Internationalisation, innovation and productivity: how do firms differ in Italy?

Transcription

1 Internationalisation, innovation and productivity: how do firms differ in Italy? Davide Castellani* and Antonello Zanfei Università di Urbino Carlo Bo June 2004 Preliminary Draft Abstract Consistently with some recent modelling in trade literature and with more consolidated views in the economics of international production, we show that, after controlling for sector, location, firm age and size, a substantial heterogeneity in economic and innovative performances characterizes Italian manufacturing firms with different involvement in foreign activities. In particular, non internationalised firms turn out as the worse performers, while multinational firms carrying out manufacturing activities abroad, are characterised by both the highest productivity premiums and the highest R&D efforts and innovative performances. Multinationals with a lower commitment to foreign markets, i.e. with non manufacturing activities abroad only, exhibit a higher productivity than exporters but they do not appear to innovate more than the latter. * Corresponding author: Davide Castellani Istituto di Scienze Economiche Università degli Studi di Urbino Carlo Bo Via Saffi, Urbino (PU) Tel. +39 (0) Tel. +39 (0) castellani@uniurb.it 1

2 1. Introduction Recent developments in trade theory have tackled the issue of firm heterogeneity and internationalisation modes. The key prediction emerging from this literature is that firms with different levels of productivity will generally engage in different modes of international activities, as these will be associated with distinct sunk costs (Helpman, Melitz and Yeaple 2004). This interpretation has clear, albeit not fully acknowledged, connections with a more consolidated view emerged in international production literature, which has stressed the role of ownership advantages in determining the choice to enter foreign markets. Firms must be endowed with some exclusive, proprietary asset, such as a superior technology, to be able to face foreign competitors and set up foreign activities (Hymer 1960, Dunning 1970, 1988, Cantwell 1989). More recent contributions have also emphasised that international activities might enable firms to gain access to foreign knowledge sources and further reinforce such advantages (see Narula and Zanfei 2004 for a review). This paper draws from complementary insights stemming from these different streams of literature and provides evidence on the relationship between firm heterogeneity and internationalisation modes, with specific reference to the Italian manufacturing industry. We introduce two main novelties in the empirical analysis of this issue. First, we extend the span of variables to capture intra-industry heterogeneity, by focusing on both productivity and measures of R&D and innovative behaviour. This per se marks a departure from most contributions which have either focused on the former or on the latter type of indicators. By combining both types of measures, we can also disentangle different aspects of firm performances and understand some of the interdependences between technological accumulation strategies, as captured by measures of innovative activities, and economic performances, as identified by productivity. The second novelty of this paper will be to analyse the relationship between (economic and innovative) performances and a further mode of internationalisation, that is the creation of non manufacturing activities abroad. That is a sort of intermediate category between pure exporters and the creation of foreign manufacturing affiliates. On the one hand, considering this further category will enable us to test the robustness of the theoretical predictions. On 2

3 the other hand, it will allow us to better highlight the relationship between internationalisation strategies and the different performance indicators. Section 2 will overview the theoretical and empirical background to this paper. Section 3 will illustrate the data we use, and section 4 will discuss the results of the empirical exercises we have carried out. Section 5 will conclude. 2. Previous research on firm heterogeneity and internationalisation Several streams of literature have considered the relationship between firm heterogeneity and internationalisation strategies. However the issue has recently got at centre stage with the increasing attention it has received in international trade studies. A paper by Helpman, Melitz and Yeaple (2004) marks an important departure in this field. In fact it combines the analysis of two issues that have had a separate treatment. The first issue concerns the choice between export and (horizontal) FDIs as a means to serve foreign markets. The trade literature has tackled this issue mostly in terms of the proximity-concentration trade-off: firms can be expected to invest abroad when the gains from avoiding transport and tariff costs outbalance the costs of maintaining capacity in multiple markets (Brainard 1993). More precisely, the prediction is that firms are more likely to expand production horizontally, rather than exporting, the higher are transport costs and trade barriers, and the lower are fixed costs of entry and the size of scale economies at the plant level, relative to the corporate level. Using data on US export and foreign affiliate sales for 63 industries in 27 countries, Brainard (1997) provides an empirical test for this model, and shows that the FDI/export ratio varies according to trade costs and fixed entry costs observed in each sector. However, as noted by Head and Ries (2003 p.2), her model does not predict which firms does which activity. Here comes the second issue at stake in the model by Helpman et al. (2004): firm diversity. A number of recent empirical studies have highlighted that more productive firms tend to self select into the export market 1. This increasing evidence on the relationship between export and firm performance has given rise to a set of theoretical trade models explicitly allowing for firm heterogeneity (Bernard, Eaton, 1 Among many others, Bernard and Jensen (1999) for US firms, Castellani (2002) for Italy and Clerides et al (1998) for Colombia, Mexico and Marocco. Tybout (2004) provides an extensive review of such evidence. 3

4 Jensen and Kortum, 2003; Melitz, 2004). Although evidence is not as extensive as in the case of exports, there are also several empirical works documenting that multinationals tend to outperform firms with no investment abroad as in the case of Doms and Jensen (1998) for the US, Barba Navaretti and Castellani (2004) for Italy, Criscuolo and Martin (2003) for the UK, De Backer and Sleuwaegen (2003) for Belgium, Pfaffermayer and Bellak (2002) for Austria. Helpman et al. (2004) are aware of some of this empirical literature and extend the original model by Melitz (2004) to account for this further heterogeneity. They assume that firms make their profit maximising choice, given their productivity level and considering the differential costs associated with the different modes of market access, some of which are sunk (entry costs) while others vary with sales (transport costs and tariffs). Relative to FDI, it is assumed that exporting involves lower sunk costs but higher per unit costs. This hierarchy of costs recalls the analysis of market access strategies proposed earlier by Horst (1971), Hirsch (1976) and Buckley and Casson (1981), who also considered the alternative route of licensing agreements. Given such cost schedules, however, the choice in this latter category of models depended primarily on industry conditions such as market size and growth, and the intensity of competitive threats. Firm level characteristics became relevant after the initial decision to enter foreign markets, as firms got more familiar with the market or acknowledge that entry costs had been sunk when taking further investment decisions. In the model by Helpman et al. (2004) firms self-select into foreign markets because they are endowed ex ante with differential productivity. This differential will enable them to extract positive profits from international activities in spite of the higher variable costs in the case of export, and of higher fixed costs in the case of FDIs, relative to production for the domestic market only. Firms with low levels of productivity can only afford to operate in domestic markets; whereas more productive firms are more profitable in all three activities, i.e. at producing for the domestic market as well as producing at home for exports or producing abroad to serve foreign markets 2. Rising trading costs will increase the critical productivity required to make exporting profitable; and will lower the critical productivity necessary to make FDI 2 This ranking of performances associated with different modes of internationalisation can be further reinforced by the processes of learning from foreign operations. In fact, a number of recent works have documented that exporting firms do not learn by their international operations (Clerides et. al, 1998; Bernard and Jensen, 1999), but some evidence supports the hypothesis that increasing degrees of learning are associated with increasing commitment to international operations (Kraay, 1999; Castellani, 2002a; 2002b; Barba Navaretti and Castellani, 2004). 4

5 preferable to exports. By the same token, higher fixed costs of operating foreign affiliates increase the level of productivity required to make FDI preferable to exporting. The model thus connects to a well established tradition in the economics of international production, which has emphasised that multinational firms must be endowed with some exclusive ownership advantage in order to be able to face competition in foreign markets. This idea goes way back to Hymer (1960) and to Dunning (1970) and has been incorporated into trade models of the multinational firm (Markusen, 2002). Two main weaknesses emerge when adopting the perspective proposed by Helpman et al. (2004) and related literature. First, the origins of productivity advantages are not explored. Productivity levels are assumed to be drawn casually from a probability distribution, and firms behaviour varies accordingly for any given level of trade costs and of fixed costs of operating abroad. This simplifying assumption is indeed most frequently adopted in recent trade models, as surveyed by Tybout (2003); and reflects a more consolidated tradition in other fields of (micro)economics, as shown in Gilbert (1989) and in Reinganum (1989). Making specific reference to this latter tradition, Nelson (1991 pp.64-65) observes that a generalised view tends to be that firms differ because of the luck of a draw, or an initial condition, which made different choices profitable given certain market circumstances. However, if the market circumstances were reversed so would be firm behaviours, even though some delay or inertia may be allowed for in the modelling. When these assumptions are made, there is not much sense attached to the word heterogeneity: There are firm differences but there is no essential autonomous quality to them (p.64). Taking firm heterogeneity and its causes into account more seriously would imply, in Nelson s view, a general reorientation in the analysis of economic behaviour, a subject which cannot be dealt with here. However, one should at least consider that one of the key sources of (productivity) advantages, and hence of heterogeneity, is firms accumulation of technological capabilities. Firms intentionally accumulate knowledge to increase their own competitiveness relative to their own main rivals in the final product market. By reducing unit costs relative to others in the same industry, firms with higher technological capabilities (and hence ownership advantages) may both increase their margins and reduce prices, thus expanding their 5

6 international market shares (Cantwell 1989, Cantwell and Sanna Randaccio 1993). By contrast, firms with the fewest or weakest ownership advantages in general hold their position more easily in domestic markets than in international markets owing to government support, consumer loyalty, the closeness of local business contact and so forth (Cantwell 2000 p.39). To summarise: technological accumulation and related advantages do affect firms positions in international markets. But this is only part of the story. The choice of internationalisation mode also interacts with the cycle of utilisation and generation of technological advantages. Given the partially tacit nature of knowledge developed by firms, its exploitation and use tends to require complex organisational devices and the creation of internal networks within firms, especially when technology is applied in different and (culturally and institutionally) distant contexts (Cantwell 1989, Kogut and Zander 1993, Vaccà and Zanfei 1995). Furthermore, a growing albeit less established literature has emphasised that internationalisation, and particularly manufacturing and R&D investments abroad, can be a fundamental means to gain access to external knowledge sources that are complementary to the specific assets a firm is already endowed with. The dynamic process of knowledge accumulation through internationalisation has been dubbed in the literature as asset seeking (Dunning 1993, Dunning and Narula 1995, Narula and Zanfei 2004) or home base augmenting (Kuemmerle 1999), or technology sourcing FDI (Driffield and Love 2002). This will generally require the combination of internal network development within multinational firms, with external networks involving partners active in the specific local markets in which foreign affiliates are active (Zanfei 2000). Overall, one can thus expect FDI and multinational expansion to reflect the need to utilise, and extract economic value from, the available knowledge, which implies some degree of retention of technology within the firm as well as some interactions with external users. Moreover, the international generation of knowledge, requires extensive R&D efforts to be carried out internally, but also contacts and collaborations with external parties possessing localised knowledge. Ex ante advantages, resulting from a firm s history of technological accumulation, will provide guidance for further research as well as abilities to absorb external complementary knowledge, wherever this may be available. This cumulative, interactive process is fuelled by, and contributes to, firm heterogeneity. As Cantwell (2000 p.29) posits it: By extending its own network, each firm extends the use of its 6

7 unique line of technological development; and by extending it into new environments, it increases the complexity of its development. The second source of weakness in Helpman et al. (2004) pertains to the domain of empirical analysis. Not many direct tests of the predictions stemming from this model are available yet, although some evidence is being produced somewhat consistently with it. The authors themselves supply what they consider a test of their own theory using data of US affiliate sales and US exports in 38 countries and 52 sectors. They do not only find evidence that sector/country specific transport costs and tariffs have a strong negative impact on export sales relative to sales of foreign affiliates, in line with earlier results obtained by Brainard (1997). They also find that more heterogeneity leads to significantly more FDI sales relative to export sales. The key issue here is that they measure heterogeneity in terms of firm size, according to a recurrent albeit criticisable practice in economics literature. The authors derive this way of measuring heterogeneity directly from their model, wherein the dispersion of firm size within a sector captures the joint effect of the dispersion of firm productivity and the elasticity of substitution, which magnifies the effect of productivity differences across firms (ibidem p.307). While identifying a relationship between size and productivity may be reasonable to simplify the analytical treatment in a theoretical model, empirical work has shown that this is far from true in real world. Firms may grow large for a variety of reasons that have nothing to do with technical efficiency, with technology or immaterial assets that translate into firm level advantages. And small firms in many sectors are more innovative than large ones (see Cohen and Levin (1989) for a review of the extensive and controversial literature on firm size and innovation). However, a growing number of studies have directly addressed the prediction of Helpman et al. (2004) that differences in productivity should correspond to different foreign market entry modes. Head and Ries (2003) use several indicators of performance to differentiate firms in a sample 1070 large Japanese companies classified in 17 two digit industries in These indicators range from sales to value added, approximate and estimated total factor productivity. They find that domestic firms, exporters and firms investing abroad exhibit the predicted hierarchy in performance levels (from the lowest to the highest performance) only when using sales data, while much weaker support is found to the theory when using measures of productivity. Furthermore, consistently with the criticism recalled above, Head and Ries (2003) find only a weak correlation between 7

8 firm size and productivity. However, they find that among overseas investors, more productive firms span a wider range of host-country income levels, reflecting higher capabilities to penetrate different markets. Girma, Gorg and Strobl (2004) compare sales, productivity and profitability of purely domestic firms, domestic exporters and domestic multinationals with specific reference to Ireland for the year They utilise a non parametric approach based on the principle of first order stochastic dominance, and find that the distribution for multinationals dominate that of other domestic firms, while they do not find clear differences in plant performance between domestic exporters and non-exporters. Girma, Kneller and Pisu (2003) apply the same non-parametric technique to UK data and find consistent evidence that the productivity distribution of multinational firms dominates that of exporting firms, which in turn dominates that of non-exporters. Apart from a more general criticism that would apply to any measure of firm level productivity when data on physical quantities (of output, capital and intermediate inputs) are not available, it seems crucial to extend the analysis to other more direct indicators of differentials in terms of technology and innovation. In line with the emphasis we have placed earlier on the key differentiating role of technological accumulation and of its interactions with internationalisation strategies, it is worth testing this relationship more explicitly. A few works do explore the innovation-internationalisation linkages at the firm level. However, most of these studies focus on how innovation is associated with individual modes of internationalisation. A number of empirical contributions have assessed the impact of different innovation measures and export propensity, as in the case of Wakelin (1998) for the UK, Sterlacchini (1999) and Basile (2001) for Italy. A lower number of works have provided evidence on the relationship between innovation and other modes of foreign market entry. Fors and Svensson (2002) examine the impact of R&D investments on foreign sales of Swedish multinationals. Frenz and Ietto (2004) focus on the relationship between the degree of multinationality of UK firms and their innovative behaviour. Basile, Giunta and Nugent (2003) build an indicator of foreign expansion of Italian firms and find that firms innovative activies are important determinants of the degree of involvement in international operations. Criscuolo, Haskel and Slaughter (2004) use a dataset similar to the one used in the present work and test for the role of global engagement in UK firms innovation, controlling for knowledge inputs. Their results are consistent with 8

9 the idea that the higher propensity to innovate in multinational firms can be attributed to a better access to internal and external stock of knowledge. However, to the best of our knowledge, there is no study analysing how differences in both productivity and innovation are associated with alternative modes of entry in foreign markets. 3. Data The empirical analysis presented in this paper is based on a dataset resulting from the intersection of two different sources: the Second Community Innovation Survey (CIS II) and ELIOS (European Linkages and Ownership Structure). The former is a survey based on a common questionnaire administered by Eurostat to firms from all European countries, which aims at assessing various aspects of firms innovative behaviour and performances. Subject to a confidentiality agreement, we were allowed to access micro data for Italy from the survey carried out in 1997 and covering innovation occurring in Innovation data were complemented with ownership, multinationality and economic performance data from ELIOS dataset developed by the University of Urbino, Italy, which combines information from Dun & Bradstreet s Who Owns Whom and Bureau Van Dijck s Amadeus. The sample resulting from this matching comprises 778 domestic-owned manufacturing firms 4. Exploiting information on exports, available from the CIS, and information on subsidiaries controlled abroad, available from Elios, we broke down the sample distinguishing between firms serving only the domestic market (DOM MKT), firms serving foreign markets only though exports (EXP) and two types of multinational firms, namely those that control manufacturing subsidiaries abroad (MNC MAN) and those which control only non manufacturing (mainly sales) subsidiaries in foreign markets (MNC NON-MAN). As illustrated in Table 1, approximately one half of the sample firms serve the foreign markets only through exports, 21% control non manufacturing activities abroad and 16% are multinational firms controlling foreign manufacturing activities. A relatively small proportion of the sample is not 3 We thank Giulio Perani from the Italian National Statistical Office for allowing us access the data. 4 The overall sample resulting from the intersection includes 1,114 firms, but for the purpose of this study, we excluded from the analysis foreign-owned firms. In a companion paper (Castellani and Zanfei, 2004) we assess differences in productivity and innovative behaviour of foreign and domestic owned multinationals. Balcet and Evangelista (2003) utilize part of the same dataset to characterize innovative patterns of foreign-owned firms in Italy. 9

10 internationalized 5. It is rather clear from the large number of average employees that our sample is biased against small firms and is very likely to under represent non internationalised firms. This bias derives from the fact that Who Owns Whom (from which we draw information on ownership and multinationality) has a large coverage of firms belonging to groups of 2 or more firms, while it does not cover small independent firms. However, the sectoral distribution of the sample firms (reported in Table A.1 in Appendix) turns out to be not significantly different from the Eurostat universe of firms over 50 employees, and the coverage of multinational firms appears to be very accurate. 4. Results Following recent models in international trade theory as well as previous developments in economic literature, in the previous sections we have argued that firms involved in exporting activity can be expected to outperform purely domestic firms, while multinational firms should reach higher performance than the former. In this section we discuss some evidence consistent with the heterogeneity of firms engaged in different internationalisation modes. In Table 1 we report the means in size, productivity and innovation of firms serving only the domestic market (DOM MKT), firms serving foreign markets though exports (EXP) and multinational firms, distinguishing firms controlling only non manufacturing activities (MN NON MAN), and firm engaged in international production (MN MAN). Such basic statistics support the idea that increasing commitment to international activities are associated with larger firm size, productivity 6, propensity to innovate and the intensity in intra muros R&D expenditures. Further support to this hypothesis can be drawn from Figures 1 and 2 where the cumulative distributions of TFP and intra-muros R&D 5 It is worth mentioning that EXP are firms internationalized only though exports and MNC NON MAN do not control any foreign manufacturing activities. However, one should take into account that the latter group includes 156 out of 164 firms which are engaged also in exporting, while out of 121 firms carrying out international production (MNC MAN) 117 are also exporting and 79 are engaged in other non manufacturing activities abroad. In other words, moving along the continuum from DOM MKT to MNC MAN there seems to be an increase in commitment to foreign markets, with firms controlling foreign subsidiaries which, to a large extent, are also engaged in exporting activities. 6 We used three measures of productivity. First, we calculated labour productivity as value added per worker. Second, we obtained an approximate TFP, calculated as value added per worker net of the contribution of capital (per worker), with an approximated elasticity of.3 (following Griliches (1998) and more recent applications by Head and Ries (2003)). Third, we estimated TFP as the residual from a gross output production function, controlling for capital labour and materials, using a AR(1) random effect model, with sector specific factor elasticities. 10

11 intensity for the four groups of firms are compared 7. In particular, productivity in multinational firms with foreign production dominates that of all the other groups of firms for almost any value of the cumulative distribution; MN NON MAN dominate EXP; and the productivity of firms serving only the domestic market lies at the right hand side of that of the other groups for at least half of the distribution. While the ordering in the distribution of internationalised firms closely matches the prediction of Helpman et al. (2004), the productivity of DOM MKT is not constantly lower than in all other cases, as we would rather expect. However, we believe that this is due to the fact that our sample is biased in favour of large firms. As purely domestic firms are usually smaller in size, this bias is likely to reduce the dampening effect that small firms tend to have on the productivity of this group of firms. If we could correct for this bias, the whole distribution of DOM MKT would most probably shift to the left, becoming stochastically dominated by the other groups of firms. A similar pattern can be observed for the distribution of intra muros R&D intensity, with the expected ordering of curves from multinationals to uni-national firms. Evidence from the other continuous measures, reported in Appendix (Figure A.1), broadly confirms these results. However, the differences we have illustrated so far might well reflect a number firms characteristics, which affect productivity, innovation and internationalisation, such as the size and age of the firm, the sectoral and geographical composition. In order to take these sources of heterogeneity into account, we rely on parametric analysis. We ran regressions of various measures of productivity and innovative behaviour on three dummies identifying DOM MKT, MN MAN and MN NON MAN, using the largest group (EXP) as baseline category and controlling for other exogenous firms characteristics, such as sector and geographic location dummies, age of the firm and dummies for the size class. Coefficients on the three dummies reflect percentage differences in productivity and innovative behaviour of the various groups relative to exporters. Results of regressions reported in Table 2 confirm that both economic and innovative performances vary according to firms engagements in international markets, with a few important qualifications. First, OLS estimates for all measures of 7 Here we rely on visual comparison of cumulative distributions, without a formal test of stochastic dominance, as in Girma, Gorg and Strobl (2004) and Girma, Kneller and Pisu (2003). Further on in the text we shall consider differences in means, controlling for various firms exogenous characteristics in a regression framework. 11

12 productivity suggest that multinationals with manufacturing activities abroad perform better than multinationals with non manufacturing activities; and the latter outperform both exporters and purely domestic firms 8. No difference in productivity emerge between exporters and purely domestic firms in our sample, but this might reflect the bias in favour of large companies we have already mentioned earlier. Second, multinationals controlling manufacturing activities abroad exhibit the highest intramuros R&D efforts and innovative output; while multinationals with non manufacturing subsidiaries abroad do not exhibit different innovative activities than mere exporters. Non internationalised firms appear to invest the least in R&D and to innovate less than exporters. It thus appears that a higher engagement in international activities is associated with greater R&D and innovativeness. It is worth mentioning that the above discussed results should not be interpreted as causal effects of the various internationalisation modes on productivity and innovation. In fact, unobserved factors might well affect both the internationalisation choice and the dependent variable, and our data do not allow to disentangle self-selection effects from the ex-post process of learning and technological accumulation. A careful causal analysis is problematic with our data, due to the lack of good instruments and of a time series in most indicators. Bearing in mind these limitations, in the first two columns of table 3 we test whether the observed heterogeneity in terms of productivity is robust to controls for innovation outputs (captured by two dummies indicating the introduction of process and product innovation) and inputs (such as intra and extra muros R&D intensity and two dummy variables indicating technological cooperation with Italian and foreign counterparts). Results of this test confirm the ranking in TFP of multinational firms, exporters and non internationalised firms. The third column of table 3 also confirms that the heterogeneity in the propensity to innovate of the four groups of firms persists even controlling for innovation inputs. The results of table 2 and 3 shed some light on the relationship between ownership advantages and internationalisation. On the one hand, our findings are 8 Productivity measures are available for the period , so panel estimators can be used in these cases. Given that we have the information internationalisation status only for 1996, we could implement only random effect estimators. Regressions using random effect estimators broadly confirm the size of the differences in productivity, but in a few cases they turn out slightly less precisely estimated. 12

13 consistent with the idea that firms involvement in foreign activities can be associated with both ex ante and ex post advantages. In other words, the more firms are engaged in international markets, the more they need to be endowed with some immaterial assets, such as a superior technology, enabling them to face their competitors and overcome the costs and risks of carrying out their operations abroad; but a higher involvement in international activities may be also associated with learning and technology sourcing, further reinforcing their ownership advantages. As we have noted already, given the limitations of our data, it is not possible to identify the direction of causality (from performance to internationalisation or viceversa), nor the whether one of the two effects prevails. Furthermore, differentials in productivity may well originate from other sources we do not control for, such as market power and proximity to final markets. However, the fact that both MN MAN exhibit higher productivity than MN NON MAN, that the latter outperform firms with a lower commitment to foreign markets and that such differentials are robust to controls for R&D effort has important interpretive implications. In fact these results are consistent with the idea that the higher the engagement in international activities, the greater the productivity, holding constant at least some of the most important sources of technological advantages. Our results may thus reflect the fact that firms which are more rooted in foreign contexts can also be in a better position to gain access to foreign sources of knowledge. Finally, when we turn to heterogeneity in R&D effort and innovative output, our results highlight that it is R&D intra-muros (and not R&D extra-muros) that makes the difference between multinationals with manufacturing activities and all other categories of firms. This is a key feature of multinationals: they are more likely than other firms to be able to autonomously develop new products and processes, but they also have more absorptive capacity to gain access to external sources of technology. 5. Conclusion The paper has highlighted how intra-industry hetereogeneity in Italy is associated with different internationalisation modes. Consistently with some recent modelling in trade theory and with more consolidated views in the economics of international production, we have shown that firms with a high engagement in foreign 13

14 activities, also exhibit better economic and innovative performances. Companies with the highest international involvement, namely firms with manufacturing activities abroad, are characterised by both the highest productivity premiums and the highest R&D efforts and innovative performances. However, multinationals with a lower commitment to foreign markets, i.e. with non manufacturing activities abroad only, do show levels of productivity that stand between those of multinationals with manufacturing activities abroad and those of mere exporters; but they do not innovate more than the latter. This might have important policy implications. In fact, by favouring the growth of firms with manufacturing activities abroad it is more likely that technological capabilities also increase, and especially intra-muros R&D activities are enhanced. This could represent a fundamental channel for the creation of innovation opportunities in this country. References Balcet G., Evangelista R. (2003) Global technology: innovative strategies of multinational affiliates in Italy, mimeo. Barba Navaretti G., Castellani D. (2004) Does investing abroad affect performance at home? Comparing Italian multinational and national enterprises, CEPR Discussion Papers. Basile R. (2001). Export behaviour of Italian manufacturing firms over the nineties: the role of innovation. Research Policy 30 (8): Basile R., Giunta A., Nugent (2003). Foreign Expansion by Italian Manufacturing Firms in the Nineties: an ordered probit analysis, Review of Industrial Organization. Bernard A., Jensen B. (1999) Exceptional Exporter Performance: Cause, Effect or Both?, Journal of International Economics, 47, Brainard (1993) A Simple Theory of Multinational Corporations and Trade with a Trade-off Between Proximity and Concentration, NBER Working Paper No. 4269, Febraury Brainard (1997) An Empirical Assessment of the Proximity-Concentration Trade Off Between Multinational Sales and Trade, American Economic Review, 87, 4, Buckley P., Casson M. (1981) The optimal timing of a Foreign Direct Investment, The Economic Journal, 91, 361, Cantwell J. (1989) Technological Innovation and Multinational Corporations, Basil Blackwell, London. Cantwell J. (2000) A survey of theories of international production, in Pitelis C., Sudgen R. (eds) The nature of the transnational firm, 2 nd Edition. 14

15 Cantwell J., Narula, R. (2001) The Eclectic Paradigm in the Global Economy, International Journal of the Economics of Business, Vol. 8. Castellani D (2002a) Firms technological trajectories and the creation of foreign subsidiaries, International Review of Applied Economics, 16, 3. Castellani D. (2002b) Export behavior and productivity growth: evidence from Italian manufacturing firms, Welwirtshaftliches Archiv, 138, 4 Castellani D., Zanfei A. (2004) Multinationals, innovation and productivity. Evidence from Italian manufacturing firms, mimeo. Clerides S.K., Lach S., Tybout J.R (1998) Is Learning by Exporting Important? Micro- Dynamic evidence from Colombia, Mexico, and Morocco, Quarterly Journal of Economics, August, Cohen W., Levin R. (1989) Empirical studies of innovation and market structure, in Schmalensee R. and Willig R. (eds) Handbook of Industrial Organization, North Holland. Criscuolo C., Martin R. (2003) Multinationals, foreign ownership and US productivity leadership: evidence from the UK, mimeo. Criscuolo C., Haskel J., Slaughter M. (2004) Why are some firms more innovative? Knolwedge inputs, knowledge stocks and the role of global engagement, mimeo. De Backer K., Sleuwagen L. (2003) Why are foreign firms more productive than domestic firms?, mimeo. Doms M., Jensen B. (1998) Comparing wages, skills, and productivity between domestically and foreign-owned manufacturing establishments in the United States, in Baldwin R., Lipsey R. and Richardson J.D. (eds) Geography and Ownership as Basis for Economic Accounting, University of Chicago Press. Driffield and Love (2002) Foreign Direct Investment, Technology Sourcing And Reverse Spillovers, The Manchester School, 71, 6, December Dunning J. (1970) Studies in International Investment, London, Allen &Unwin Dunning J. (1993) Multinational Enterprises and the Global Economy, Addison Wesley, Wokingham, England. Dunning J., Narula, R. (1995) The R&D Activities of Foreign Firms in the United States, International Studies in Management & Organization, 25, 1-2, pp Fors & Svensson (2002) R&D and foreign sales in Swedish multinationals: a simultaneous relationship?, Research Policy, pp Frenz M., Ietto G. (2003) The impact of multinationality on the propensity to innovate: an analysis of the UK Community Innovation Survey 3, Presented at the Workshop on Empirical studies on innovation in Europe, Urbino, 1-2/12/2003 Gilbert R. (1989) Mobility barriers and the value of incumbency, in Schmalensee R. and Willig R. (eds) Handbook of Industrial Organization, North Holland. Girma S., Gorg H. and Strobl E. (2004) Exports, international investment and plant performance. Evidence from a non-parametric test, Economics Letters, forthcoming. Girma S., Kneller S. and Pisu M. (2003) Exports versus FDI: An Empirical Test, GEP Research Papers 2003/21, University of Notthingham. Griliches Z. (1998) R&D and Productivity: The Econometric Evidence, The University of Chicago Press. 15

16 Head K., Ries J. (2003) Heterogeneity and the FDI versus Export Decision of Japanese Manufacturers, NBER Working Paper Helpman E., Meliz M., Yeaple S. (2004) Export versus FDI with Heterogenous Firms, American Economic Review, 94, 1, Hirsh S. (1976) An international trade and investment theory of the firm, Oxford Economic Papers, 28, Horst T. (1971) The theory of the multinational firm: optimal behaviour under different tariff and tax rates, Journal of Political Economy, 79, Hymer S. (1960) The international operations of national firms: a study of direct investment, MIT Press, Cambridge, MA Kraay A. (1999) Exports and Economic Performance: Evidence from a Panel of Chinese Enterprises, Revue d Economie du Developpement, 1-2/1999, pp Kuemmerle W. (1999) The Drivers of Foreign Direct Investments into Research and Development An Empirical Investigation, Journal of International Business Studies, 30, pp Markusen J. (2002) Multinational firms and the theory of international trade, MIT Press, Cambridge, MA. Melitz M. (2004) The Impact of Trade on Aggregate Industry Productivity and Intra- Industry Reallocation, Econometrica, 71, 6, Narula R., Zanfei A. (2004) The international dimension of innovation, in Fagerberg J., Mowery D. and Nelson R. (eds) Handbook of Innovation, Oxford University Press, forthcoming. Nelson R. (1991) Why do firms differ and how does it matter?, Strategic Management Journal, 12, Pfaffermayer M., Bellak C. (2002) Why foreign-owned are different: a conceptual framework and empirical evidence for Austria, in R. Jungnickel (ed.) Foreign-owned firms: are they different?, Palgrave-Macmillan, Houndmills. Reinganum J. (1989) The timing of innovation: research, development and diffusion, in Schmalensee R. and Willig R. (eds) Handbook of Industrial Organization, North Holland. Sterlacchini A. (2002). The Determinants of Export Performance: A Firm-Level Study in Italian Manufacturing. Welwirtschaftliches Archiv 137 (3). Tybout (2003) Plant and Firm-Level Evidence on New Trade Theories, in Kwan Choi E. and Harrigan J., eds, Handbook of International Trade, Basil Blackwell. Vaccà S., Zanfei A. (1995) Capturing value from local context: the decentralization of decision within transnational corporations, in Schiattarella R. (editor) New Challanges for European and International Business, Proceeding of the EIBA Annual Conference, Urbino, December 1995 Wakelin K. (1998). Innovation and Export Behaviour at Firm Level. Research Policy 26: Zanfei A. (2000), Transnational Firms and Changing Organisation of Innovative Activities Cambridge Journal of Economics, vol

17 Figure 1 Cumulative distribution of the log of TFP, by firm type 1 dom_mkt export mn_nonman mn_man ltfp2 Figure 1 Cumulative distribution of the intra-muros R&D intensity, by firm type 1 dom_mkt export mn_nonman mn_man rdintxy 17

18 Table 1 Distribution of the sample and basic statistics, by type Labour productivity Approximate TFP N. firms Employees TFP % DOM MKT 98 13% EXP % MN NON MAN % MN MAN % Total % Innovating products Share of firms Innovating processes Carrying out R&D Total R&D R&D intensity Internal R&D External R&D DOM MKT 28% 40% 34% 1.9% 0.2% 1.7% EXP 59% 59% 65% 2.5% 0.7% 1.7% MN NON MAN 69% 66% 73% 3.0% 0.9% 2.1% MN MAN 80% 78% 89% 3.1% 1.2% 1.8% Total 60% 61% 66% 2.6% 0.8% 1.8% 18

19 Table 2 Firm performances and innovative behaviour, by type Dependent variable (1) (1) (2) (2) (3) (3) (4) (5) (6) (7) (8) (9) Estimation method OLS RE OLS RE OLS RE PROBIT PROBIT PROBIT TOBIT TOBIT TOBIT MN MAN 0.109** 0.158** 0.101** 0.136** 0.061** 0.067** 0.171** 0.137** 0.210** 0.010* 0.007** (4.52) (3.18) (4.31) (2.93) (4.77) (2.49) (2.99) (2.45) (3.78) (1.78) (2.08) (1.47) MN NON MAN 0.057** ** 0.094** 0.029** (2.80) (1.38) (4.62) (2.24) (2.63) (1.08) (1.53) (0.90) (1.21) (1.28) (0.27) (1.64) DOM MKT ** ** ** ** ** (0.55) (0.49) (0.46) (0.19) (0.39) (0.80) (4.81) (2.42) (4.06) (2.44) (4.00) (1.47) Constant ** ** (0.61) (0.83) (3.03) (0.35) (0.49) (0.34) (0.83) (2.55) (0.99) Observations Test on coefficients 670 MN MAN MN NON MAN 0.05* 0.10* ** ** * 0.00 p-value [0.06] [0.08] [0.64] [0.40] [0.02] [0.17] [0.12] [0.12] [0.01] [0.56] [0.09] [1.00] Baseline category: Exporters. All regressions are estimated controlling for the age of the firm, sector, region and size dummies. Robust t-statistics in parenteses below estimates. Asterisks denote confidence levels (**: p<.05; *: p<.10) RE: Random effect panel estimator; Marginal effects are reported Dependent variables: (1) Log of Labour productivity (2) Log of Approximate TFP (3) Log of (estimated) TFP (4) =1 if the firm introduced a product innovation in (5) =1 if the firm introduced a process innovation in (6) =1 if the firm carried out R&D in (7) Total innovation expenditures in 1996 as a share of turnover (8) Intra-muros R&D in 1996 as a share of turnover (9) External R&D in 1996 as a share of turnover 19

20 Table 3 Internationalisation, innovation and productivity Dependent Variable Log of TFP Log of TFP Product Innovator (dummy) Estimation Method OLS RE PROBIT MN MAN 0.050** 0.056** 0.141** (3.89) (2.06) (2.28) MN NON MAN 0.024** (2.21) (1.05) (1.29) DOM MKT ** (1.10) (0.96) (3.55) Intra-muros R&D intensity ** (1.30) (1.26) (3.73) Extra-muros R&D intensity ** ** (2.09) (1.48) (6.49) Product innovator dummy 0.042** (3.24) (1.08) Process innovator dummy (0.38) (0.05) Technological cooperation with national counterparts 0.029** ** (2.05) (0.75) (2.37) Technological cooperation with international counterparts 0.043** ** (2.99) (1.53) (2.42) Constant (0.79) (0.19) Observations R-squared 0.17 Test on coefficients MN MAN MN NON MAN 0.03* p-value [0.06] [0.30] 0.27 Baseline category: Exporters. All regressions are estimated controlling for the age of the firm, sector, region and size dummies. Robust t-statistics in parenteses below estimates. Asterisks denote confidence levels (**: p<.05; *: p<.10) RE: Random effect panel estimator; Marginal effects are reported 20