic and financial, contexts which can severely impact on firms' performances, especially in terms of innovation. Future developments would consist in expanding the research in other contexts in order to assess whether the results might be applied to other countries.

Second, data regard only manufacturing firms; as seen, opinions regarding manufacturing sectors are much more unanimous than for other industries (e.g. Barrios et al., 2003; Lachenmaier & Woessmann, 2004; Nguyen et al., 2008). Therefore, it might be interesting to gauge whether this research's results also apply to other industries, such as service.

Third, this study uses solely dummy variables to describe both technology agreements and export. On the contrary, the use of measures of export intensity (such as export normalized on sales) and of number of and/or expenditures for technology agreements would give further details on the relationship. Indeed, they would help in having a more comprehensive picture on how much technology agreements affect innovation.

Fourth, future research would be well served to explore the nature of the agreement stipulated. As stated above, the variable of technology agreements regards both the purchase of a license and of technical assistance. It can be presumed that the positive effects these elements might have on innovation are quite different. Thus, there is room for further investigation on differences in magnitude, in terms of positive influences, that agreements might have.

Finally, the nature of the "transferor" is not taken into consideration. With the term nature, I mostly refer to two aspects: first of all, the transferor might be a multinational firm, a small medium enterprise, a research center, etc. Investigating the nature of the transferor might give new insights on strategic decisions of firms: as an example, firms might decide to purchase technology from a small firm focused on R&D, instead of developing R&D inside. The second aspect I refer to with the term nature is the country of origin of the transferor: several researches (e.g. Alvarez & Robertson, 2004; Trofimenko, 2008) showed additional benefits, in terms of innovation, when exporting to developed countries. Therefore, if firms export to developed countries, the likelihood that they increase their knowledge and innovation is higher compared to exporting to developing country. So it can be assumed that, if the firm engages in a relationship with a foreign counterparts coming from a developed country, the amount and the value of the knowledge incorporated in a technology agreement is higher, thus having more positive impacts on innovation.

In spite of the aforementioned caveats, this study offers a number of contributions to the literature. In the first place, albeit with limitations, it partially fills the void left by the literature regarding the mechanism that may underlie learning by exporting. Prior research has attempted to explain how firms learn from international trade, assuming that the informal communication with foreign buyers allows exporters to gain access to foreign technology and improved knowledge (MacGarvie, 2006). While results based on the above assumptions were quite unsatisfactory, this study proves that the formal purchase of technology from abroad permit the firms to increase their knowledge, thus improving the number of patents registered. Second, it provides further evidences for learning by exporting. As stated, export, consistently improves firms' innovative performances. Firms increase their patent registrations and product innovations after starting to export. Organizational learning - strategic management