Comparative Institutional Advantage As A Determinant Of Fdi Economics Essay
The Varieties of Capitalism literature proposes a concept of institutional arbitrage: as a result of comparative institutional advantages due to different investment incentives provided by types of institutions, companies relocate activities to countries, where the institutional set-up best supports a given activity. We test this proposition using German outward FDI figures on a sectoral level. We find evidence of institutional arbitrage using different ways to operationalize system differences between countries. Evidence crucially depend on the size of the receiving sector. FDI in a sector with comparative institutional advantage increases with sector size, while for a sector without a comparative institutional advantage the size of the sector does not matter.
1. Introduction
The comparative capitalism literature around Hall and Soskice’s (2001) Varieties of Capitalism puts forward a strong claim of comparative institutional advantage.The institutional set-ups in coordinated (CME) and liberal market economies (LME) are said to provide firms with comparative advantages in specific industries, in which dominating research patterns are more prominently utilized than others. Specifically, in LMEs institutions facilitate radical innovations and hence, firms have a comparative advantage in the very industries, which require radical innovation to prosper. On the other end of the spectrum, CMEs exhibit institutional features, which give the opportunity to engage in more incremental innovation strategies, which are utilized more in some industries than in others. The literature provides a considerable number of empirical investigations into the existence and the extent of comparative institutional advantage. The overall picture is mixed, but taking everything into consideration there seems to be considerable evidence in favor of comparative advantage of nations.
A further logical step for Hall and Soskice (2001) is then to propose a concept of institutional arbitrage: firms should exploit existing comparative institutional advantages by shifting production to countries, where the institutional set-up is best for the activity. Decreasing barriers to the flow of production factors and goods due to “globalization” make it easier to shift production to other countries. Apart from “traditional” determinants of foreign direct investment such as differences in labor cost, market size, transport costs and generally the distance between countries, the concept of institutional arbitrage implies an institutional motive for locational production choices of companies.
The empirical literature on the new institutional economics has long introduced institutional variables in order to explain foreign direct investment and trade patterns. The focus there, however, is on the quality of institutions in the colloquial sense of the word meaning “good” institutions. For instance, it can be shown that lower corruption in countries correlates with higher FDI inflows; similarly, the quality of contract enforcement in countries can explain parts of the production pattern of countries. This resonates with the large Doing Business project led by the World Bank (2007),which collects worldwide data on institutional differences in a wide range of areas. These data are then utilized to show the impact of institutional quality on financial development and growth.
The comparative capitalism literature has a different focus: different types of institutions exist as a result of multiple equilibria with complementary institutions. As stated above, the focus here is on the different investment incentives provided by different varieties of institutions. Hence, simply put, institutions in advanced capitalist countries are not better or worse, but lead to different outcomes with respect to industrial structures and dynamics or distribution of income. While it can be shown that institutional differences in kind lead to different sectoral growth rates and investment strategies (Carlin and Meyer 2003),the impact on cross-country investment and thereby the existence of institutional arbitrage has not been tested empirically, yet, on a cross-sectoral level. We make a first step of doing so using data of German outward foreign direct investment (FDI) on a sectoral breakdown. We follow the gravity equations approach, which explains cross-country variation of FDI flows on the basis of market size, labor costs and distance between host country and the country of origin. In order to model the institutional set-up of the host-country we include a number of institutional variables ranging from a simple binary system variable showing whether a host country is classified as CME or LME to indicators for specific institutional sub-systems of the host countries. Hence, the basic measure says whether countries are liberal market economies or coordinated market economies. This is amended by continuous measures of degrees of coordination using an indicator developed by Knell and Srholec (2007)and an expanded and updated version of the coordination index proposed by Hall and Gingerich (2004).In addition, we include indicators of industrial relations, corporate governance and labour laws. In a similar fashion as Allen et al. (2006),we group industrial sectors according to the dominant innovation pattern.
Our analysis suggests that there is a systematic institutional arbitrage to be observed by German companies. In other words, German companies contemplating relocation do seem to be influenced by the institutional endowment of host countries. However, this effect depends on the size of the receiving sector: larger sectors receive more German FDI if they exhibit a comparative advantage. If the advantage is not present, the size does not matter.
The analysis proceeds as follows: chapter 2 sets the stage by incorporating the concept of comparative institutional advantage of the comparative capitalism literature into its usage within the broader stream of new institutional economics. Against this backdrop, we then discuss the concept of institutional arbitrage as advocated by Hall and Soskice and give an overview about the empirical determinants of FDI. Chapter 3 presents our methodology, which follows widely used empirical methods of international economics to model direction of investment; chapter 4 presents the data. The results of our data are presented in chapter 5, while chapter 6 concludes
2. Comparative institutional advantage and foreign direct investment
Comparative advantage refers to lower opportunity cost of production in certain goods and services. In different strands of trade theory, comparative advantage based on labour productivity (Ricardo) and relative factor abundance (Heckscher/Ohlin), explains specialization and the direction of trade. Simply put, firms specialize in industries, for which in their respective countries the opportunity costs of producing are lowest and export those goods. Introducing institutions into the concept of comparative advantage of nations gives rise to the notion of comparative institutional advantage. It states that institutions (North 1990)provide companies in a given country with comparative advantages in the production of specific goods and services (Hall and Soskice 2001, Belloc 2006).For our purposes it is less relevant, whether countries have comparative or absolute advantages from the institutional environment. What is more, empirically observing trade or investment patterns of countries allows to infer a comparative advantage and not an absolute advantage (see also discussion in Franzese and Mosher 2002).
The theoretical and empirical literature on comparative institutional advantage can be broadly classified along two lines. On the one hand, some authors attempt to show that a higher quality of institutions in one country provides firms with a better contracting environment relatively to firms in a country with lower quality of institutions. This then has an effect on the whole economy. The second line of research, to which the present paper attempts to make a contribution, puts the main emphasis on different types of contracting institutions, which create different institutional environments entailing sectoral comparative advantages.
Including quality differences of institutions in theoretical and empirical studies to explain growth and development differentials around the world has substantially advanced growth theories. It has been shown that the quality of institutions, in particular of contracting institutions, has a long-term impact on the development of countries (Acemoglu et al. 2001, Acemoglu and Johnson 2005, North 1990).Also, trade theory is increasingly incorporating institutions and differences in contract enforcement as a way to reconcile theoretical predictions with actually observed trade patterns (Belloc 2006).
The theoretical background is provided by institutional economics approaches such as property rights theory (Hart 1995, Hart and Moore 1990) and transaction cost economics (Williamson 1985, 2000). There, the unifying feature is the importance of relation-specific investments for the choice of specific types of contractual arrangements. Property rights literature emphasizes the crucial role of residual control rights in an enterprise. Since contracts are incomplete, it is the owner who decides over the distribution of total surplus in cases of ambiguities or imprecision. Hence, there will always be under-investment of the non-owner as compared to a first-best solution. Low quality of contract enforcement in a country aggravates this because under-investment in relation-specific assets of non-owners will be even greater. Ceteris paribus, it is argued that the better the contract enforcement in a country, the more relation-specific investments are undertaken. As a consequence, countries with better contract enforcement (for instance less corruption in the judiciary) enjoy a comparative institutional advantage in those economic activities, which heavily rely on relation-specific investments. Oxley (1999) makes a related observation in a transaction cost economics framework: the quality of intellectual property rights protection in a host country has an influence on the governance choice by the foreign investor: the weaker the protection the more hierarchical the chosen governance structure.
Recently, a rapidly growing literature has used trade data to investigate whether trade flows follow the theoretically predicted lines of comparative institutional advantage. All of them use similar indicators to capture the differential need for contracting institutions, such as the contract intensity of industries as proxied by the thickness of input product markets (Nunn 2007),the complexity of products in industries measured by the concentration of input goods (Levchenko 2007)or survey results about the complexity of tasks in a given industry (Costinot 2009).They find that countries with better contracting institutions export relatively more of contract-intensive products than countries with worse institutions.
Differences in the quality of financial systems can also impact on comparative advantages of countries otherwise equally endowed (Baldwin 1989, Bardhan and Kletzer 1987).Earlier theoretical contributions are more broadly echoed in the Law and Finance literature (see an overview by Beck and Levine 2005 and La Porta et al. 2008). Empirical investigations that explicitly model comparative advantages arising of better (or worse) financial systems confirm this (e.g. Svaleryd and Vlachos 2005).
Including quality measures of institutions in theoretical and empirical papers in a wide range of economic applications is commonplace. Less attention has been paid to different types of institutions, such as different degrees of non-market coordination. This will be discussed in the following section.
2.1 Type of institutions: Varieties of Capitalism
Contributors in this tradition emphasize the point that institutions of coordination can protect relation-specific investments and thereby lower the degree of under-investment. That is, given equally good contract enforcement characteristics, different institutional settings result in incentives to invest in relation-specific investments. Hence, national institutions affect the production capacity differently across sectors of the economy.
Tightly coupled with this is the notion of institutional complementarities: simply put, two institutions are complementary when the existence of one raises the efficiency of a second institution. (Amable 2000, Aoki 2001, Aoki 1994, Hall and Soskice 2001, Höpner2005). A related interpretation is that one institution within a domain A can only function efficiently, when a second institution is present in domain B. The mathematical foundation is based on the notion of super-modularity, which essentially means that elements are linked in discrete structures.
An encompassing framework within this line of work is provided by the Varieties of Capitalism (VoC) literature (Hall and Soskice 2001, Estevez-Abe et al.2001). It takes a firm-centered view of the economy by analyzing the predominant coordinating device used by firms. In liberal market economies (LME) firms rely predominantly on the use of markets, hierarchies and arm’s length contracting. In coordinated market economies (CME) there are additional non-market institutions, which facilitate coordination among companies. Table 1 provides a stylized overview of the predominant institutions in CMEs and LMEs.
As already postulated by North (1990), the institutional environment provides incentives, in what kind of skills and knowledge to invest. While North is primarily concerned with the development of economies and hence lays great emphasis on the distinction of “good” vs. “bad” institutions, here the focus is on the type of investment decisions facilitated by the institutional environment. For Hall and Soskice this is connected to the specificity of investments and accompanying innovation strategies. In short, it is argued that in CMEs the institutional environment gives incentives to invest in highly specific assets, both for employees and for firms. On the contrary, in LMEs individuals will invest comparatively more in general assets, which can be put to use in a broad range of activities without losing value. The institutional settings in the sub-system of corporate governance and social security can serve as an example: in a CME the high protection of both employment and unemployment of workers gives incentives to invest in specific human capital on the part of the workers.
Table 1
CME
LME
Financial System
Bank-based
Capital market
Corporate Governance
Stakeholder
Shareholder
Industrial Relations
Country-, industry-level
Firm-level
Social Security
High protection, low flexibility
Low protection, high flexibility
Vocational training
Specific human capital
Generic human capital
Inter-firm relations
Cooperation
Competition
They expect that the appropriation of the quasi-rent of specific assets, i.e. being fired and having to accept a job with a different skill profile, is less likely than elsewhere. The management of the firm can furthermore commit itself to retaining the workforce during economic hardships, because the corporate governance system is built around the principle of ‘patient capital’. This means that banks and stakeholders secure credit through dense webs of cross-shareholding and other monitoring devices (Vitols2001, 2004). Hence, the company can pursue investment strategies, which rely on incremental product innovations, such as diversified quality production (Streeck 1991). Here, it can be seen that the focus is on outcomes of institutional complementarities: as a result of “fitting” institutions an innovation strategy of incremental innovation is feasible. For LMEs, the logic runs the other way: here employment and unemployment protection is not existent or low, hence employees will invest in portable, generic skills. They cannot be sure that investment in higher specificity of assets would pay off. For the management of firms in LMEs, it is of utmost importance to be able to present good performance measures to shareholders. Flexible labor laws make it possible to ‘hire and fire’, which is one way of achieving good short-term profitability figures. Here, the resulting innovation strategies for firms are radical innovations. Liquid financial markets and hostile takeovers make it easy to move fast into new markets and change innovation strategies quickly. Also, as stated above, fluid labor markets foster radical shifts of companies into new business fields. Hence, institutional complementarities both constrain and enable actors to engage in particular activities.
To sum up the basic rationale of VoC, the institutional environment in CMEs and LMEs facilitates different predominant types of assets in both systems. These, in turn, go together with different innovation strategies. Apart from Germany as the prime example other European and Asian countries are considered a coordinated market economy, while the Anglo-Saxon countries form the group of liberal market economies.
Table 2
Coordinated Market Economies
Liberal Market Economies
Mixed Market Economies
Germany
Netherlands
Belgium
Austria
Switzerland
Denmark
Finland
Sweden
Norway
United Kingdom
United States
Canada
New Zealand
Australia
Italy
France
Spain
Portugal
Greece
Apart from that, some countries are not placed in the CME camp, because they lack some decisive features of it. Greece, Spain, Portugal, Italy and also France are such ‘mixed market economies’ (Hall and Soskice 2001; Molina and Rhodes2007). See Table 2 for an overview. Some authors argue that it makes more sense to talk about liberal market economies and lump everything else into one ‘non-liberal’ group, whose defining feature is the stark contrast to countries, which predominantly rely on market relationships (cf. in Streeck and Yamamura2001).
Having defined sectoral innovation strategies and institutional differences between countries, it is argued that those innovation strategies give rise to comparative institutional advantages of nations, because different sectors rely more on one type of innovation than others. The support from institutional complementarities for certain activities as described above is different in LMEs from CMEs. It can be expected that this is reflected in the structure of the economy. Hall and Soskice (2001) submit the hypothesis that CMEs are relatively better at product strategies connected to incremental innovation, such as machine tools, consumer durables, engines and so on. LMEs, by contrast, are seen as relatively better in activities such as biotechnology, semiconductors, IT and so on. To provide evidence, Hall and Soskice provide patent data from the European Patent Office for patents from Germany and the US are used to calculate a specialization index. For the respective country, an industry is listed if that country specializes in the respective technology field. Specialization is given when the share of patents of a particular industry in total domestic patents is larger than the respective global share.
Table 3
CME (Germany
LME (USA)
Civil engineering
Consumer goods
Weapons
Nuclear engineering
Transport
Agricultural machines
Handling
Mechanical elements
Engines
Machine tools
Environment
Thermal processes
Material processing
Surfaces
Process engineering
Basic materials
Pharmaceuticals
Polymers
Control systems
Electrical energy
Surfaces
Basic materials
Agriculture, food
New materials
Biotechnology
Pharmaceuticals
Organic chemistry
Medical engineering
Control systems
Optics
Semiconductors
Information technology
Telecommunications
As can be seen from Table 3, Germany is specialized in the very areas, which are usually conferred to as incrementally advanced diversified quality production. On the other hand, American firms are relatively better in industries, which rely on radical innovations.
Taylor (2004)in a first step confirms Hall and Soskice’s (2001) classification of industries in either relying on incremental or radical innovation. Extending the analysis over a longer time span and a larger country set, however, he finds that the significance of results of clear comparative advantages between Germany and the USA exist hinges on the inclusion of the United States.
Apart from patents, trade data are also used to map comparative institutional advantage. Panuescu and Schneider (2004)use the share of high-tech and medium-tech technology sectors in total exports of a country as a dependent variable. The intuition is that LME countries will specialize more on high-tech industries with a large R&D-intensity, while CME countries specialize on medium-tech activities characterized by a lower R&D intensity. To correct for imports they include the relative advantage of both for a total of 20 countries. The results are overall supportive for the claim that LMEs specialize in industries dependent on high R&D expenditures and hence radical innovations, while the comparative advantage of CMEs of high relative to medium-tech categories is significantly lower than for LMEs (52-55). More rigorously, Allen et al. (2006) and Allen (2006)classify economic activities as being either based on specialized supplier relations or science-based industries. The former are associated with incremental innovation strategies, while the later rely on radical innovation. The authors find broad support for VoC’s claim of comparative advantages for all countries studied by the literature. Watson (2003) in turn, does not find support for trade flows following the pattern of institutional differences using a social-security index and a size-adjusted trade index of country pairs. Similarly, Beyer (2006) reports mixed results.
Carlin and Mayer (2003) undertake a related approach. They investigate empirically the relationship between institutions of the financial system (such as disclosure requirements), structural characteristics (such as ownership concentration), and sectoral growth and R&D investments. Sectors with a high dependence on equity and high skills grow stronger and display higher R&D expenditures with more information disclosure and higher ownership concentration. The results suggest that differences in corporate governance systems could provide comparative advantages for technologies, which require different kinds of funding and/or commitments.
2.2 Institutional arbitrage
The notion of comparative institutional advantages among nations has strong theoretical appeal. The empirical record with regard to quality differences of institutions appears to be strong, while evidence on the impact of types of institutions and systems differences are mixed. At the same time, international factor movements have increased since the 1980s with foreign direct investments playing an increasingly strong economic role. In the second half of the 1990s, FDI inflows worldwide grew annually by 40% on average. The worldwide inward stock of FDI is 23% of the World’s real GDP in 2005 (UNCTAD 2006).
Against this backdrop, the institutional arbitrage hypothesis expects investments to be shifted to those countries, where the institutional environment and corresponding comparative institutional advantage suits them best. As multinational corporations (MNCs) can now easier relocate production to other economies exploiting advantages in resource endowments, factor costs and market sizes, in addition VoC predicts that it is easier to utilize institutional differences across countries and accompanying comparative institutional advantages (Soskice 1999:118, Hall and Soskice2003:248, Hancké et al.2007). The opportunity of institutional arbitrage is also much emphasized by the strategic management literature, both in theoretical elaboration (Porter1990, 1996, Hoskisson et al. 2004) and strategic advice for companies to exploit institutional differences more consistently (Ghemawat 2003).
However, the extent of the opportunity for arbitrage crucially depends on the capacities of the local sectors. In order for the comparative advantage to play out fully, there has to be an “industrial tradition” of a sector (Resmini2000), which provides skilled labor and company networks. Ideally, sectoral data measuring for instance skill level or skill intensity would be desirable, but such data do not exist. But. a straightforward way to proxy the strength of a sector in that respect is to look at the value-added it produces. In general, a sector that produces a greater value-added can be expected to be of significant importance and therefore possess the institutional capacities needed to fully reap the benefits of a comparative institutional advantage. So, we will expect the institutional advantage to be most clearly visible in large and potent sectors rather than niche sectors. Econometrically, as will be explained in more detail below, this means introducing interaction terms between value-added and comparative advantage indices.
Accordingly, in order to test the proposition we will suppose that in addition to traditional determinants of foreign direct investment location such as distance, labor costs and market size, the institutional environment should play a decisive role in the investment decision. The next section introduces the empirical literature on the determinants of FDI.
2.2 Determinants of foreign direct investment
FDI activities can be categorized as horizontal and vertical FDI. Horizontal FDI refers to a situation when a horizontal stage of the production process is duplicated in a foreign country. Vertical FDI, in contrast, occurs when the production process is split up and a part of the production chain is entirely relocated.
In both cases, trade-offs arise, which a firm contemplating any of the two types of foreign direct investment must resolve. For horizontal investments the firm loses economies of scale on the plant level, because the singular plants become smaller. However, the firm gains better market access. In the second case of vertical foreign direct investment costs of disintegration arise through the split of the production chain. On the other hand, differences of relative factor supplies, hence cheaper factor costs in the host country, are the potential gain for this kind of investment. Form the considerations in the previous section it follows that not only differences in factor endowments but also institutional differences can be seen as determinants of (vertical) FDI. Both views can also be treated simultaneously in a unified model, in which horizontal and vertical FDI are considered special cases (Markusenand Maskus 2002).
Both theoretical and empirical problems arise when trying to disentangle country determinants for horizontal vs. vertical FDI. On the one hand exact classification of the type of investment poses problems, since the distinction is not always clear-cut. Then, theoretical predictions of the effects of country traits on either type of investment are unclear. Higher trade costs should encourage horizontal FDI while it should discourage vertical FDI. Larger markets of the host country will most certainly lead to higher horizontal investment, but it can also be shown for vertical FDI (Zhang and Markusen1999). Cheaper factor costs encourage vertical investments, but no predictions can be made for horizontal.
Apart from that it is very difficult to distinguish between horizontal and vertical types of FDI from the data. One would need to have detailed data on the destination of sales of host country affiliates, whether they are within the host country, are directed to the home country or to a third country. Since such data are not available as in our case, we concentrate on the traditional way to model country determinants using gravity equations. In general, gravity equations attempt to explain the distribution of FDI across countries mainly by factors such as distance, GDP and factor costs (see overview in Barba Narvaretti and Venables2004, chapter 6). In such a setting, researchers have to live with the fact that their data include both horizontal and vertical FDI. Traditionally, gravity models were utilized to explain trade flows and volumes; the first application to investments flows is from Eeaton and Tamura(1994).
While in trade theory a recent literature attempts to model institutional diversity of countries and accompanying comparative institutional advantages in order to explain discrepancy between theory and empirics of the traditional trade models (see section 2.1), to our knowledge this has not been done yet in the FDI literature. Studies, which take into account institutions of the host country focus on FDI flows from advanced countries to developing and transition countries (Garibaldi et al.2002, Kinoshita and Campos2003, Wei2000, Du et al. 2008). In contrast, Habib and Zurawicki(2002) and BŽnassy-QuŽrŽ et al.(2007) have a broader scope by including also bilateral flows between advanced market economies. What these studies have in common is that they focus on the quality of institutions by including World Bank governance indicators or by measuring an institutional distance of quality and do not explicitly focus on comparative advantages. As explained above, our focus is on differences of institutional types exemplified by distinctive institutional systems sewn together by institutional complementarities.
3. Methodology
We estimate the amount of German FDI as a function of a vector of explanatory variables including both country and sector characteristics. We estimate the size of German (log) FDI stocks to sector i of country j as a function of country and sector characteristics. We postulate that FDI should tend to flow to sectors enjoying a comparative advantage in the host country. The VoC-Dummy is constructed such that it takes the value 1 if the sector has a postulated comparative advantage, and 0 if not:
Table 4 shows which economic activity of the FDI classification (according to ISIC Rev. 3) we classified as either being favored by a CME and an LME, respectively. The classification follows Allen et al.’s (2006) distinction between radical innovation sectors and incremental innovation sectors and it is in line with Hall and Soskice’s (2001) results given in Table 3 for Germany and the USA.
The estimation equation then is:
As stated above, it is expected that the comparative institutional advantage should be strongest in large sectors. In order to capture that effect, VoC and value-added are introduced as interaction terms. The coefficient on value-added is assumed to be positive, since a larger sector should attract more FDI. Also, VoC is expected to enter with a positive sign.
There are two ways to interpret the interaction effect. We expect that ceteris paribus a larger sector will receive more FDI if the sector enjoys a comparative advantage. A second way to state the same thing is to expect that the size of the sector is only relevant if the advantage is actually present. As a result, we will have to interpret the marginal effects if the conditioning effect when the comparative advantage is present (VoC = 1) and when it is not (VoC = 0). This implies that the marginal effect of the size of the sector is given by if there is no comparative advantage. For a sector with an advantage (so VoC = 1) the marginal effect will be . The marginal effect of moving to a sector with a comparative advantage is given by as it depends on the size of the sector.
Table 4
Incremental innovation
Radical innovation
ISIC Rev. 3 Code
Industry
ISIC Rev. 3 Code
Industry
D28
Fabricated metal products, except machinery and equipment
D24
Chemicals and chemical products
D29
Machinery and equipment, n.e.c.
D33
Medical, precision and optical instruments, watches and clocks
D30
Office, accounting and computing machinery
D353
Aircraft and spacecraft
D31
Electrical machinery and apparatus, n.e.c
J65
Financial intermediation except insurance and pension funding
D32
Radio, television and communication equipment
J 66
Insurance and pension funding, except compulsory social security
D 34
Motor vehicles, trailers and semi-trailers
J 67
Activities related to financial intermediation
D 35
Other transport equipment
K 72
Computer and related activities
D 359
Railroad equipment and transport equipment n.e.c.
K 73
Research and development
D 36
Manufacturing n.e.c
Note: Based on Allen et al. (2006)
The vector gravity includes the distance to Germany, GDP, GDP per capita and unit labor costs as controlling variables. The absolute GDP measure the overall market size of the host country and is expexted to have a positive sign. In order to proxy for overall institutional capacities and infrastructure, we include GDP per capita of the host country, which is also expected to enter positively. Unit labor cost differences point at cost considerations of production and is expected to be negative. We include a vector Z of industry dummies. All parameters except for the VoC indicator are included as logarithms. In addition to that, in order to minimize possible multicollinearity, value-added and the VoC indicator are mean-centered.
4. Data
Table 5 provides some summary statistics of the sample. Panel A of the table shows overall summary statistics of the dependent and independent variables. Panel B shows the countries in the sample and the number of sectors per country that are classified as having a comparative institutional advantage in this country and the number of those that do not. Our dependent variable represents German outward direct FDI stocks in Euro. The industry-level data by host country are taken from the Deutsche Bundesbank Micro database Direct Investment (Lipponer 2003). We aggregate firm-level panel data over the period 1996 to 2001. This transformation into cross-sectional data is justified because we are examining systemic aspects of the allocation of German FDI. We thus eliminate time-variation in the data.
The Bundesbank measure for FDI differs slightly from the one used by the OECD and the IMF. The difference between the two measures is that the Bundesbank measure “excludes loans to shareholders, affiliated enterprises and enterprises linked with the party required to report through participating interests; and claims on shareholders, affiliated enterprises and enterprises linked with the party required to report through participating interests” (reverse loan capital; see Lipponer 2003: 19). For robustness, we estimate all equations using both measures. The results do not differ qualitatively, which is why we only report estimations with the Bundesbank measure. The main explanatory variable of interest – the VoC indicator – will be operationalized using different measures. The base estimation includes a binary variable as explained above, where the sector either takes the value one, when it is in a country, whose type of capitalism ought to facilitate the sector specific innovation pattern, or zero if otherwise. In the next section we introduce further measures of system differences as robustness checks.
GDP per capita is the mean over the period 1996 to 2001 from World Bank (2005). Mean unit labor costs were calculated as labor compensation over value-added, both from the STAN database for Industrial Analysis (OECD). Data on geographical distances to Germany are from Mayer and Zignago (2006).
Table 5
A
Observations
Mean
Std. Dev.
Min
Max
FDI
723
8.05
1.86
0.46
13.07
FDI if VoC = 1
311
8.24
1.72
2.11
13.07
FDI if VoC = 0
412
7.90
1.94
0.46
12.51
Value-added
618
8.99
1.63
3.06
13.74
Distance
754
6.83
1.21
5.16
9.84
GDP
754
27.01
1.24
24.82
29.88
GDP p.c.
701
10.11
0.28
9.26
10.62
ULC
576
0.45
0.13
0.02
.70
All variables are in logarithms
B
Number of sectors if VoC = 0
Number of sectors if VoC = 1
Total
Australia
4
5
9
Austria
32
17
49
Belgium
46
31
77
Canada
12
18
30
Denmark
30
14
44
Finland
15
10
25
France
36
16
52
Greece
4
1
5
Ireland
5
19
24
Italy
22
13
35
Japan
23
13
36
Korea
5
6
11
Netherlands
39
15
54
New Zealand
3
3
Norway
14
11
25
Portugal
3
3
6
Spain
20
12
32
Sweden
26
14
40
Switzerland
36
16
52
UK
21
40
61
USA
19
34
53
Total
412
311
723
5. Results
The results of the estimation using OLS are given in Table 6. Columns 1-3 give the results of different specifications using the full sample. Columns 4-6 show results for a smaller sample excluding observations for FDI to Spain, Portugal and Greece. The remaining columns 7-9 list the results, which are obtained including only the “pure” LMEs and CMEs (see Table 2).
In all models, the market size measured by the absolute GDP shows the expected positive sign. The effect is the largest of all independent variables: for a 1%-increase in GDP, FDI into a sector increases by approximately 0.5%. The distance to Germany also has the expected negative sign. Both are highly significant throughout all models. The unit labor costs in a given sector show a small negative coefficient, but it is never significantly different from zero, nor is the coefficient for GDP per capita. Both can be explained by the fact that the countries in the sample are all comparatively rich countries, which do not differ that greatly in overall institutional capacity (proxied by GDP per capita) and productivity (proxied by unit labor cost). We observe that the size of the receiving sector proxied by the value-added of the sector has the expected sign, but it is not significant when entered in models 2, 5 and 8, respectively. The interpretation in models 3, 6 and 9, however, must take into account the introduction of the interaction effect.
The marginal effect of the size of the sector on FDI is given by the coefficient on value-added alone only if VoC is equal to zero. Hence, we can already conclude that the size of the sector does not seem to play a role if the sector does not enjoy a comparative advantage: the coefficient is never significantly different from zero. It does not make much sense to interpret the coefficient on VoC in isolation in models 3, 6 and 9, because that would mean that value-added were zero. For the sake of argument, however, we see that the comparative advantage does not seem to influence the decision to invest in a particular sector, if that sector is small.
The interesting interpretation comes from looking at the joint effect. The coefficient of the interaction effect is positive and significant. It is larger and attains a higher significance level when Spain, Portugal and Greece are excluded, which to a certain extent fulfill expectations we had about those countries. The fact that none of the coefficients of the other variables change greatly after the introduction of the interaction term, lends belief to the claim that this is a genuine effect not driven by possible multicollinearity or by a large significant effect by one influencing variable that would overlie the effect of the other.
Table 6
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
full sample
excluding Spain, Portugal and Greece
classical LMEs and CMEs only
Value-added
0.10
(0.09)
0.11
(0.09)
0.09
(0.09)
0.10
(0.09)
0.07
(0.10)
0.07
(0.10)
VoC
0.01
(0.01)
0.01
(0.01)
0.00
(0.01)
0.00
(0.01)
0.00
(0.01)
-0.00
(0.01)
0.00
(0.01)
0.00
(0.01)
-0.00
(0.01)
VoC Ã- VA
0.19*
(0.10)
0.23**
(0.10)
0.20*
(0.11)
Marginal effect of VA if VoC=1
0.30**
(0.13)
0.33***
(0.13)
0.27*
(0.14)
GDP
0.67***
(0.14)
0.54***
(0.17)
0.50***
(0.17)
0.64***
(0.14)
0.54***
(0.17)
0.49***
(0.17)
0.69***
(0.16)
0.63***
(0.18)
0.57***
(0.18)
Distance
-0.15***
(0.05)
-0.16***
(0.05)
-0.17***
(0.05)
-0.14***
(0.05)
-0.15***
(0.05)
-0.16***
(0.05)
-0.16***
(0.06)
-0.17***
(0.06)
-0.17***
(0.06)
ULC
-0.02
(0.06)
-0.02
(0.06)
-0.02
(0.06)
-0.03
(0.06)
-0.03
(0.06)
-0.03
(0.06)
-0.01
(0.07)
-0.01
(0.07)
-0.01
(0.07)
GDP p.c.
0.10
(0.25)
0.01
(0.26)
-0.01
(0.26)
0.06
(0.34)
-0.02
(0.36)
-0.03
(0.35)
0.13
(0.41)
-0.01
(0.47)
0.05
(0.47)
Constant
0.14
(0.24)
0.36
(0.31)
0.42
(0.30)
0.20
(0.33)
0.39
(0.40)
0.46
(0.39)
0.09
(0.37)
0.30
(0.48)
0.31
(0.47)
Obs.
554
554
554
511
511
511
428
428
428
R²
0.39
0.39
0.40
0.40
0.41
0.41
0.41
0.41
0.41
F
11.17***
11.22***
11.85***
10.69***
10.73***
11.55***
16.77***
15.97***
14.51***
Robust standard errors in parentheses.
*** p<0.01, ** p<0.05, * p<0.1
Note: All models include industry dummies. All variables except VoC are in log form and have been normalized between 0 and 1. Value-added and VoC have been mean-centered.
Table 6 reports the coefficient of the joint effect if VoC=1 and corresponding standard errors. It is significant in models 3, 6 and 9. Accounting for comparative advantage, the size of the sector matters: ceteris paribus, for a 10% increase in value-added a sector receives about 3% more FDI if the sector has a comparative advantage. If we take the coefficients at face value, this effect is smaller than the effect of the total market size, but it is larger in absolute terms than the effect of the distance of a country.
A similar interpretation can be put forward looking at the marginal effect of comparative advantage. Figure 1 shows the marginal effect of a sector with a comparative advantage, which is conditioned by the size of the sector. The left-hand diagram shows the effect from model 3, which uses the entire sample, while the right-hand side reports results from model 6. The dashed line in both graphs show the upper and lower bounds of a 95%-confidence interval. The graph for model 9 is similar to the one shown for model 3.
It can be seen from the graphs in Figure 1 that the effect of the comparative advantage increases as the sector size increases and that it is not significant for small sectors. So for investments in relatively small sectors the advantage does not matter. A careful comparison of the two graphs reveals that the effect “frays” less for the sample excluding Spain, Portugal and Greece as the value-added increases, which is reflected in the higher significance in Table 6.
Taken together, the data suggest that German multinationals do follow the comparative advantage of a receiving sector when making investment decisions. This depends crucially on the size of a sector. However, the results also suggest that for medium sized sectors this effect does not trump “traditional” gravity-equation explanatory variables such as market size and the distance.
Panel A
Panel B
Figure 1
To check robustness, we use alternative indices of differences in institutions between host and home country. As opposed to the binary indicator that takes values 0 and 1, these are indicators that vary between 0 and 1. Hence, these indicators can capture different degrees of suitability of institutions. Table 7 shows for each of the additional indices the results for the full sample excluding Greece, Portugal and Spain and the core sample of LMEs and CMEs. In other words, Table 7 shows the models that correspond to models 6 and 9 in Table 6.
First, shown in columns 1 and 2 of Table 7, we use a measure constructed from an updated coordination index following Hall and Gingerich (2004). It is a composite index including indicators for the institutional spheres of industrial relations and corporate governance. The index is the outcome of a factor analysis using measures of centralization of wage bargaining, job tenure, the size of stock market, dispersion of control and shareholder power [] . In columns 3 and 4, we base our measure on the coordination index provided by Knell and Srholec (2007), which is also the outcome of a factor analysis, but with a slightly different focus than Hall and Gingerich’s. Their coordination index is comprised of indicators from the spheres of social systems, labor markets and business regulation. For the two composite indices the results are fairly similar. In models 2 and 4 the sign of the interaction effect is not significant anymore, but the joint effect reaches some significance for model 2, while it is not significant in model 4. The economic effect here is somewhat smaller than in the models using the binary indicator.
As a last step we undertake a robustness test by using institutional indicators for single spheres of the economy instead of global systems indicators. Hence, in order to check whether there are any differences when looking at complementarity “unbundled”, we split up our updated coordination index on the basis of Hall and Gingerich (2004) into its two elements: industrial relations (models 5 and 6) and corporate governance (models 7 and 8). Also, in order to check whether labor market regulation alone might account for the sectoral distribution of foreign direct investment, we include Botero et al.’s (2004) labor law index, which is shown in columns 9 and 10. Unbundling the composite index shows that industrial relations and corporate governance indicators have a higher impact on the marginal effect of sector size than the composite indices.
Figure 2
In addition this effect stays significant, albeit economically smaller, when the mixed market economies are excluded. The highest effect seems to stem from labor laws. For a 10% increase in value-added of a sector, FDI into this sector increases by roughly 3 – 4% depending on the chosen sample. This is effect holds only when the labor laws in the country fully correspond to the claimed comparative advantage.
Since the indicator is not a binary one, but takes values between 0 and 1, we can visualize the effect of intermediate values in Figure 2. It shows the marginal effect of value-added (and corresponding upper and lower confidence bounds) for different values of the labor law index. A straightforward interpretation of the horizontal axis is that as one moves to the right the correspondence with postulated comparative advantages increases. As expected, the marginal sectoral size effect increases in correspondence.
6. Conclusion
The analysis in the paper suggests that there are institutional arbitrage arguments involved in the locational investment decision of firms. Controlling for “traditional” gravity-equation determinants of FDI on both country and sector level, we find that a sector, which in a given country enjoys a comparative advantage due to the institutional surroundings receives a larger chunk of German FDI than a sector without this advantage. This effect depends on the size of the sector: if the comparative advantage is present, a larger sector receives more FDI. On the other hand, the size of a sector is shown to be irrelevant if there is no comparative advantage. This effect has been shown to be robust to several specifications and a considerable number of operationalizations of institutional advantage. Of course, a wider cross-country set of FDI data involving more home countries than only Germany seems desirable. In general, a sample of bilateral FDI stocks with detailed information on the actual activity undertaken, such as subsidiary sales, could help to better identify multinational strategies. In such a setting, it would be possible to disentangle horizontal and vertical FDI along the lines of the model by Markusen and Maskus (2002). To our knowledge, such data of investment do not exist on a detailed sectoral level. However, that fact that companies from a sample of firms from the biggest European economy do seem to engage partly in institutional arbitrage makes a fairly strong point.
The comparative institutional economics and capitalism literature can help to explain how specific industries or production methods are more likely to be found in some countries rather than others. Taken one step further, we show that by way of institutional arbitrage the effect translates into international investment decisions as well. In a broader context still, the institutional arbitrage argument lies at the heart of a notion of persistent divergence of institutional systems. That is, if globalization makes international capital movements easier, institutional arbitrage should cement institutional differences and not erode them, because country competition for capital investments would not occur in such a scenario.
Order Now