The International Expansion of Emerging Market Firms:
Evidence from the BRICs Using an Institutional- and
Resource-Based View
Master Thesis International Economics and Business
University of Groningen, Faculty of Economics and Business
Frank Aanstoot
S1803069
Supervisor: dr. M.S.S. Krammer
Abstract
In this Thesis the internationalization of emerging market firms stands central and this is
measured in terms of exports. Both resources and institutions will be taken into account, as these two features are identified by the extant literature as having an important effect on firm strategy in the context of emerging markets. The contribution of this Thesis is that it combines theories with regard to both developed and emerging nations, and it applies these to a dataset of over 5700 emerging market firms. The results indicate that firms from emerging markets which have developed more capabilities are relatively more internationalized, as these provide a source of productivity and hence competitiveness. At the same time, institutions still have a profound effect on the actions undertaken by emerging market firms, thus affecting their
internationalization.
Content
Introduction
1
Theory and hypothesis development
4
The Resource-Based View The Institutional-Based View
Combining the Resource-Based View and the Institutional-Based View Hypothesis development
Method
12
Data
Measurement of variables Model and Methodology
Emerging markets gain more and more importance and prominence in today’s
economy. While their share of world Gross Domestic Product (GDP) is on the rise, firms from these countries are more and more at the forefront of international business. One has only to think of some of the world-class brands which have their or
SabMiller and Tata) to see that firms from emerging markets are as prominent nowadays as firms from the developed world. It is therefore safe to say that overall, emerging markets have become more integrated with the world economy and nowadays can influence and shape the global competitive landscape to a large extent (OECD, 2009).
This statement especially applies to Brazil, Russia, India and China, the so countries or BRICs, a term first coined b
(O’Neil, 2001). These countries have rapidly expanded exports over the past two decades, primarily because they started to integrate in the world economy around twenty years ago. is illustrated in the next graph. Here it can be seen that especially the rise of exports by China stands out, that Russia and India gain a slightly improved share and that Brazil stayed on an equal level over the last three decades.
Graph 1: rising share of BRICs in terms
Africa is included. Source: Naudé et al., 2012.
Introduction
arkets gain more and more importance and prominence in today’s globalized economy. While their share of world Gross Domestic Product (GDP) is on the rise, firms from these countries are more and more at the forefront of international business. One has only to
class brands which have their origin in these countries (e.g. Lenovo, SabMiller and Tata) to see that firms from emerging markets are as prominent nowadays as firms from the developed world. It is therefore safe to say that overall, emerging markets have become
orld economy and nowadays can influence and shape the global competitive landscape to a large extent (OECD, 2009).
This statement especially applies to Brazil, Russia, India and China, the so
countries or BRICs, a term first coined by Jim O’Neil from the investment bank Goldman Sachs (O’Neil, 2001). These countries have rapidly expanded exports over the past two decades, primarily because they started to integrate in the world economy around twenty years ago.
e next graph. Here it can be seen that especially the rise of exports by China stands out, that Russia and India gain a slightly improved share and that Brazil stayed on an equal level over the last three decades.
Graph 1: rising share of BRICs in terms of world total exports. Note that in this graph South Africa is included. Source: Naudé et al., 2012.
globalized economy. While their share of world Gross Domestic Product (GDP) is on the rise, firms from these countries are more and more at the forefront of international business. One has only to
these countries (e.g. Lenovo, SabMiller and Tata) to see that firms from emerging markets are as prominent nowadays as firms from the developed world. It is therefore safe to say that overall, emerging markets have become
orld economy and nowadays can influence and shape the global This statement especially applies to Brazil, Russia, India and China, the so-called
BRIC-y Jim O’Neil from the investment bank Goldman Sachs (O’Neil, 2001). These countries have rapidly expanded exports over the past two decades, primarily because they started to integrate in the world economy around twenty years ago. This
e next graph. Here it can be seen that especially the rise of exports by China stands out, that Russia and India gain a slightly improved share and that Brazil stayed on an
South-This integration into the world economy being the prime example - in the BRICs developed countries over the past decade. crucial role in the structural change
For example, while the BRICs’ share of world exports was the 1980s, their combined share reached 13 percent
projected that the four countries collectively will overtake the G
2040-2050 (UNIDO, 2013). To see this, the next graph provides a nice illustration of the rising share of world GDP by the BRIC
Graph 2: rising share of BRIC-countries in world GDP. Note that i included. Source: Haraguchi and Rezonja, 2012
Although the BRICs will probably increase this share further over the next decades very likely do so following a different path than developed economies du
legacies and their late arrival on the world stage.
shifted policies based on Western practices, they stilldiffer to a considera countries. Furthermore, although the BRIC
overall basis, they have experienced varying degrees of liberalization since the 1980s
2012). Typically, emerging economies are still in the process of transforming their economies from largely state-controlled to a market based economy. For example, Russia (as part of the former Soviet Union) had a planned economy until 1991
government decided what products should be produced and how much. Overnight this system was overthrown and an economy based on market forces was introduced, demanding a lot of This integration into the world economy has resulted in staggering growth rates
-in the BRICs, especially when compared to the economic growth of developed countries over the past decade. In other words, trade, particularly exports,
crucial role in the structural change and growth rates of the BRICs, especially since the 1990s. while the BRICs’ share of world exports was less than 4 percent at the beginning of
mbined share reached 13 percent by 2010 (UNIDO, 2012). In fact, it is projected that the four countries collectively will overtake the G-7 in terms of global GDP by
To see this, the next graph provides a nice illustration of the rising share of world GDP by the BRIC-countries over the last two decades.
countries in world GDP. Note that in this graph south included. Source: Haraguchi and Rezonja, 2012.
Although the BRICs will probably increase this share further over the next decades
do so following a different path than developed economies due to their historical legacies and their late arrival on the world stage. That is,even though the BRIC-countries have shifted policies based on Western practices, they stilldiffer to a considerable extent with OECD
Furthermore, although the BRICs tend to converge to Western-style policies on an have experienced varying degrees of liberalization since the 1980s
emerging economies are still in the process of transforming their economies controlled to a market based economy. For example, Russia (as part of the
had a planned economy until 1991. For more than seventy years the government decided what products should be produced and how much. Overnight this system was overthrown and an economy based on market forces was introduced, demanding a lot of
- with China , especially when compared to the economic growth of
In other words, trade, particularly exports, has played a since the 1990s.
at the beginning of In fact, it is
n terms of global GDP by To see this, the next graph provides a nice illustration of the rising
n this graph south-Africa is also
Although the BRICs will probably increase this share further over the next decades, they will e to their historical
countries have ble extent with
OECD-style policies on an have experienced varying degrees of liberalization since the 1980s (UNIDO, emerging economies are still in the process of transforming their economies
changes in the mindset and actions of people and firms. Accordingly, what this means is that the institutional environments of the firms residing in the BRIC-countries have undergone major shifts and currently still run through a process of constant change. Hence, with the liberalization efforts and the opening up of the respective economies, market-based resources and firm
capabilities are becoming more important, while one also still needs to take into account the institutional context since there remain challenges in the BRIC-countries (Chari and David, 2011). This is the reason why Hoskisson et al. urged scholars to focus on understanding the relationship between firms’ assets and the changing nature of the countries’ institutional infrastructure (2000).
Due to the growing importance of the BRICs in the world economy it is important to study the phenomenon and to understand what drives their success in terms of growth rates. In this Thesis this success will be investigated through exports, since these have played a key role in the economic growth of emerging economies. Or, to be more precise, their economic growth has for a large part been underpinned by export growth (UNIDO, 2012). Therefore, in this Thesis the internationalization of Emerging Market Firms (EMFs) stands central, in particular the export activities of firms residing in the BRICs. A micro-based approach will be applied, as various authors have pointed out that there exists a wide variation among firms (see e.g. Jensen et al., 2011). In other words, firm heterogeneity will be accounted for. This will be done by measuring two features that have an impact on firm strategy and which have been identified by
otherscholars to be important in an emerging market context, the first being firm resources and capabilities and the second being the institutional environment of the firms. Accordingly, the main research question of this Thesis is: to what extent do firm capabilities and the institutional environment influence the internationalization of emerging market firms?
By answering this question, this Thesis contributes to the literature in two important ways. First, there is a research gap with respect to export behaviors of firms from emerging economies. Although exporting is the easiest and often less costly way for firms to internationalize (as opposed to licensing or foreign direct investment), there have been few empirical studies
2008). Accordingly, this Thesis will help answering important questions in the discipline of international economics and business, such as “why do firms differ?” and “how do governments influence business decisions?”
This Thesis is organized as follows. The next section will first of all give an overview of the relevant literature, and will thereafter present the hypotheses which will be tested in the following chapters. Section three will present the methodology and the underlying economic and econometric logic behind it, as well as a description of the data. The fourth section will show and discuss the results. Finally, section five concludes.
Theory and hypothesis development
This part starts with a separate discussion of the Research Based View (RBV) and the
Institutional Based View (IBV), and thereafter these two views will jointly be discussed in the context of emerging markets and it will be argued why in this Thesis these two perspectives stand central. Following and building upon these first parts, the hypotheses will be developed and posited.
The Resource-Based View
However, not all resources have the potential to create a sustained competitive advantage, i.e. an advantage which is hard to duplicate by other firms. In order to create such an advantage four requirements are needed. The first is that the resource is valuable, the second that it is rare, the third is that it should be hard to imitate and the fourth is that the resource is in-substitutable (Barney, 1991). What this basically means is that the RBV aims to explain how a sustained competitive advantage can be obtained relative to other firms operating in the same market, and this is attributed to developing and thereafter exploiting unique capabilities. In other words, the main issue of the RBV is about how firms actually operate and on the basis of this it continues with accounting for firm heterogeneity (Loicket et al., 2009).
It is important to note that the resources a firm could develop depend on the resource environment of that firm. What this means is that resources, and on the basis of this the
capabilities built, will vary between countries.That is, some factors which create a competitive advantage will be present in some countries but not in others. Consequently, due to the general lack of resources in emerging markets, it is crucial for firms to obtain them and to transform into productive use in order to become successful (Wan, 2005). However, it should be noted that emerging economies are placing more value on market based policies, so it is reasonable to expect that resources will become relatively less scarce. This will in turn ensure that more capabilities can be developed, and firms can exploit these domestically developed capabilities internationally by means of exporting. This is the reason why Hoskisson et al. suggest that EMFs “must understand the relationship between its company assets and the changing nature of the institutional environment” (2000).
The Institutional-Based View
In short, the main concern of the IBV is the influence the institutional architecture of a country has on firms and how this shapes the behavior of organizations. To put it differently, institutions can be defined as the rules of the game in a society (North, 1990). They are therefore not static but dynamic, in the sense that rules of the game can change over time. The foregoing means that the institutional environments of countries change over time, focusing for example less on informal regulations and ties between the government and business but more on formal laws and the gradual shift from state-owned enterprises dominating the economy towards private
choices made by firms are not only based on their capabilities, but also on the constraints and opportunities offered by the institutional infrastructure of countries (Peng et al., 2008).
It is worth mentioning here that research focusing on developed economies also takes into account the institutional framework, but this is done in a much more indirect way, assuming the smooth functioning of institutions as a background condition. This assumption has to be radically altered when doing research in emerging economies, given the important variance in terms of institutional infrastructures between emerging markets and mature economies (Ibidem). Accordingly, just as firms differ in terms of their capabilities and resource endowments, so do countries differ in terms of their institutional apparatus.
Hence, it is necessary to be aware that emerging economies are not one homogeneous group but that vast heterogeneity exists among these countries (Hoskisson et al., 2012). This means that EMFs cannot take for granted the smooth functioning of the institutional environment in which they operate, which in turn implies that institutions are not mere background conditions but have a direct impact on firm actions in emerging economies. To put it like this, institutions are nowadays recognized as key characteristics of the economic performance of nations in general and of firms in specific, because they mold the decisions of enterprises (Van Hoorn, 2013). In other words, institutions matter. In a nutshell, institutions affect the decisions of firms on a wide range of topics, such as investment decisions, resource and capability development and internationalization efforts. To give one example, the degree to which proprietary resources are protected by the government can have a profound effect on the innovation efforts by firms. That is, when intellectual property rights are not thoroughly protected by law or when such laws are not enforced, firms have a smaller incentive to create such resources because they can be misappropriated by other firms (Kshetri, 2009). Again, firm resources and the institutional environment need to be analyzed in coherence instead of examining them separately.
Combining the Resource-Based View and the Institutional-Based View
As indicated in the introduction, Hoskisson et al. (2000) urge scholars to integrate the Resource-Based View (RBV) and the Institutional-Resource-Based View (IBV) to analyze firm behavior in
emerging economies. In fact, according to these authors the IBV is seen as one of the three most insightful when research is done in the context of emerging market (the others being a
environment (e.g. pro-market reforms in the case of the BRICs) resources of firms are crucial in order to stay competitive and to continue business. So, when this type of reforms are being implemented the need for more sustainable competitive advantages is increased since firms can no longer rely on the government for help but need to invest more in tangible and especially intangible assets. This leads them to conclude that in emerging economies the RBV is gaining more importance, while the IBV is also important as the rules of the game do not change radically, but more evolutionary over time. In other words, it takes time to change the
institutional architecture. They are supported in their view by a host of other scholars in the field, all stressing the point that the IBV is pivotal when doing research with a focus on emerging markets.
For example, Wright et al. (2005) suggest that scholars should take into account four perspectives: transaction cost theory, agency theory, resource-based theory and institutional theory. They analyze four strategic options with regard to conducting business in emerging economies, and in all these four options the RBV or the IBV are included separately or jointly, while the other two views are not consistently present. Another example would be Peng et al. (2008), who posit that the IBV is a third leg in the strategy tripod, next to the traditional RBV and the industry-based view. Again, the point is made that institutions matter and they have an impact on firms’ strategic choices and therefore need to be included when analyzing actions of firms. This argument is in line with that of Meyer (2004), who states that “corporate strategies, institutional change and the development of local resources and capabilities are mutually interdependent”. In other words, when looking at EMFs, one should not look at these different features separately but rather analyzing them in interplay, as they impact on each other. For that reason, it is not sufficient to focus on just one attribute in emerging markets.
It stems from these different works that the RBV and the IBV are features which are consistently taken into account when research is conducted with respect to emerging countries. Hence, in this Thesis the focus will be on these both views.
Hypothesis development
predominantly been driven by export growth. The question then becomes how resources and the institutional infrastructure impact on the internationalization efforts by EMFs. This question is taken up to investigate whether or not a firm in one of the BRIC-countries engages in exporting and how many it exports of its total sales. The former is also called export propensity, and the latter is also referred to as export intensity. Accordingly, when a firm does not export at all this is equal to exports=0, and when its exports are a positive share of its total sales this is equal to exports > 0.
Prior research has shown that technological capabilities and the sophistication of the workforce are important determinants for the long-run competitiveness of the firm and
consequently are resources which are considered as having an effect on the internationalization of the firm (Singh, 2009 & Wagner, 2007). In fact, Alvarez finds that technological resources can createfirm-specific advantages, especially when the firm serves markets that are not the home-market (2004). Hence, these capabilities transform resources, both tangible and intangible, of the firm into new products and technologies and by doing soimprovethe competitive position of the firm. To put it like this, the technological capabilities of a firm affect the flow of products and services across nations (Singh, 2009).
With regard to the composition of the workforce, characteristics of the individual workers will have an impact on the labor productivity of the firm as a whole, and therefore these are taken as a determinant influencing the internationalization of the firm. That is, the more productive firms go abroad by serving foreign markets while the less productive firms stay at their home market. This is the case because firms with a higher productivity level can incur the additional costs associated with exporting. Given these theoretical underpinnings, I hypothesize that
Hypothesis 1: the technological level of EMFs is positively related to their internationalization. Hypothesis 2: the workforce level in terms of skills and knowledge of EMFs is positively related to their internationalization.
Another characteristic of the firm which features frequently in research with regard to the RBV are management capabilities. These are considered especially important with regard to managing human resources and knowledge of the industry in which the firm operates. That is, more
which can improve firm performance and innovation (Yiu et al., 2007). So, by managing the employees more effectively, a firm can develop its human capitaland as was shown above this is critical in creating and sustaining a competitive advantage. Furthermore, more capable and experienced managers know what is going on in the industry. Specifically, they have more experience in evaluating the potential for exporting within the industry (Sapienza et. al, 2006). Accordingly, I hypothesize that
Hypothesis 3: management capabilities of EMFs are positively related to their internationalization.
As a last characteristic with regard to the RBV I will look at firm quality. This feature is
important for EMFs that want to internationalize, since the general lack of resources in emerging markets limits the degree of sophistication and hence products from EMFs can be seen as having less quality. This in turn will make it harder for these firms to export. However, when they possess an internationally recognized standard for quality, economic efficiency is improved as these provide a basis for lowering information related transaction costs (Nadvi, 2008). So, a quality signal can enhance trade between countries from different nations ‘‘when a supplier’s capability to design and supply conforming products need to be demonstrated”
(Clougherty&Grajek, 2008). In other words, a quality signal in emerging markets can foster trade, specifically exports, because it sends a clear message to external actors about the
sophistication of the establishment and its production process. Consequently, this will lower the barrier to internationalize with regard to EMFs and will induce them to start exporting or
increase their exports as a share of total sales. Therefore I posit that
Hypothesis 4: there will be a positive relationship between EMFs possessing an international standard and their internationalization.
host of different determinants are included in the extant literature when it comes to exporting. However, what seems to be a common theme is that emerging markets are typically regarded as higher-risk and higher-regulated countries. What this basically means is that there is both a higher degree of uncertainty and that there are constraints in the home environment with regard to doing business. These generate incentives for companies in emerging economies to look for other markets and by going to other countries they can to some extent overcome the market failure of the home nation (Contractor et al., 2007). In other words, the market failure at home poses a constraint on EMFs to grow and to become more competitive. This will lead firms to search for other markets where they can profitable sell their products. Hence, by exporting EMFs can create more value as they can increase sales, and by this performance (He et al., 2013).
One such a characteristic that drives firms to search for other markets due to uncertainty in the home market is the degree of political instability. It is a widely held view that such instability is detrimental for economic growth. In other words, this means that there exists a negative relationship between the degree of political instability and the prospect for economic growth at home (Jong-A-Pin, 2009). This will in turn ensure that firms will look for other markets in order to overcome dependence on a market which is expected to slow down or to shrink. To put it like this, to overcome this constraint EMFs will start looking for other markets to sell their products and services. So, a higher degree of political risk, ceteris paribus, will lead firms to search for other markets in order to generate revenue. It is important to note that political risk in this sense does not mean a complete shutdown of the country as happened for example in Egypt during the protests in 2012 and 2013. Such an event would most likely lead firms to internationalize less, due to the impossibility to reach important logistical points because of road blockades and a shortage of gasoline, or workers not showing up at work. Rather, here it should be interpreted asbeing that emerging markets are typically seen as having a lower degree of rule of law and a higher degree of clièntelism1. It thus pushes EMFs that are less well connected to home government officialsto search for other profitable markets, as managers could perceive such a political situation as a constraint to grow at home due to a higher degree of uncertainty or because other firms are favored. So, because political instability is often more intense in
emerging markets than in developed countries, it drives competitive firms to search for other countries where they can profitably sell (Guillaumont, 1999). This leads to the following
hypothesis:
Hypothesis 5: there will be a positive relationship between political instability in the home market and the internationalization of EMFs
Another important feature of emerging economies is that a key role is being played by the
informal economy. In fact, for some emerging markets it is estimated that the vast majority of the population operates predominantly in the large, but hidden, informal sector whichis not recorded in official GDP statistics (London & Hart, 2004). The incentives encouraging entrepreneurs to participate in the informal, as opposed to the formal, economy in emerging markets are in general driven by costs, since for many it is simply too costly or complicated to engage in the formal economy. In other words, only the most productive and competitive firms can afford to conduct business in a formal way in many emerging economies (Ibidem). In such an economy, where sometimes informal firms set the rules of the game which formally registered firms need to follow,the former players demonstrate their power by setting boundaries and safeguarding them (Ibidem). This drives formal firms to search for other markets where they can find a level-playing field with other formally registered firms and to overcome market failure in their home country. In other words, EMFs can diversify their business through exporting and this provides them with a manner to exploit internationally their competitiveness build at home (Buckley et al., 2007). This leads me to the following hypothesis:
Hypothesis 6: there will be a positive relationship between the degree of informal sector players dominating the economy of the home country and the internationalization of EMFs.
perspective acts like a strategic tool which improves the competitiveness of the firm (Martine et al., 2007). Hence, this will lead to a higher export intensity, as Bernard and Jensen showed that the more productive firms export a higher share of total sales (2004). The latter proposition argues that when a firm bribes in its home country, it gets tied more heavily to the home market since this will enforce its relative position in this country. In other words, they will focus more on the home country, leveraging their relative strengths in this market, something which diverts attention away from profitable overseas markets. To put it like this, firms can achieve a
preferential treatment by their home government when bribing and this decreases
internationalization (Hundley & Jacobson, 1998). Hence, this would lead to a lower export intensity, since the attractiveness of serving foreign markets will be lower in such an instance. According to prior empirical research, the latter view dominates (Lee &Weng, 2013). That is
Hypothesis 7: there will be a negative relationship between home country bribes and the internationalization of an EMF.
Method
In this part I will first of all describe the data source which is used to perform the subsequent analysis. Thereafter a measurement of the dependent, independent and control variables will be given. Finally, the last part of this section will provide the model and the methodology which is used to present the results in the following section.
Data
I obtained data for my analysis from the World Bank’s Enterprise Surveys. Enterprise Surveys collect information about a country's business environment, how it is perceived by individual firms and about the various constraints to firm performance and growth. These data are provided by the top-manager of the establishment. For firms to be included in the survey they were first screened to assess their eligibility. When eligible, the answers given by the top-managers were controlled for quality, as local contractor knowledge (who conducted the interviews) was used to perform the interviews. In all four countries, the data provides information of the
non-agricultural economy of firms with at least 5 employees and positive amounts of private
sector, services sector, the retail sector,the transport, storage and communications sector and the IT sector. Note that this definition excludes the following sectors: financial intermediation, real estate and renting activities and all public or utilities-sectors. It is also important to mention that state-owned enterprises are not included in the sample.
Data for Brazil was obtained in 2009, for Russia it was obtained in 2012, for India it was obtained in 2005 and for China it was obtained in 2012. The data for the four countries was selected using stratified random sampling. This means that the full sample in each country is obtained by separating the population elements into non-overlapping groups (e.g. firm size, industry and geographic location), called strata, and then selecting a simple random sample from each stratum. The full sample (grouping all observations for the four countries together) consists of 11008 observations (Brazil 1802, Russia 4220, India 2286 and China 2700). After dropping firms for which no observations were available on the respective variables I end up with a total of 5738 and 2443 observations for the Tobit and Probit model (depending on the explanatory variables being included) and 511 observations for the OLS model (see below).
Measurement of variables
Dependent variable(s)
Following the literature (see e.g. Estrin et al., 2008 & Yong Gao et al., 2010), I will in first instance measure the internationalization of EMFs by export intensity. Export intensity basically means how much of its total sales a firm exports. In other words, this is expressed as a
Independent variables
To measure firm resources and capabilities I include five explanatory variables. The first two variables will measure an EMFs’ technological level. As posited frequently by the RBV, investments in technological capabilities are important in shaping firm capabilities. The first variable will be expressed in terms of total expenditure of EMFs on new machinery and equipment, which comprises investments made in these assets. In the questionnaires the top-managers are asked to provide this number in local currency units. In order to have one common measurement these numbers are converted to US dollars, for which the respective exchange rates will be taken at the end of the year in which the questionnaire was held2. Also, the observations are reported in millions of dollars in order to interpret the results more easily. The expectation is that higher expenditures of this kind will lead to more internationalization. The question in the surveys is: “in the last complete fiscal year, how much did this establishment spend on purchases of: machinery, vehicles and equipment (new and/or used)”? Hence, this is a continuous random variable and it can therefore take any value within the interval zero to infinity.
The second measure of an EMFs’ technological level will be analyzed by looking at whether a firm uses or licenses technology from a foreign firm. Accordingly, this will be a dummy variable and it takes the value 1 if a firm does so, and it takes the value 0 if it does not. So, when a firm uses or licenses technology from a foreign firm it can be expected that it has more capabilities, and hence is more likely to internationalize. The question used in the surveys is: “does this establishment at present use technology licensed from a foreign-owned company”?
Secondly, I will include a variable which measures the workforce level in terms of skills and knowledge. In other words, firm competencies can be measured by asking managers to what degree he or she thinks skills and knowledge pose a problem to the operations of the firm (Yiu et. al, 2007). Managers indicated this in the survey along a scale ranging from 0 to 4, with 0 indicating no obstacle and 4 indicating a very severe obstacle. Past literature has supported the view that a firm's capability in terms of human resources is a key factor in enhancing firm
capabilities (see e.g. Lau and Ngo, 2004). Accordingly, I expect that when the number is close to 4 firms will have a lower level of internationalization and vice versa. In other words, due to how this variable is measured in the questionnaire a negative sign indicates that the EMF exports a
2Exchange rates were used from the website www.freecurrencyrates.comand the last available day of the
greater share of its total sales.This would confirm the theory that EMFs having a more capable workforce export a larger share of their total sales, i.e. that there is a positive correlation between a more capable workforce and internationalization. The question in the surveys is: “is an
inadequately educated workforce: No Obstacle, a Minor Obstacle, a Moderate Obstacle, a Major Obstacle, or a Very Severe Obstacle to the current operations of this establishment”?
Thirdly, management experience will be measured to explain firm capabilities, as more experienced managers have in general more knowledge about how to handle their personnel and they possess knowledge about the industry and specifically they have more experience in
evaluating the potential for internationalization within the industry (Sapienza et. al, 2006). Accordingly, this explanatory variable will be measured in terms of how many years of experience the top managementhas working in the specific sector the firm is active. So, the expectation is that the higher this number, the greater the export propensity and intensity of the EMF will be. The question in the surveys is: “how many years of experience working in this industry does the top manager have”? In other words, this explanatory variable also is a continuous variable.
Lastly, as a fifth variable measuring firm capabilities, an indicator variable will be included which indicates whether a firm holds an internationally-recognized quality
certification, such as ISO 9000. In other words, this will be a dummy variable taking the value 1
if an EMF holds such a quality certification and 0 otherwise. It is expected that EMFs having such a standard, which indicates a higher degree of capabilities, will internationalize relatively more. The question in the survey is: “does this establishment have an internationally-recognized quality certification”?
To measure the institutional environment of the EMFs I will include three explanatory variables. The first variable will be the degree of political instability as experienced by managers. The measure is on a scale from 0 to 4, with zero indicating no political instability and 4 signaling very severe political instability. I expect that EMFs facing more political instability in their home country will be more likely to export and to export a higher share of total sales3. The question in
3As explained in the theory section, this is due to how this variable is being measured in the questionnaires. It is
the surveys is: “to what degree is political instability: No Obstacle, a Minor Obstacle, a Moderate Obstacle, a Major Obstacle or a Very Severe Obstacle to the current operations of this
establishment”?
Secondly, I will include a variable which measures the impact of practices from
companies in the informal sector on EMFs, as experienced by managers. Again, the measure is
on a scale from 0-4, with zero indicating no obstacle and 4 signaling a very severe obstacle for the operations of the EMF. The expectation is that when managers perceive that informal competitors pose a problem to the operations of the establishment, EMFs will internationalize relatively more. The question in the surveys is: “are practices of competitors in the informal sector: No Obstacle, a Minor Obstacle, a Moderate Obstacle, Major Obstacle, or a Very Severe Obstacle to the current operations of this establishment”?
Lastly, as a third variable measuring the institutional architecture the degree of
corruption will be included. This is measured by asking managers what percentage of total
annual sales firms need to pay to government officials to get things done. Accordingly, when this number is high it is expected that EMFs get more heavily tied to the home market, and therefore I expect that their export propensity and export intensity is lower. However, some observations in the surveys indicated this variable in total annual value. To have on common measurement, these numbers will be transformed into percentages by taking the total annual value as a
percentage of total annual sales which are also given in the surveys. The question in the surveys is: “on average, what percentage of total annual sales, or estimated total annual value, do
establishments like this one pay in informal payments or gifts to public officials for this purpose (i.e. to get things done)”?
Control variables
As is common, I will also include a set of control variables. The first one will control for firm
size, as bigger firms tend to internationalize relatively more than smaller firms (see e.g. Bernard
et al. 2007). Firm size will be measured by how many permanent, full-time individuals work in the respective establishment. The question in the survey is: “at the end of the last complete fiscal year, how many permanent, full-time individuals worked in this establishment”? Following the
literature, I will transform this number taking the natural logarithm (see e.g. Yiu et al., 2007 &Estrin et al., 2008).
Secondly, firm age will be controlled for, withage being measured as the year when the survey was held, and then the year when the firm was founded will be subtracted from this former year. The question in the survey is: “in what year did this establishment begin
operations”?Firm age is important to take into account in a transition economy, because older firms that have been established in the pre-reformed period are in general more risk-averse (Yiu et al. 2007). Since exporting is riskier than selling domestically, older firms are less likely to export and they are less likely to export a greater share of their total sales.
Thirdly, a dummy control variable will be included whether the firm is a foreign wholly
owned subsidiaries (FWOS). It will take the value 1 if private foreign individuals, companies or
organizations own 100% of the common stock and 0 otherwise. When a firm is a wholly owned subsidiary it is more likely to internationalize, since these are often used by their parent firms as export platforms. That is, they are making use of the lower production costs in emerging
economies. The question in the surveys is: “what percentage is owned by: private foreign individuals, companies or organizations”?
Related to the previous control variable, I will also include a set of dummy variables measuring in which industry the EMF is active. To do so I follow the questionnaires, and include sixdummy variables which indicate whether the EMF is active in the manufacturing,
construction, services, retail and wholesale,the transport, storage, and communications or the IT sector. So, these variables will take the value 1 if the EMF is active in the respective sector and 0 otherwise4. In the estimations I include five dummies to avoid the dummy variable trap. These dummy variablesare included because it is relatively easier to export manufactures than for example service or retail and wholesale products, which are often non-tradable (see e.g. Riedl, 2010). Hence, when a firm is in the manufacturing sector it is more likely to internationalize. In the survey, both the interviewee and the interviewer indicated in which sector the establishment is active. When I found a discrepancy between both answers, I took the sector which was indicated by the interviewer. In the tables reporting the results the industry dummies are not
shown, but they are included throughout the different estimations and can accordingly be interpreted as industry fixed effects.
Finally, I will control for the country in which the respective EMF is active. That is, in the estimations I will include three dummy variables for the four countries which are included in this Thesis: Brazil, Russia, India and China.Table 1 in the appendix summarizes the description of the variables listed above.
Model and Methodology
To estimate the impact of firm resources and the institutional environment on the internationalization of EMFs I follow the literature (see e.g. Fernandez and Nieto, 2006
&Nassimbeni, 2001) and will in first instance use a Tobit model given the fact that I have data of two sorts, limit observations (y=0) and non-limit observations (y>0). In other words, for all the observations it is known whether or not an EMF exports while for the EMFs that do, it is known how much they export (Breen, 1996). This leads to the following expression:
= 0 ∗≤ 0∗ ∗≥ 0
y* is a latent variable, and censoring takes place from below with a lower limit of 0. Hence, in this equation the two types of observations whichare in the data, the limit observations and those that are positive, are generated by the latent variable y* crossing the zero threshold or not
crossing that threshold (Hill et. al, 2011). In this case, when I would drop observation for the dependent variable when it does not cross the threshold, theleast squares estimators of the regression parameters are biased and inconsistent as there will be a correlation between the independent variables and the error term.
1985). I explore for the presence of non-normality and heteroskedasticity (see next part), and when these two conditions are not met I will use a Heckman two-step estimation model which can provide a better fit to the data as it relaxes the non-normality condition and can correct for heteroskedasticity.
When using a two-step model the first part consists of an estimation using a binary outcome equation (i.e. does the EMF export or not) and the second part uses a linear regression (i.e. conditional upon if the EMF exports, how much does it export). In other words, I will use in the first part a Probit model, and in the second part an OLS method. What this boils down to is the following:
( | ) = PrPr( = 1| ), ( | ∗= 1, ) ( = 0| ) ∗= 0∗> 0
That is, the probability of y depends on x, a set of explanatory variables, and y will take the value 0 if the latent variable y* also takes the value 0, and y will take the value 1 if y* is above 0. Hence, when y*=0you could only observe Pr(y=0) and when y*>0, f(y|y*=1) is the conditional density of y, again dependent on x, a set of explanatory variables (Cameron & Trivedi, 2009).
Thus, in simple terms the model I estimate is:
= 0 + 1 + 2 + 1 +
= 0 + 1 ℎ + 2 + 3
+ 4 + 5 + 6 + 7
+ 8 + 9(ln) + 10 + 11 + 12 1 + 13 2
+ 14 3 + 15 4 + 16 5 + 17 6 + 18 + 19
+ 20 + 21 ℎ +
Where internationalization measures the same limited dependent variable as described above, β0 represents the intercept, β1-5 represent the variables measuring the resources and capabilities of the firm (with β2 and β5 being dummy variables and β1 is expressed in millions of dollars), β6-8 represent the variables measuring the institutional architecture in which the EMF operates and β9-21 capture the control variables (with ln representing the natural logarithm of firm size with regard to β9) with β11-21 all being dummy variables. When estimating the model I do not include the dummy variable measuring whether an EMF resides in China and whether it is active in the IT sector, as in both instances this would make me fall in the dummy variable trap. In other words, by including only three of the indicator variables representing the possible four countries in which a firm can be active and including only five of the six industry dummies, the omitted variables (in this case China and the IT sector) define the reference group and I avoid the problem (Hill et. al, 2011). The error term is again represented by ε. With regard to both
equations, the positive signs in front of the variables are chosen arbitrarily, and do not say anything about the expected sign of the respective variables.
Results
managementas I had one observations indicating that a manager had 80 years of experience
working in the respective industry, and another 100 year. These seemed to be unrealistic and accordingly I dropped these two observations from the data. From the table it also stems that around 63% of the EMFs in the sample is active in the manufacturing industry (see D1), while the other industries have far less observations.
Table 1: Summary Statistics
Furthermore, I looked at the correlation matrix (see table 2 in the appendix), and no values were found which indicated that collinearity would pose a problem to my estimation. For the EMFs that report positive amounts of exports (i.e. y>0) a histogram is provided in the next graph. On the y-axis the percentage of EMFs can be found, while on the x-axis the export intensity is showed. It seems that EMFs by and large export a relatively small percentage of total sales, or
D6 5809 .0382166 .1917352 0 1 D5 5809 .0220348 .1468092 0 1 D4 5809 .1797211 .3839881 0 1 D3 5809 .0766053 .265987 0 1 D2 5809 .0499225 .2178037 0 1 D1 5809 .6334997 .4818898 0 1 chinaDV 5809 .3454984 .4755714 0 1 russiaDV 5809 .3336202 .4715465 0 1 brazilDV 5809 .1592357 .3659272 0 1 indiaDV 5809 .1616457 .3681572 0 1 FWOS 5798 2.926638 15.41375 0 100 age 5802 14.4657 11.60712 0 183 lsize 5755 3.530235 1.389776 0 10.30895 corruption 5809 1.26608 5.410318 0 100 informal 5808 1.047348 1.244238 0 4 instability 5801 1.017238 1.321981 0 4 qualityDV 5802 .3240262 .4680502 0 1 management 5692 15.79454 9.852898 0 60 workforce 5808 1.331267 1.359748 0 4 licenseDV 4780 .123431 .3289657 0 1 newmachinery 3058 .6251274 6.460729 0 296.1 intensity 5746 5.580753 19.34807 0 100
they export 100% of total sales since the middle range is relatively underrepresented in the histogram.
Graph 3: histogram of EMFs that report positive amounts of exports
Tobit estimation
related to the RBV, the workforce variable is significant at the 0.05 level, and the management variable is significant at a 0.01 level. Both dummy variables (license and quality) measuring the capabilities of the EMFs do match their expected signs, but they are not statistically significant at any conventional level. As for the newmachineryvariable, it is also not statistically significant atany conventional level and its sign also differs from what was expected. Hence, it seems from the results that EMFs that invest in new and/or used machinery, vehicles and equipment, ceteris
paribus, do export less although the effect is very small. The IBV variables are all significant,
the instability and informal variables at a 0.01 level and the corruption variable at a 0.1 level, thus confirming the hypotheses.
However, as indicated in the methodology sub-section, I will test for non-normality and
heteroskedasticity as the Tobit model yields inconsistent estimates when these two conditions are violated. To test for non-normality I follow Drukker (2002), who developed a conditional
moment test for evaluating whether the disturbances in a Tobit model have a normal distribution. Hence, this leads to the following:
=
= −
The following result was obtained using the tobcmcommand in Stata, which allows for the testing of normality when using a Tobit model with a lower limit of 0.
Table 2: testing for normality
Accordingly, because the calculated p-value is smaller than any conventional critical p-value, I reject the null hypothesis and therefore conclude that the errors are non-normally distributed. To test for heteroskedasticity I follow Cameron and Trivedi (2009) who propose a test of homoskedasticity when using a Tobit model. This test follows the conventional chi-squared distribution and so I compare the actual value against the critical value. Hence, this leads to the following:
12.793 0.00167
CM Prob > chi2
= ℎ
= ℎ −
To obtain the result in Stata a long list of commands is needed. This list is stored in a log-file and is not provided here, but is of course available on request. The 5% critical value is 5.99 while the calculated value is411.22 with a p-value close to 0. Hence, this outcome leads to a strong
rejection of the null hypothesis and I therefore conclude that my data is heteroskedastic. Due to the existence of non-normality and heteroskedasticity, the Tobit estimation procedure discussed above contains weaknesses. Therefore, I continue with the Heckman two-step model introduced in the methodology section. This estimation procedure relaxes the Tobit model assumptions, and is therefore better suited to analyze the data.
Two-step estimation
The first part of the estimation procedure when using a two-step model is using a binary equation (indicating whether or not the EMF exports), which basically models the probability that export intensity is bigger than 0. Hence, this is about export propensity. Here I will use a Probit model to estimate whether or not a EMF exports. The second part uses a linear regression model to estimate the export intensity, conditional upon whether or not the EMF exports. Or, to put it like this:
Pr( > 0) and
( | > 0)
Hence, I construct a new dependent variable (Dintensity) which effectively is a binary indicator of positive exports, so that it takes the value 1 if intensity > 0 and it will take the value 0 when
intensity=0(Cameron and Trivedi, 2009). For the model to be identified, I need at least one
equation (Ibidem). From table 4 in the appendix it stems that this variable satisfies this criterion. The second reason is theoretically derived, and first of all posits that such a quality signal reduces transaction-costs and hence can provide EMFs with a critical push to start exporting since the adoption of such a signal enhances the overall quality and reliability of exporters and accordingly provides buyers improved information about the EMF. That is, the adoption of an international standard matters more for the EMF selling the product than for the firm buying the product (Clougherty&Grajek, 2008). In other words, holding such a quality signal could
influence the decision of an EMF to start exporting, but does not necessarily say anything about the level of exports. The summary statistics for the binary indicator are provided in the next table.
Table 3: summary statistics for binary indicator
Results of the Probit estimation are provided in table 4 in the appendix. Again, I did control for industry fixed effects but the results are not included in the table. In this model the dependent variable is Dintensity, thus only measuring export propensity. In the first model (column 1) I again only regressed the binary indicator on the control variables, in the second column I
additionally include the variables related to the RBV, in the third column of table 4 I again added the variables related to the IBV to the equation and left out the variables with respect to the RBV and in column 4 finally all the variables listed in the equation in the methodology part are included. To start with, three of the control variables, these being size, FWOS andInviaDV are again highly statistically significant while the variables indicating the age of the EMF and the two dummy variables of Russia and Brazil are not significant at any conventional level.
However, these three latter variables do have their expected signs. For the RBV-variables almost the same results are obtained as in column 2, with qualityDVand managementbeing significant at a 0.01 level. However, workforce now also is significant at a 0.1 level and in comparison to column 2 also has its predicted sign. The other two variables are not significant, but do have the expected signs. With respect to the IBV-variables, instability and informal are both significant at
Dintensity 5809 .1468411 .3539779 0 1
a 0.01 level, and they have the expected sign. Corruption however is not significant at any conventional level although it does have its expected sign.
In the second step I will use an OLS estimation where I exclude the qualityDVvariable but where I include the Inverse Mills Ratio (IMR). The IMR can be interpreted as an additional variable containing the generalized residuals from the Probit estimation. The coefficient of this additional variable is a function of the correlation between the two error terms of the model.So, if the IMR is significant it is a sign of a sample selection problem and it also indicates the direction of this correlation. In this case the selection problem would stem from the fact that in the OLS model only EMFs are included which pass the critical threshold, defined here as y*>0. On the other hand, if the IMR is not significant it rules out the possibility of a non-random sub-sample. To put it like this, the IMR is created from the first step Probit estimation and accounts for the fact that theobserved sample of exporting EMFs is not random (Hill et. al, 2011).
Results of the linear regression estimation can be found in the appendix, table 5. Again, the industry dummies are not included in the table but I did control for the industry in which the EMFs are active. Here the dependent variable is intensity and only EMFs are included which report positive amounts of exports. Hence, it measures the export intensity of EMFs, which is conditional on the fact that only EMFs are included that passed the threshold of y*>0. That is, only EMFs are included which do export. Again, in the first column the results can be found where the intensity variable is regressed on the control variables and the IMR, in the second column I included the RBV variable additionally and in the third column I included the IBV variables and excluded the RBV variables.
Finally, in column 4 of table 5 the results of the full model are reported. All the control variables are statistically significant and have their expected signs, except for the dummy
variable indicating that the EMF operates in India. Again, it seems that Indian EMFs export more that the reference group, which are EMFs located in China. This canbe attributed to the nature of the data, as almost all Indian firms are located in the manufacturing sector. In other words, from the results it seems that ceteris paribus EMFs operating in India export more than EMFs
Although the management and license variables are not significant at any conventional level, they do have their expected signs. Again, the variable indicating whether the EMF has introduced any new machinery has a different sign than predicted, although it is statistically highly insignificant. The variables related to the IBV all three have their predicted sign, and the
instability and the corruption variable are significant at the 0.1 level while the informal variable
is significant at the 0.01 level. Furthermore, it is important to note that the IMR is significant at a 0.05 level. That is, there seems to be sample selection bias and hence the method applied here is the appropriate one as the error terms of the Probit and OLS models are not independently determined. In other words, when using an OLS estimation over the EMFs which reported non-zero values for the intensity variable and excluding the EMFs which reported non-zero values for this variable5, without first estimating the Probit model and computing the IMR, would have
produced biased estimates. Here this would mean that the estimates would be upwardly biased since the IMR has a positive value.
I will again test for homoscedasticity with respect to the second step in the two-step estimation, since the presence of heteroskedasticity when using an OLS method would imply that my estimates are inefficient, although they will be consistent. This leads to the following:
= ℎ
= ℎ −
With a linear regression there is a simple command in Stata to test for heteroskedasticity, and the command is hettest. I executed the command, which yields the following result:
Table 4: testing for heteroskedasticity
As can be seen, the chi-squared value exceeds the critical value of 5.99 and the p-value is very close to zero. Hence, this leads me to reject the null hypothesis and I therefore conclude that
Prob > chi2 = 0.0000
chi2(1) = 29.57
Variables: fitted values of intensity Ho: Constant variance
heteroskedasticiy is present in my data. I also tested for normality with respect to the second step in the two-step estimation. Again, the hypotheses are the following:
=
= −
To test for non-normality in Stata when using OLS there is also a simple command, which in this instance is performed by sktest.I executed the test, which yields the following result:
Table 5: testing for non-normality
As can be seen in table 5, the p-value is again very close to zero, leading me to reject the null hypothesis and accordingly I conclude that the errors are non-normally distributed in the sample. However, when using an OLS estimation normality is not a condition to obtain consistent
estimates (see e.g. Cameron & Trivedi, 2009 and Hill et al., 2011). Nonetheless, due to the presence of heteroskedasticity I re-estimated the full model in the second step with robust
standard errors. The results can be seen in the last column of table 5in the appendix.Accordingly, my main results can be found in table 4, column 4, with respect to export propensity and in table 5, column 5, with respect to export intensity. Hence, when not controlling for
heteroskedasticityin the second step the standard errors are not valid and generally too small. Moreover, the corruptionvariable is now significant at a 0.01 level. The country dummies are still all three significant at the 0.01 level, indicating that ceteris paribus EMFs located in India export around 36% more relative to Chinese EMFs while EMFs located in Russia and Brazil export 30% less relative to Chinese EMFs.
Robustness checks
In table 6 I executed some robust checks. Accordingly, all the results in the table (columns 1-4) are corrected for the fact that the errors have a non-constant variance. Also, the industry fixed effects are not reported but were included throughout the different estimations. The dependent
yhat 2.4e+03 0.0404 0.0000 54.26 0.0000
Variable Obs Pr(Skewness) Pr(Kurtosis) adj chi2(2) Prob>chi2 joint
variable is intensity, thus measuring export intensity, conditional on y passing the threshold of y*>0. First of all, in column 1 I tested for non-linearity with regard to those variables which are indicated in the extant literature as potentially being different from linear. I did this by taking the squared values of the respective variables. As can be seen they all have the effect of diminishing the original effect except for the variable informal, which seems to reinforce the original effect. That is, the more practices of informal competitors in the home marketpresent an obstacle to the operations of the EMFs, the higher the export intensity of these firms. In other words, this variable seems to have an increasing marginal effect on export intensity, although it is not statistically significant at any conventional level. The diminishing effect for the other squared variables is represented by the respective signs which are the oppositein comparison to the non-squared variables. For example, although political instability leads EMFs to search for other profitable markets, it does so at a diminishing rate. Put differently, political instability has a diminishing marginal effect on the export intensity of EMFs. However, only the squared value of age is statistically significant, and it is so at a 0.1 level. Moreover, the effect of including these squared values undoes the statistically significant effect of most RBV and IBV variables which were significant in column 1.
Therefore, in columns 2 and 3 the squared values are left out again. In both columns the variable
newmachineryis also left out, and two other measures of productivity are included to see whether
these provide a result which is consistent with prior literature. That is, EMFs investing more in resources should have a higher productivity, and more productive firms should export more relative to less productive firms (and not less as is indicated by the negative sign in front of the variable newmachinery). Hence, the variable productivity is added in column 2. This variable measures average productivity for all EMFs and is obtained by dividing the total sales (measured in millions of dollars) of an EMFby its permanent, full-time employees. Contrary to what would be predicted however, this variable also has a negative sign, indicating that ceteris paribus more productive EMFs export a smaller share of total sales. The variable is however not statistically significant at any conventional level.
would be expected, although it is again not statistically significant (see column 3). Hence, from the results it seems that more productive EMFs export a smaller share of their total sales, although none of the estimated variables is significant at any conventional level.
As a last robustness check in table 6 I again included the variable newmachineryand the variable age2, as from column 1 it stems that the latter has an impact on the export intensity of EMFs. The results can be seen in column 4 of the table. The former variable still has a different sign as was predicted, although it is again not statistically significant. The latter is still significant at a 0.1 level, indicating that younger firms export a greater share of their total sales relative to older firms, although this variable has a diminishing marginal effect. Also, for the other variables their significance levels and signs stay the same, indicating that the results in table 5, column 5, are robust.
Finally, in table 7 I executed some additional regressions to check the robustness of my argument to include the qualityDVvariable in the Probit model, but to exclude it from my OLS estimation. That is, I looked for other variables which, from a pure theoretical point of view, potentially could impact the decision to start exporting (i.e. export propensity), but which do not impact the amount of exports (i.e. export intensity). In the table the results are corrected for
heteroskedasticity and industry fixed effects were controlled for in both estimations, although they are not reported in the table. The dependent variable is intensity, thus measuring only export intensity, again conditional on y passing the threshold of y*>0.
Hence, in table 7, column 1, I excluded an additional variable, which is the one measuring the experience of the top management. I picked this variable firstly because the excluded variable should have a significant impact on the probability of starting to export, i.e. on the selection equation. As can be seen in table 4, column 4, this variable has a statistically significant (at a 0.01 level) impact on the decision to start exporting. However, in all the results in tables 5 and 6 the management variable did not show up significant at any conventional level, although it consistently matched its expected sign throughout the estimations. Furthermore, from table 6, column 1, it stems that management experience has a declining marginal effect on export intensity, although it is not statistically significant. Still, this could be an indication that
management experience isrelatively more important with regard to export propensity than it is with respect to export intensity.
although most of the statistically significant variables lose significance in the sense that they are now significant at a lower level. However, the variable newmachineryis now also statistically significant at a 0.1 level and still has a sign different than would be predicted. In other words, EMFs that are more productive due to their investments in new machinery and equipment,
ceteris paribus, export less. Because thisclearly contradicts prior literature, it does not seem to be
a plausible outcome.
Hence, in column 2 of table 7 I again included the management variable but now excluded the instability variable. Again, I picked this variable first of all because the excluded variable should have a significant impact on the probability of starting to export, i.e. on the selection equation. As can be seen in table 4, column 4, this variable has a statistically significant (at a 0.01 level) impact on the decision to start exporting. Furthermore, from table 6, column 1, it stems that political instability has a declining marginal effect on export intensity, although just like the management variable its effect was not statistically significant. However, this declining marginal product of political instability could indicate that its effect is most likely felt with respect to the decision to start exporting, and not so much with regard to the decision of the amount of exports.
The results do not change much with respect to the results in table 5, column 5, although most of the statistically significant variables again lose significance in the sense that they are now significant at a lower level. However, with respect to column 2 the newmachineryvariable no longer is statistically significant, and the variable indicating the capabilities of the workforce is moreover no longer significant at any conventional level. This is not consistent with the results obtained in table 6, where this variable always shows up significant. Hence, this result seems to be inconsistent in comparison to the other estimations. Moreover, the instability variable itself shows up significantly throughout most of the estimations with regard to the OLS method in the second step of the Heckman procedure. That is, it seems that political instability has an effect on the export intensity. Hence, it would not seem feasible either to exclude this variable form the second step estimation.
Accordingly, with regard to export propensity I find support for hypothesis 2 at a 90% confidence leveland I find support for hypotheses 3, 4, 5, and 6 at a 99% confidence level. Furthermore, with respect to export intensity I find support for hypothesis 5 at a 90% confidence level, for hypothesis 2 at a 95% confidence level and for hypotheses 6 and 7 at a 99% confidence level. That is, export propensity is affected by management capabilities and holding a quality signal, but no support was found with regard to these variables affecting the export intensity of EMFs. Also, export propensity does not seem to be influenced by corruption, but it does affect the export intensity of EMFs. Moreover, both export propensity and export intensity are affected by the capabilities of the workforce, political instability in the home country and informal players dominating the economy in the home country. Lastly, I found no evidence in support of hypothesis 1. That is, the technological capabilities of EMFs do not seem to influence the internationalization of these organizations, nor with respect to export propensity nor with regard to export intensity. The above is summarized in table 6, where it can be seen which hypotheses are confirmed at their respective confidence levels, both with regard to export propensity and export intensity.
Table 6: hypothesis support
Hypothesis no:
Feature which stands central
Hypothesized sign
of correlation Propensity Intensity
1 technological level + No support No support
2 workforce level + 90% 95%
3 management capabilities + 99% No support
4 international standard + 99% No support
5 political instability + 99% 90%
6 informal sector players + 99% 99%
7 corruption - No support 99%
