The food industry is the main manufacturing industry in Europe, representing 15.6% of total sales and over one third of world trade in agricultural products and food ( Serrano & Pinilla, 2014 ). In Spain, the sector making the largest contribution to gross domestic product after metal is the agro-industrial complex representing over 22% of sales in the manufacturing sector and employing around 13% of the workforce (21% in manufacturing sector) (Industrial Survey, Spanish Statistical Office, INE). According to data from the economic report by the the Spanish Federation of Food and Drink Industries (FIAB, Federación Española de Industrias de Alimentación y Bebidas ) it was consolidated as the top export sector in 2015. Food industry exports reached 25.4 billion euros in 2015, representing 10.2% of total exports in Spain.

The development of exports in the Spanish food and drink industry indicates a major shift towards internationalization. Advances in expanding the industry to other countries are reflected in both the growing intensity of sales abroad (between 1970 and 2012, the mean annual rate of increase in exports was around 4%) and the rising number of destination countries for the exports (over 175 geographical locations in 2016).

This process of internationalization can be explained by the existence of a home market effect in the food and drink industries in Spain ( Serrano et al. , 2015 ) and the intensification of intra-regional trade. This type of trade refers to that taking place between countries that have signed preferential agreements, a phenomenon dubbed the "spaghetti bowl" by Bhagwati et al. (1999) and Baldwin (2006) . There is some consensus that regional trade agreements (RTAs) have boosted trade between the signatory countries, sparking intense regionalization. Thus, in the last few decades a significant part of the trade in agricultural and food products has occurred within regional blocs ( Sarker & Jayasinghe, 2007 ; Grant & Lambert, 2008 ; Jayasinghe & Sarker, 2008 ; Serrano & Pinilla, 2012 , 2016 ). Some authors suggest that RTAs have increased partners' bilateral agricultural and food exports by 30-40% on average. As can be seen in Table 1, the European Union (EU) in particular was especially successful in liberalising the exchange of agri-food products among its members. Intra-EEC trade between the member states accounted for just 17% of total world agricultural and food trade when the community was founded. By the end of the twentieth century, this figure was close to 30% ( Pinilla & Serrano, 2009 ). After Spain joined the EU, the percentage of exports to this destination climbed to over 80% of the total. Since the end of the integration and transition period in 1992, the restructuring of Spanish exports has consolidated the community market as the natural destination for its products, while the rest of the world has decreased in importance ( Contreras & Bacaria, 2000 ; Clar et al. , 2015 ).

According to the normalised Herfindahl-Hirschman Index (HHI) ¹ of geographical diversification of exports in terms of volume (taking 175 destinations into account, right axis of Fig. 1), the growth in concentration during most of the last few decades is considerable, as the greater importance of flows with European partners shows. Finally, some geographical diversification has been seen in the past decade comprising both increasing exports to new EU partners from Eastern Europe and more importance placed recently on non-European exports.

The Spanish food and drink industry has focused its exports on large-scale regional markets in developed and geographically close countries ( Clar et al. , 2015 ). Breaking down Spanish agricultural exports by destination region highlights the fact that removal of barriers to entering the EU intensified the degree of internationalization and the importance of this region; however, it did not change the main destination markets for Spanish exports (France, Italy, Germany and the UK have always been very important countries for Spanish trade). The only significant exception is Portugal, which has purchased more since it joined the EU, to take second place in around 2016. The process of redirecting flows towards the region especially affected other countries in the OECD, which have gradually decreased in importance, especially the USA in the last decade. Also noticeable is the recent rising trend in exports to other countries outside the EU, namely Russia, China and the Middle East, although their share is still relatively small. This integration process and its effects on the foreign trade of agri-food products have been addressed in some studies. Research has been focused on the impact of liberalising trade barriers ( Contreras & Bacaria, 2000 ; Sanz & Gil, 2001 ; Gil Pareja et al. , 2005 ; Selva & Álvarez-Coque, 2011 ). More recent studies have shown the presence of the home market effect to be a key factor in the upsurge of Spanish agri-food exports ( Clar et al. , 2015 ; Serrano et al. , 2015 ).

To sum up, the internationalization of Spanish agri-food companies has increased, especially in exports to traditional European markets, due to a significant decrease in transaction costs in these markets from the removal of trade barriers and the uniformity of food safety standards in the region. Although Spanish companies are increasingly exporting to more countries, it is only recently that the geographical diversification of Spain's exports can be considered of any significance. The recession in Europe has led Spanish firms to try to increase their sales in markets that differ greatly from their traditional ones, e.g. the Middle East, China and Russia. However, companies operating outside the European region are confronted with highly protected markets commercially and the steep costs of adapting products to local regulations ( Serrano & Pinilla, 2011 ).

This section covers the process of company internationalization from a dynamic perspective. According to the Uppsala Model ( Johanson & Vahlne, 1977 ), the internationalization of companies follows a sequential process that depends on the gradual accumulation of knowledge of external markets. Johanson & Vahlne (1977) acknowledge the costs of acquiring foreign market knowledge. Early internationalizations have large learning costs because of unfamiliarity with foreign markets, cultures and environments ( Johanson & Vahlne, 1990 ). To explain the expansion of operations into new markets, the Scandinavian school used the concept of 'psychological distance', comprising a set of barriers that hinder a company's internationalization.

This model, together with the learning by exporting theory, presents the benefits of internationalization: the appearance of new business opportunities as new commercial relationships are established, while the existing ones are exploited ( Johanson & Vahlne, 1990 ). Furthermore, following the theory of transaction costs, exporting companies face the costs of entering new markets and increased coordination costs arising from international expansion. The Uppsala school used the concept of 'psychological distance', comprising a set of barriers that hinder a company's internationalization ( Johanson & Wiedersheim-Paul, 1975 ). According to the Uppsala model, firms first choose to enter nearby markets in terms of psychological distance with low market commitment, followed by subsequent entries into markets with greater psychological distance. In other words, internationalization is an evolutionary process linked to better knowledge and stronger commitment to foreign markets.

The negative effect in the initial stages and the high coordination costs of some over-internationalized firms form the basis of the three-stage theory of international expansion that some authors have proposed for other sectors ( Contractor et al. , 2003 ; Lu & Beamish, 2004 ).

A variety of benefits stem from the geographical expansion of company operations. An incremental focus and obtaining economies of scale, scope and experience are probably the most common reasons in the literature to explain why companies internationalize. Investment in the company's property, plant and equipment and costs sunk into overseas expansion require companies to have a minimum volume of operations to reach the profitability threshold. Increasing the intensity of international operations and entering new markets sequentially form the usual procedure to achieve economies of scale. In some cases, the increase in results arose from the opportunity to use several assets across product families (economies of scope). Lastly, as mentioned above, as companies increase their foreign operations, they acquire more market knowledge, which translates into taking advantage of economies of experience. From an incremental point of view, an increase in the volume of operations abroad results in a decrease in the mean cost.

A second group of arguments that are well documented in the literature can explain the international expansion of companies. These include: diversification of risks ( Annavarjula & Beldona, 2000 ) by operating in markets with a changing cycle, for example; improving performance by leveraging specific assets abroad, for example exploiting the competitive edge of innovations in various markets ( Buckley, 1988 ; Caves, 1996 ; Delios & Beamish, 1999 ); and obtaining benefits from exploiting market imperfections ( Caves, 1971 ).

Following the theory of transaction costs, companies immersed in an internationalization process face the costs of entering new markets and an increase in coordination costs arising from international expansion.

On the one hand, related to the 'psychological distance' concept, companies tackle markets they do not know. In the initial stages especially, companies incur 'research and information costs' since they have to spend time and effort on researching the target market and adapting their products to local consumer tastes and very diverse regulations. On the other, they also incur an increase in 'negotiation and decision costs' as they have to contact intermediaries, compile information on offers, travel to establish trade networks, decide, negotiate, draft contracts and take out risk cover.

Lastly, as the company increases the volume of its foreign operations and diversifies, it incurs more 'supervision and control costs' to comply with agreed conditions. Transaction costs rise, especially in very distant markets, and transport costs are also steep, since the environments differ greatly and the barriers to market entry, associated with trade protection policies or a wide variety of regulatory and technical reasons, are difficult to overcome. For some authors, the key to decreasing coordination costs is the stability of the regulatory framework, or compliance with the guarantees of property rights.

Internationalization, therefore, is the consequence of a process of incremental adjustments to the changing conditions of the company and its environment ( Johanson & Vahlne, 1977 ). The issue of whether there is a positive relationship between internationalization and performance, or whether there is an optimal degree of internationalization are very controversial subjects ( Hsu et al. , 2013 ; Powell, 2014 ). Ruigrok & Wagner (2005) and Wagner (2012) reviewed papers covering this aspect with very disparate results. Some have a positive linear relationship, others a negative effect, and yet others curvilinear U-shaped or inverted U-shaped relationships with a positive or negative effect based on the company's degree of internationalization.

In an attempt to synthesise previous research, some authors have proposed a horizontal S-shaped relationship between performance (y-axis) and internationalization (x-axis) ( Contractor et al. , 2003 ; Lu & Beamish, 2004 ; Li, 2005 ). Contractor et al. (2007) classified companies into various stages depending on their degree of internationalization: new exporters (Phase 1) are firms that have just started expanding abroad and are facing the initial costs of crossing the border; mature exporters (Phase 2) are companies intensifying the internationalization process and reaping positive benefits from an increased volume of business; and lastly, highly internationalized firms (Phase 3) are companies whose export and geographical diversification have grown to the extent that they face additional costs arising from greater management complexity. Consequently, firms in the third phase revert to a negative effect on their performance.

First phase (negative slope)

As is well known, international markets represent an important source of new opportunities for improving firm performance. Most of these opportunities are derived from the resource-based view of the firm, since international expansion gives the firm the chance to explore the interdependencies between business units, developing new capabilities and creating a competitive advantage.

Despite these benefits, operating across countries usually creates several new costs for the firm. In the early stages these costs are associated with the liability of foreignness. The most important relate to distance (transportation, coordination at a distance and over time zones, etc.), and the company's lack of familiarity with the destination's markets and legal systems ( Zaheer, 1995 ). The importance of these liabilities of foreignness depends of the different cultural values, levels of development, quality of institutions, and the skills and experience of the firm in foreign markets ( Hitt et al., 2006 ). The international expansion of firms involves entering unknown markets and incurring an increase in transaction costs. Firms face costs in seeking information, researching markets ( Hofstede, 1980 ), visiting sales representatives and intermediaries, negotiating contracts and supervising compliance. The cost of this learning goes together with an increase in coordination and communication costs ( Rosenzweig & Singh, 1991 ).

From the sector's point of view, the food exporter also needs to adapt the product to technical safety and quality standards, to the requirements of a new label, a new language in some cases, and the logistics needed to preserve products, since some are perishable. Firms face steep initial costs in this first stage, which narrow their margins. Implicit behind this argument is the incremental internationalization theory proposed by Johanson & Vahlne (1977) . Obviously, many costs will occur at any stage in international expansion. However, in terms of impact on performance, the negative influence in this first phase is greater, since the initial high costs are distributed over a volume of business that is still very small. Consequently, the negative effect on performance caused by reduced margins in the first phase, comes from a combination of the steep initial costs of crossing the frontier and an insufficient scale.

Second phase (positive slope)

In the second phase, the company benefits from the different opportunities arising from international operations. The firm advantages to the firm are the economies of scale, scope and experience linked to more foreign activity ( Buckley & Casson, 1976 ; Caves, 1996 ). Furthermore, as in the other stages, it is well documented in the literature that internationalized companies perform better due to the diversification of risks ( Annavarjula & Beldona, 2000 ), leveraging of specific assets abroad (for example innovations), which are the source of competitive advantage ( Buckley, 1988 ; Caves, 1996 ; Delios & Beamish, 1999 ).

Economies of scale occur when activity leads to an increase in profits through greater production factor specialization (technological know-how, organizational know-how, brand names). Following Ghoshal (1987) , scale can foster dynamic benefits such as experience and the learning effect because higher volumes allow for progressive cost reduction.

Expanding international volumes can also lead to the appearance of economies of scope as a result of the common use of production factors (assets, know-how and brand names) above and beyond the potential offered through product diversification ( Grant et al. , 1988 ; Kim et al. , 1993 ; Caves, 1996 ).

Hamel & Prahalad (1985) and Kim et al. (1993) have also shown that market diversification provides a firm with multiple national market bases from which it can retaliate against any aggressive moves by its competitors, therefore reducing its risk.

Another advantage of international expansion is the learning effect of operating in different markets and the ability to develop more diverse capabilities. Thus, international diversity fosters innovation and prepares firms for achieving good results in dynamic environments ( Kogut, 1983 ; Ghoshal, 1987 ; Kim et al., 1993 ).

The agri-food firm in these phases has already overcome initial product adaptation costs, has established more stable distribution networks and can diversify geographically by taking advantage of economies of scale and scope. The positive effects of international expansion occur once the firm has crossed the threshold of necessary business, after the export volume that reduces the mean cost of the operations has been reached, and it has also accumulated experience to exploit opportunities in other nearby markets, in this case, slow but steady expansion throughout the European region.

Third phase (return to a negative slope)

Despite the fact that the costs of exporting to new markets tend to reduce as international activity increases, some new costs appear which are associated with the greater geographical dispersion of operations. Following the theory of transaction costs, geographical dispersion increases managerial information and processing demands. Coordination between units, essential in the exploitation of economies of scope and scale, becomes much more difficult, and firms must develop a strong ability to manage their global needs.

Highly geographically diversified companies again leading to a negative performance slope due to a new reduction in margins. There are several reasons for this. Some authors highlight the increase in coordination costs associated with geographical dispersion as they have to deal with several different regulations ( Sundaram & Black, 1992 ). Others emphasise that directors face more complexity ( Grant, 1987 ), information overload ( Hitt et al. , 1997 ), loss of information or distortion in governance ( Hoskisson & Turk, 1990 ). Lastly, the institutional and cultural diversity of entering more diverse environments has a negative influence on costs ( Bartlett & Ghoshal, 1989 ). Following Contractor et al. (2003) and the Uppsala school's 'internationalization theory' hypothesis, over-expanding (over-diversified) firms face incremental costs of further expansion into peripheral nations (smaller and less profitable markets) and the growth of coordination and governance costs may exceed the benefits of further expansion, because of the complexity of global operations.

The food trade is also subject to a series of special characteristics that can help us understand the negative slope faced by companies in this last phase. As mentioned above, it is a highly protected sector and regulated outside the frontiers of regional blocks ( Serrano & Pinilla, 2014 ). Therefore, companies that decide to export on a more global scale face higher costs due to an increase in trade barriers. If we add in the higher cost of operations resulting from a longer physical and psychological distance, and that in some cases geographical diversification occurs in very small markets, the outcome is lower margins in foreign operations. The recent literature analysing the determining factors in international trade flows, generally using gravity models, has highlighted the importance that distance, as a proxy for transport costs and other associated costs, continues to have in the current second wave of globalization ( Jacks et al., 2011 ). Also in the case of agri-food products, there is clear evidence in the same direction ( Serrano & Pinilla, 2012 , 2014 ). Finally, also for Spanish agri-food exports, a recent study using gravity models has shown that for the existing 13 agri-food subsectors, between 1970 and 2012 and with a destination of 175 markets, distance is significant and has a negative coefficient ( Serrano et al. , 2015 ). In other words, trading to distant destinations entails additional costs. This is not surprising, as there is empirical evidence that for agri-food products, transport costs have remained fairly stable over the long term in the second globalizing wave ( Serrano & Pinilla, 2010 ).

In this context, this paper proposes to test the following hypothesis:

Hypothesis 1: The relationship between internationalization and performance in food companies results in a horizontal S-shaped curve graph representation (performance y-axis - internationalisation x-axis) with three phases:

-In the first phase , the degree of internationalization has a negative impact on company performance.

-In the second phase , the degree of internationalization has a positive impact on company performance.

-In the third phase , the degree of internationalization has a negative impact on company performance.

The data we used came from the ESEE ( Fundación SEPI, 2014 ), a panel survey on business strategies conducted by the SEPI Foundation backed by the Spanish Ministry of Industry. This survey provides information on the strategies of Spanish manufacturing companies for the 1990-2012 period. The ESEE has already been used in several other papers. These have highlighted its representativeness as a sample and its unique information on business strategy. In our case, it contains relevant information on the internationalization of companies.

The empirical work uses a longitudinal panel between 1994 and 2012 comprising companies in the following business sectors: (1) meat industry; (2) food and tobacco products; and (3) beverages. The three groups selected from the ESEE correspond to the groups in the food, beverages and tobacco industry contained in NACE-2009.

Dependent variable : in the majority of studies, company performance is based on ROA, which is also the dependent variable used in this study. It is an indicator of how profitable a company is relative to its total assets. ROA gives an idea of how efficient a company's management is at using its assets to generate earnings: ROA=Net Income/Total Assets. In order to ensure the robustness of our results, we estimated an additional model using Return on Sales (ROS) as an alternative dependent variable. ROS is a ratio used to evaluate a company's operational efficiency, also known as the operating profit margin: ROS= Operating Profit / Net Sales. This is the profitability indicator presented by ESEE, which is earnings before interest, taxes and depreciation of sales. Some previous studies have already used this indicator successfully ( Almodóvar, 2012 ; Almodóvar & Rugman, 2014 ; Fernández-Olmos & Díez-Vial, 2015 ).

Independent variables : defining and measuring the degree of internationalization has become a controversial and unresolved issue among international business researchers ( Fernández-Olmos and & Díez-Vial, 2015 ). Since the degree of internationalization refers to the performance of a firm in foreign markets several authors have argued that a proxy must be found to measure the dimensions of the geographical scale and scope of foreign operations ( Grant, 1987 ; Sullivan, 1994 ).

In order to reflect the internationalization process of the firms selected for our study, we designed an operational measure combining both the geographical scale and scope of their foreign operations (Degree of Internationalization or DOI). In particular, we focused on exports instead of foreign direct investment, since exporting is their primary vehicle of internationalization.

The foreign sales ratio, defined as the ratio of a firm´s foreign sales to its total sales, is the most commonly used variable for capturing the geographical scale of foreign operations (see for instance Grant et al. , 1988 ). In order to capture the geographical scale of foreign operations, some researchers have used the number of export countries as a proxy for internationalization ( Delios & Beamish, 1999 ). However, this measure is significantly limited, as it does not distinguish between the markets, either in terms of intensity of the market or distance of the market from the home country ( Pangarkar, 2008 ). Therefore, this measure makes it difficult to accurately assess the extent of a firm´s international activities ( Kim et al. , 1989 ). To correct this deficiency, other measures of internationalization have also been used, such as the number of dissimilar geographical regions in which the firm operates. For example, Kim et al. (1989) and Hitt et al. (1997) used an entropy index weighted by foreign sales which includes not only the multiplicity of foreign markets, but also the extent to which a company's activities are spread across similar foreign countries within regions ( Qian et al. , 2010 ).

Based on the previous reasoning, this paper uses a measure that considers both the traditional proportion of the foreign sales variable and the dispersion of foreign sales across geographical regions, as shown in the equation below. Other authors that have used similar indexes are Grant et al. (1988) , Pangarkar (2008) , Fernández-Olmos (2011) and Fernández-Olmos & Díez-Vial (2015) .

The sample only includes exporting companies and, therefore, these were included in the empirical model. The turnover of these companies was considerably higher than non-exporting firms ² .

Table 3 contains the values of certain variables related to internationalization between 1994 and 2012. Looking at the variables in detail, it is observed that food exporters can be classified as large companies. In 2012, they employed a mean of 227 workers. Although size is a determining business characteristic in internationalization, innovation or marketing capacity are factors that influence a firm's growth strategies and the success of its international expansion ( Altomonte & Nicolini, 2012 ). As seen in Table 3, food companies tend to invest in innovation. According to 2012 data, on average, around 20% of the surveyed exporters introduced product innovations during the year, and 41% process innovations. In 1998, for example, 4.6% of their sales was invested in advertising investment. These companies with an average age of around 36 years. Lastly, exporting companies allocate a higher volume of resources to marketing. Despite the recent drop, the percentage is always higher than the industry average.