Institutional Dynamics in the Global FDI Network : Examining The Co-evolution of Institutions and FDI with Stochastic Actor-Oriented Modelling

(1)

Linköping University | Department of Management and Engineering Master’s Thesis in Economics Master’s programme in Economics Spring 2021 | LIU-IEI-FIL-A--21/03677--SE

Institutional Dynamics in the Global

FDI Network

Examining the Co-Evolution of Institutions and FDI with

Stochastic Actor-Oriented Modelling

Axel Norgren

Martin Olsson

Supervisor: Gazi Salah Uddin

Linköping University SE–581 83 Linköping +46 13 28 10 00, www.liu.se

(2)

Acknowledgements

We would like to direct a special thanks to our supervisor Gazi Salah Uddin that have offered us valuable advise and assistance. Furthermore, we wish to thank Tom Snijders for useful methodological and theoretical insights. Finally, we would also want to extent our gratitude to everyone that has read and reviewed this thesis, your input was much appreciated.

(3)

Abstract

This thesis addresses the relationship between institutions and foreign direct investments (FDI). While the issue of how institutions attract FDI (selection) is quite well-researched, the empirical evidence for institutions spreading through FDI (influence) is more ambiguous. We argue that past studies have neglected issues of endogeneity and interdependence in their modelling. We amend these issues by using a Stochastic Actor-Oriented network model which allows for interdependent and endogenous processes. The thesis also addresses the mechanisms governing the general relation between FDI and institutions and what these can tell us about institutional change and the process of globalisation. The model provides no evidence that FDI helps to spread institutions from home to host countries, but it does provide evidence that the selection effect can be an important dynamic between FDI and a certain set of institutions. Finally, we argue that FDI does not seem to be a contributory factor to institutional convergence.

Keywords: Institutions, Foreign Direct Investment, Economic development, Social network analysis, Stochastic Actor-Oriented Modelling, Institutional Convergence.

(4)

1. Introduction

At the heart of globalisation lies economic motives, plausibly the main driver for all inter-national exchange. Next to trade, foreign direct investment (FDI) can be seen as one of the most important transnational economic interactions. The global FDI stock has grown substantially in the world over the last couple of decades from 24 per cent of world GDP in 2005 to 42 per cent of world GDP in 2019 (OECD, 2021). This growth in FDI is primarily driven by developing countries that are increasing their share in both inward and outward FDI, capturing approximately 40-50 per cent of inward FDI flows the latest decade while also representing 30-40 per cent of the outwards FDI flows (Carril-Caccia, Pavlova, et al., 2018). Due to the sharp decline in FDI during the Covid-19 pandemic in 2020, developing countries now capture 70 per cent of all FDI inflows, further magnifying the importance of FDI in developing countries as well as the importance of developing countries in the global markets (UNCTAD, 2021).

The long-run economic benefits of attracting FDI is well known, but the liberalisation of international capital also have strong opponents from several political spheres. Critics argue that the race between countries to attract FDI results in a downward spiral of regulatory quality, a ”race to the bottom” (Karlewicz, 2020; Lawder, 2021). On the contrary, scholars debate this one-dimensional description of the dynamics between FDI and institutions; some even argue for a ”race to the top”, where the two help to promote each other (Ali et al., 2011).

Sometimes, efforts are also made to deliberately spread certain institutions through FDI. The Chinese government, alongside other authoritarian regimes, are being accused of trying to shape the world through investments and aid (Naim, 2009). Another example is the efforts of the European Bank for Reconstruction and Development (EBRD) to promote systems of market economy in former Soviet states through investments (Kilpatrick, 2020). However, institutions are not always spread deliberately, FDI with strictly financial motives may also have the potential to spread institutions.

Both North (1993) and Bush (1987) describe institutional change as incremental, driven by new knowledge and changing perceptions of the world. FDI can be seen as a carrier of

(7)

tech-nology and human capital from one country to another (see e.g. Blomström and Kokko, 1998; Veugelers and Cassiman, 2004). Therefore, it is not unreasonable to theorise that FDI might give rise to new opportunities and incentive structures, which in turn leads to insti-tutional change. The importance of the instiinsti-tutional environment for economic performance should not be underestimated either; North (1990, 1997) explain that institutions are key in fostering economic growth. Similarly to North, Jones and Romer (2010) attribute institu-tional differences as one of the main explanations for differences in growth between countries and argue that bad institutions are detrimental to total factor productivity, hindering the adoption of technology and knowledge.

Since FDI is a carrier of new ideas and technology, institutions can theoretically be thought of as a moderator between FDI and growth. Endogenous growth theory depicts the growth process as driven by global economic integration, and institutions as the moderator between economic exchange and growth. Thus, there are strong economic incentives for institutional convergence as well. Fukuyama (1989) predicted an ”end of history” where western liberal democracy is the final form of government, implying a non-reversing process of institutional convergence. Since FDI is one of the leading drivers of economic globalisation, it might also be crucial in shaping the global institutional framework.

Empirically it has been well established that institutions are an important determinant for FDI (Knack & Keefer, 1995; Campos & Kinoshita, 2003; Buchanan et al., 2012). However, some scholars suggest that a reversal of this relationship exists (Kwok & Tadesse, 2006), but the empirical findings for FDI having an impact on the host countries’ institutions are inconclusive (Malesky, 2009; Ali et al., 2011; Demir, 2016). A bidirectional relationship also implies methodological challenges and raise concerns of endogeneity in the econometric models. Another methodological issue comes in the form of network effects. Granovetter (1985) point out the importance of embeddedness when analysing economic networks, that an economic actor’s position in a network influence its behaviour. While network modelling is commonly used in sociological research, where it is used to study complex interdependent social relationships, it is still a relatively new approach in economics. Some efforts to study FDI networks have been made, both Koskinen et al. (2015) and Schoeneman et al. (2020) show that global FDI network formation is driven by interdependent processes, with indirect ties being important for the network formation. But no studies have, to our knowledge, investigated the link between FDI and institutions with a network approach. Thus, there is a gap in the literature on FDI networks and institutional change. We will try to bridge this gap by modelling global FDI as a network, letting FDI represent network ties between nodes represented by individual countries. The network approach allows us to study the co-evolution of FDI and institutions and separate the effects into a selection and an influence effect, similarly to how Steglich et al. (2010) study peer influence and homophily in relation

(8)

to substance use among teenagers.

Another factor that complicates the usage of the traditional econometric models is the com-plex time order of cause and effect between FDI and institutions. We raise the question in this thesis that the time order between the two variables possibly makes causality an unsuit-able concept to understand the relationship. Furthermore, the theoretical mechanisms that may explain institutional change induced by FDI are best described through institutional isomorphism (DiMaggio & Powell, 1983; Beckert, 2010). This theoretical framework plots the course for an even more interesting question than simply if and how FDI in general affect host country institutions, but rather if and how countries can spread their own institutions to host countries through FDI.

The inconclusive literature and the methodological concerns cloud our understanding of the impact that FDI can have on host country institutions and by extent the relationship between FDI and institutions in general. If these challenges could be met, it is possible to derive the more technical mechanisms governing this relationship which would help us understand the dynamics of globalisation and institutional change.

Primarily, the aim of this thesis is to explore how institutions spread through FDI, using a methodology that has the potential to mitigate possible endogeneity concerns in previous literature. Additionally, we aim to explore the underlying mechanisms of the relationship between FDI and institutions to generate insights into the dynamics of globalisation and institutional change. The primary research question will be stated as the following:

Are selection and influence processes important for the relationship between FDI and institu-tions?

The research question is divided into two hypotheses regarding selection and influence re-spectively, and answered through the use of Stochastic Actor-Oriented Modelling, SAOM (Snijders et al., 2010). Simulating the network formation process allows us to examine the complex and simultaneously evolving relationship between FDI and institutions, separating selection and influence processes from each other. The network model compensates econo-metric modelling where the econoecono-metric models lack functionality i.e. endogeneity, embed-dedness and interpretation. Beyond understanding the importance of selection and influence respectively we will, through the influence effect, get an understanding of how the possi-ble institutional diffusion through FDI contributes to the theorised process of institutional convergence.

With this thesis, we contribute to the literature in several ways. We bring further insights into the ambiguous role FDI have for institutional change by specifically examining if FDI spread institutions and we introduce a method relatively new to the economic discipline to examine this. To the best of our knowledge, no previous research has used network analysis

(9)

in general and SAOM in particular to examine this relationship. We also contribute to the literature by exploring some of the mechanisms governing the general relationship between FDI and institutions, answering if selection and influence matter for the relationship between FDI and institutions. Finally, we provide insights on the dynamics of globalisation and institutional change, discussing if FDI contributes to institutional convergence and ”the race to the bottom”.

The findings from our models indicate that selection seems to be the primary mechanism governing the relationship between FDI and institutions. But only certain types of insti-tutions attract FDI, such as corruption, while others such as economic freedom might even deter it. We also find that FDI does not seem to help spread institutions from one country to another. Since we utilise a methodology that circumvents issues of interdependence and endogeneity, we question previous findings on the subject that have found results in support of FDI affecting host country institutions. Finally, as our results indicate that there is no influence effect we do not find it probable that FDI can be a possible contributory cause to the theorised process of institutional convergence nor that ”the race to the bottom” is driven by FDI.

The thesis is laid out as follows: Chapter 2 outlines the theoretical framework for the thesis, introducing the theories of institutional change, the channels of institutional isomorphism, and the process of institutional convergence. Chapter 3 gives a brief introduction to concepts of social network analysis that we deem relevant for this thesis. Chapter 4 reviews previous literature on the issue, mainly in regard to institutions attracting FDI and the opposite of how FDI affects host country institutions, but also covering the extent of FDI network literature. Chapter 5 states and motivates our main hypotheses. Chapter 6 describes why previous econometric approaches might be flawed, describes the data we use, and outlines the network modelling approach and SAOM specification. Results of the analyses are presented and analysed together with a methodological discussion in chapter 7.

(10)

2. Theory

As a foundation for exploring the relationship between FDI and institutional change, we base the theoretical framework on theories on institutions and institutional change combined with endogenous growth theory, organisational theory, and network theory. We separate the relationship between FDI and institutions into two main directions.

The impact institutions may have on FDI can in a network setting be thought of as a selection effect, i.e nodes connect with nodes that have similar/certain attributes. We mainly base the explanation for this direction on transaction cost theory derived from Williamson (1989), assuming that investors will try to minimise the costs associated with the investment, consequently choosing to invest in countries with beneficial institutional environments. This creates a selection mechanism where the institutional quality of a country determines the level of inward FDI.

The second effect can instead be interpreted as an influence effect in a network, where the ”ties” of FDI influence host country institutions with home country institutions. For this direction, we follow the framework set up by DiMaggio and Powell (1983) and Beckert (2010) and characterise four channels of transmission through which FDI can affect institutions: competition, bargaining power, professionalisation, and mimesis. These four channels create the influence mechanism where FDI may spread institutions from home to host countries. Finally, we relate the relationship between institutional change and FDI to Fukuyama (1989) and discuss his predictions of global institutional convergence and how it relates to the previous theory.

2.1 Institutional change

Institutions can be understood as humanly derived constraints that shape both general and economic interactions, i.e ”the rules of the game” as famously stated by North (1990). North distinguish between formal and informal institutions where informal institutions can be cul-ture and norms while formal rules are the governing laws and regulations in a society. This

(11)

can also be described with the framework of transaction cost economics (Williamson, 1989), where uncertainty comes as cost and the institutions are a way of organising the interaction to reduce uncertainty and the cost of transaction.

The perceptions of the process of institutional change are diverse. North (1994, 1995) describe it as the result of a struggle between agents with different incentives. He describes the process as incremental where the bargaining power of the agents determines the outcome. The competition between the actors will force them to continually invest in knowledge which in turn might change their perception of the world, and by extent their actions. North is grounded in theories of new institutional economics and his emphasis is on the role of informal institutions, i.e. the perceptions and incentives of the actors. In his later work, he became more nuanced and emphasised the uncertainty of institutional change and described the world as non-ergodic due to the increasing interdependence and complexity in the global economy (North, 2005).

Bush (1987), on the other hand, emanates his reasoning from the school of old institutional economics and argue for a more structural perspective, where the stock of knowledge is a central component. He sets up a framework that is based on a conflict between instrumental and ceremonial values in society. Similarly to North’s theories, Bush describe how new challenges and technology necessitates a change in the behaviours and practices of economic actors to adapt to these new conditions, giving rise to a change in the instrumental values. Ceremonial values, on the other hand, originate from the desire to maintain and create hierarchies. According to Bush, the stock of knowledge and the level of ceremonial dominance determine to which extent instrumental values can change the institutions in relation to the ceremonial values. Powerful actors can hold back instrumental institutional change in a society with strong ceremonial dominance, this is similar to how North (1993, 2005) argue that institutions reflect the bargaining power of the players, implying that institutional change can be held back by influential actors who profit from the prevailing institutional setting. Therefore, a powerful foreign investor may come with a lot of bargaining power and may be able to shape institutions for the investors own benefit. Bush’s perception of institutional change emphasises an evolutionary process, with less focus on individual actions and more on the values in society. Bush and North are useful to describe a fundamental conflict between evolution and design in the theory of institutional change and can describe how both conscious and unconscious institutional change may occur through FDI.

(12)

2.2 Institutions impact on FDI

The theoretical case for why institutions can attract or deter FDI is most easily understood by the transaction cost approach by Williamson (1989), where a well-functioning institu-tional arrangement reduces the cost of the interaction. For example, a multinainstitu-tional firm might perceive a country with strong intellectual property rights as a less costly investment compared to a country with weaker property rights given the risk of intellectual property theft (Sala-i-Martin & Barro, 1995). Another example focusing more on informal institu-tions can be that a culturally distant country might deter an investor due to barriers such as local customs and different business ethics which the investor perceives as a transaction cost (Bailey, 2018). Put simplistically it can be said that the investors want to avoid uncertainty when investing, and a certain set of institutions will be perceived as a less costly option given that they reduce this uncertainty.

If the institutions are thought of as attributes and the FDI as connections, the causality running from institutions to FDI inflow can theoretically be interpreted as the selection effect. What is interesting with this theoretical lens is that institutions do not only shape the economic actions but also that they subvert the direction of the economic incentives, meaning that host countries have an incentive to reduce their transaction costs in order to attract investors. As an example, a German company might prefer to invest in Estonia rather than Bosnia and Herzegovina simply because the property rights in Estonia are stronger and more similar to those in Germany than those in Bosnia and Herzegovina. This creates an incentive for Bosnia and Herzegovina to improve the strength in property rights to the level of Germany and Estonia in order to attract investments. This mechanism will be touched upon in the next section under the transmission channel of competition but also in the channels of professionalisation and mimesis.

2.3 Four channels of influence

To understand how FDI may impact institutions, we complement theories of institutional change derived from institutional economics with theories from organisational theory. In-spired by Beckert (2010), we present four theoretical transmission channels through which FDI can impact institutions. The first channel is a channel of competitive pressure related to the transaction cost approach. The other three channels: bargaining power, mimesis and professionalisation, are the classic isomorphic processes presented by DiMaggio and Powell (1983). We also connect these transmission channels to the theories of institutional change in order to highlight how the channels relate to the evolution and design paradigms respectively.

(13)

We also relate them to if the influence effect is contingent or non-contingent on the selection effect in the network setting. Finally, it is important to note that this typology is not a strict law of nature but rather a convenient way of structuring the theory in a way that is useful for our analysis. In reality, the boundaries of these four channels are more diffuse.

2.3.1 Competition

The first transmission channel of influence stems partly from the works of Weber (1920) who laid the foundation for a theory of increasing bureaucratisation in society due to a ratio-nalistic spirit. Weber’s theories have long dominated the field of organisational theory with the competitive pressures of capitalism as the main driver of organisational change. This could also be understood through the framework of transaction cost economics where the competitive pressure of having the ”best” institutional structure, i.e. the lowest transaction costs, create strong incentives for institutional change. This is fundamental for the race to the bottom hypothesis that argues that competitive pressure forces countries to abolish eco-nomic and environmental regulation, lower capital taxes and provide flexible labour (Beckert, 2010). The same competitive pressure could theoretically also create incentives for stronger property rights, less corruption and more democratic institutions as it is empirically proven that these factors attract investors (Bailey, 2018). However, Beckert (2010) argue that this pressure might not naturally lead to increasing institutional convergence and that under some conditions, it can increase divergence through institutional specialisation.

A premise of this channel of influence is that actors can identify the institutions together with their cost structures and have the ability to change them when necessary (Beckert, 2010). This might be somewhat unbelievable, not only due to collective action issues, but also because it assumes full rationality. If this is the case, it could be a strong argument for the conscious design theory of institutional change. An alternative premise for this channel is to look at it through an evolutionary lens, where the theory of natural selection can explain the institutional adaption the actors make, as discussed by Beckert (2010). This would give substance to Bush (1987) theory of institutional change where the structural competitive forces give rise to winning and losing organisational ideas. The two premises are not mutually exclusive and in reality there is likely both a survival of the fittest dimension to the competitive market and rational agents that can study and change institutional structures to their advantage.

In a network setting, the competition channel represents a part of the influence effect just like the other channels. However, the influence effect could be argued to take place prior to the selection, i.e. the individuals ”get dressed up before the dance” to increase the chance of connection. The selection effect creates competitive pressure because investors will choose

(14)

the countries with beneficial institutions and low transaction costs over countries with dis-advantageous institutions. This creates incentives for countries to adapt their institutions to become more similar to the institutions from the ”hand that feeds them”, or to find an institutional niche that can attract investment. Therefore, the influence effect in this channel should conceptually be thought of as contingent on the selection effect.

2.3.2 Bargaining power

Following the notion set forth by North (1993, 2005), that institutions mirror the bargaining power of the players, a clear channel for institutional change through FDI emerge. Pow-erful actors entering a new market are likely to come with a lot of bargaining power, and thus they might have the ability to induce institutional change in their favour. Ali et al. (2011) argue, based on North (1981), that FDI flows can come with substantial tax revenues for the host country government, thus governments might be incentivised to cater for for-eign investors in order to accumulate revenues and strengthen their standing. With enough bargaining power, an actor can improve or worsen the institutional quality, fuelling either convergence or divergence depending on the actor’s incentives. As an example, Bues and Theesfeld (2012) describe a situation where foreign investors have used their influence to gain advantages over the domestic population. Studying a small-scale irrigation scheme in Ethiopia, the authors found that foreign investment farms managed to overtake the rights to a water resource through their extensive bargaining power. DiMaggio and Powell (1983) define a process of coercive isomorphism, where institutional change stems from formal and informal inter-organisational influences. DiMaggio and Powell (1983) mainly discuss how multinational corporations (MNC) home country institutions have a profound effect on the firm’s organisational structure, which in turn are spread to subsidiary firms in the host coun-try through requirements imposed by the MNC. Similarly, the MNC might be reluctant to pay out bribes since the practice is not in compliance with home country formal or informal institutions, effectively making corrupt practices such as bribery less viable in the host coun-try. This creates a process of institutional convergence between the home and host country, with home country institutions spreading through the MNC’s FDI.

Beckert (2010) discuss coercive isomorphism and explain that if existing institutions are discredited, the new institutional design is likely to be shaped by the most powerful actors. However, Beckert (2010) also argue against the coercive channel as purely a mechanism of convergence. He argues that if those in power have no interest in institutional homogenisation, coercive pressure may just as well lead to divergence instead. As an example, Beckert brings up colonialists that implement a different institutional setting in their colonies than they have at home. What determines convergence or divergence in institutions when it comes to a

(15)

powerful actor and its subject? Using the transaction cost approach, it might be interpreted as that the actor chooses the least costly approach. If there are large gains and low costs of forcing institutional structures on a country, the actor will do it, otherwise, the actor might refrain from it. This is similar to Acemoglu et al. (2001) findings where the settler mortality in colonies, i.e. the cost of settling, is found to be a determinant of institutional structure. If the settler mortality is low, the chances of a coloniser implementing a more similar institutional structure to the one it has at home increases. The actions outlined in this channel are most likely the result of rational calculations, therefore this channel lies closer to North’s theories of institutional change.

2.3.3 Professionalisation

Another channel is the professionalisation channel, DiMaggio and Powell (1983) describe the drivers of this process as two-fold. One part of it is the increasing legitimacy of formal edu-cation as a cognitive base in business practices, and the other is the growth of ”professional networks”. The first driver creates pressure on countries to adopt an institutional structure that can foster formal education since it becomes increasingly important for business prac-tices. This because universities conduct important organisational research and train future business talent. There will also be a feedback mechanism back to institutions from the in-creasing recognition of academia since the cognitive frames of the individuals that attend a university tend to become more homogeneous (DiMaggio & Powell, 1983). This homogenisa-tion of mental frames will affect the instituhomogenisa-tional environment in the long term since formal institutions are derived from our mental constructs (North, 1990). The second driver is sim-ply the influence effect within professional networks, i.e homogenisation among peers. These two drivers combine into a sort of normative process that arises due to gains of scale and homogenisation in a globalised market economy. It can be thought of as a process of eroding differences. This normative process will affect institutions the way North (1990) describes it, with institutions being the construct of underlying cognitive scripts.

Even though this channel’s theoretical underpinning imply homogenisation, Beckert (2010) argue that it can imply divergence among institutional structures. This through the idea that previous differences in institutional arrangements can enhance divergence if they need complementary institutions to the institutional framework they already endowed with. Even though we borrow the four typologies from Beckert (2010), we choose to depart from this reasoning and characterise them slightly differently. Instead, we define the divergence effect stemming from different countries emulating different institutions due to cognitive or cultural differences in the mimesis channel rather than in the professionalisation channel. We believe that the professionalisation channel implies convergence since the assumption is that the pull

(16)

factors stem from global integration. Also, the ideas of institutional complementarities make more intuitive sense in the mimesis channel, which we will touch upon in the next section. In the light of North (1994) and Bush (1987) the professionalisation channel does seem to fit better with the evolutionary perspective. The homogenisation due to normative pressure and the institutional change that follows does not appear as a deliberate consequence of an agent’s plan, but rather as an unintended consequence from the victory of instrumental values, i.e when formal education becomes a cognitive base and professionalisation through peers occur.

In a network setting, this can be understood as partly contingent on the selection effect. This since the pull factors that create the normative pressures are to attract FDI. However, it is also to integrate more smoothly with the foreign investors as well and reap the rewards from already existing investment in the form of positive spillover effects.

2.3.4 Mimesis

DiMaggio and Powell (1983) describe mimetic isomorphism as a process stemming from uncertainty. Unable to come up with satisfactory solutions themselves, perhaps even failing to observe the problem, organisations may look towards more successful organisations and mimic their practices. Thus, the incentive for mimicking can be as simple as observing a more successful organisation doing something differently, why not do the same then? Furthermore, if FDI brings new technology and knowledge to the country, the simple mechanics of a competitive market might be enough to not only incentivise mimicking, it might even create a situation where domestic firms are forced to mimic to survive.

Interestingly, as DiMaggio and Powell (1983) discuss, citing Alchian (1950), mistakes in attempted imitations may lead to new innovations. Veblen (1898) describes a ”cumulative causality”, arguing for an evolutionary perspective on economics and that innovations are driven by constantly building upon previous innovations. Citing Veblen, Bush (1987) explains how this implies that ”the solution to one problem creates a whole host of new problems”. This cumulative causality can be thought of as the stock of knowledge. Consequently, Bush see the stock of knowledge not only as solutions to problems but also as something that helps a society to identify new problems. The host country already has a stock of knowledge prior to the FDI influx, likely with vastly different knowledge compared to that in the home country. In the process of trying to merge this knowledge with the FDI spillover effects, the stock of knowledge might evolve very differently in the host country compared to how it has evolved in the home country, identifying different problems, and thus creating innovations. In this view, the ”unwitting” innovations, as Alchian (1950) and DiMaggio and Powell (1983)

(17)

see them, may not be unintentional in all cases. Instead, they reflect solutions to problems identified by a dissimilar stock of knowledge.

Thus, it is likely that knowledge spillovers from FDI can create institutional change, but it might be naive to expect that they will induce strict institutional convergence. Diver-gence is also likely, given different perceptions of how to solve a problem or, as mentioned in the professionalisation channel, due to historical path dependence that constrains the al-ternative solutions. However, if the theories of old institutional economics, as described by Bush (1987), are valid, FDI spillovers that increase the stock of knowledge will still lead to instrumental institutional change. If examining aggregate institutional variables that are universally instrumental to the business environment, knowledge spillovers are most likely going to have a positive impact, regardless of the specific peculiarities of the host country institutions.

Regarding the theories of institutional change, while this channel could be seen as reflect-ing Bush’s evolutionary theory better since it is the growth in knowledge that leads to new instrumental values, it could also be argued that it is a consequence of a new and knowledge-able foreign actor establishing itself. Creating a new institutional setting and uncertainty on the market that force organisations to adapt to be able to survive. Thus, there can be a conscious rational dimension to this as well, perhaps the new actor established itself be-cause the opportunity was identified by superior knowledge? Unlike the competition channel though, the resulting adaption is more of a passive adaption process that is primarily de-fensive rather than aggressive. It can be thought of as a Nash equilibrium where no actors have an incentive to deviate from a certain institutional structure, mostly because the payoff of the deviant route is uncertain. Whereas in the competition channel the payoffs are less uncertain, and the actors are chasing upside and avoiding costs. The herd behaviour in the mimicking channel is due to disoriented actors that mimic to survive Beckert (2010).

In a network setting, this can be understood as partly contingent on the selection effect. This since the uncertainty of which institutional structure attracts FDI can lead to mimicking. But also, not contingent on the selection effect since the uncertainty of how to reap the gains of already existing investments also can lead to mimicking.

2.3.5 Summary of channels

In table 2.1 some of the key attributes for each channel is summarised. Elements of both old and new theories of institutional change can be found in almost all channels except for the professionalisation channel that we judge as better suited to Bush’s theory. However, the mimesis channel leans more towards Bush as well while the bargaining power channel

(18)

leans more towards North. All channels except for the professionalisation channel imply both converging and diverging effects for institutions. The influence effect in the competition channel is heavily contingent on the selection effect while the professionalisation channel and mimesis channel only are partly contingent on it. The influence effect in the bargaining power channel is not contingent on selection first.

Table 2.1: Attributes of the channels of influence

Competition Bargaining Power Professionalisation Mimesis Bush vs North: Both Primarily North Primarily Bush Both

Convergence vs Divergence: Both Both Convergence Both

Contingent on selection: Yes No Yes and No Yes and No

2.4 Dynamics of globalisation and institutional

conver-gence

Both FDI and institutions are important components in the endogenous growth theory (Jones & Romer, 2010). As it outlines the dynamics of the growth process, we can understand the role of FDI and institutions in the globalisation process through this theory. Especially in the sense that the increasing extent of markets is a central part of growth. As FDI can be seen as a carrier of technology and human capital (Blomström & Kokko, 1998; Veugelers & Cassiman, 2004), it is a key component for the increasing extent of global markets. Jones and Romer (2010) also discuss the importance and returns to scale of ideas. If FDI is these ideas, there are strong economic incentives for attracting FDI. The scale effect of FDI implies there will always be gains from FDI as long as the world has not yet reached total connectivity and there are no more useful ideas. According to this theory, FDI flows will continue to increase to connect the world. FDI is per se a tangible economic part of the concept of globalisation. Jones and Romer (2010) argue that institutions can disrupt or favour the adoption of new ideas in an economy. If growth is mostly driven by innovation, the institutional framework must be able to facilitate these innovations. If institutions are the structure that determines the adjustment to new ideas, they are a sort of mediator between ideas and growth. As the growth process is a process of increasing connectedness between countries and individuals, there are economic incentives to have an institutional structure that eases the transmission of ideas in the most frictionless way possible. If these incentives take the form of a conscious or unconscious process of change remains ambiguous, but we know that institutions are the

(19)

infrastructure for connectedness, i.e. FDI and economic integration, and thus, there are economic incentives for actors to have this infrastructure built after a certain mould.

Given the theoretical assumptions of increasing returns to scale of connectedness (increasing FDI flows) and that institutions mediate the success of the increasing connectedness, there is a theoretical framework complete that explains possible institutional convergence. Given that countries are rational actors, they would understand these mechanisms and adapt the best institutional structure to facilitate new knowledge. The assumptions for this theoretical explanation are strong since it at a glance implies that the institutional change is driven by conscious rational actors. But, as expanded upon in the four mechanisms, the process does not necessarily need to be conscious, rational, or even imply convergence. Several conscious or unconscious processes may be happening at the same time. However, the theories on institutional convergence, like Fukuyama (1989), are intuitively easy to understand and have important implications if they have an empirical bearing. As institutions are a fundamental driver of economic growth, institutional convergence would imply economic convergence as well (Quah, 1996; Elert & Halvarsson, 2012). This is interesting because it implies that research on institutional convergence is generalisable to economic convergence as well. Since the literature on convergence is somewhat mixed and the sources of convergence and diver-gence are relatively unknown (Rassekh, 1998; Heichel et al., 2005), any further insights can only be welcomed.

The work of Fukuyama (1989) might be interpreted as somewhat normative but it is an interesting work that has caught a lot of attention. He argues that liberal democracy with a market economy is the final form of human government and that the world is converging towards this. According to Fukuyama (1989), the convergence towards a market economy is happening because history is an accumulative process where increasing innovation and knowl-edge ultimately will lead humanity to the best way of organising the economy. He argues that democracy is a different process where the human need for recognition in society (thymos) creates a force for democratic change. He also specifies that the further a society has come in economic development, people will to a higher degree demand recognition, thus the pro-cesses are conjoined. Interpreting this within our research, the increasing spread of knowledge through FDI will create convergence in the institutions toward a market economy primarily and by extent towards democratic governance. Relating it to the four transmission channels, this convergence can primarily be related to mimesis, professionalisation and competition. This since Fukuyama’s argumentation can be interpreted as that the accumulative learning process in society implies that actors have learned that a certain institutional structure is the best structure to reap economic benefits. This is similar to that the optimal institutional structure attracts investments, and is contingent on the selection effect first (mimesis, pro-fessionalisation and competition). It is also relevant that these three channels are leaning

(20)

towards Bush’s theory of institutional change and the victory of instrumental values through an evolutionary struggle, which is also the main premise of Fukuyama’s theory that the trial and error of history has led to this specific institutional structure (market economy) to be the final form of government. This rather deterministic view of institutional change is in sharp contrast with the later works of North (2005) that emphasises the unpredictability of institutional change.

(21)

3. Networks

Social network analysis is commonly used in analytical sociology to investigate social struc-tures. Here, we will not provide a complete guide to social network analysis, but some concepts are useful to understand in order to follow this thesis. Global FDI can be thought of as a global network, with individual countries connected by FDI positions. Social net-work analysis is used to study social relations (ties) between actors (nodes) in a netnet-work, using graph theory to mathematically analyse the formation, maintenance and structure of these relations. Working with networks implies the assumption that observations are inter-dependent, that ties in a network are contingent on other ties. This, in turn, implies that studying network data through regular econometric techniques is not possible since the data will violate assumptions of independence. Granovetter (1985) explains that while the issue of interdependence in economic relations have long been considered important in several fields of study such as sociology and political science, economists have generally rejected the notion of economic exchange as a social phenomenon and rather seen it as a purely instrumental prac-tice, with economic behaviour often assumed to be driven by rationality and self-interest. Granovetter (1985) argues that while he considers embeddedness to be lower in economic markets than other forms of interaction, it is not insignificant in shaping the decisions of market actors. Jackson (2007) explains that for a long time, economists largely ignored the arguments laid out by Granovetter (1985). But a shift, according to Jackson (2007), came from economists ”having pushed many economic models to their limits”. He discusses that this newfound interest in networks stems from similar limitations of economic theory that sparked the interest in behavioural economics.

3.1 Interdependence

Networks are in their simplest form built out of a set of nodes and a set of ties between these nodes. Networks can be undirected, where every tie represent a mutual relationship, or directed, where ties do not necessarily need to be reciprocated. Networks can also be binary

(22)

or weighted, wherein weighted networks, values are attached to the ties, representing the strength of the tie. For the interests of this thesis, we will only discuss directed networks. There are several ways to characterise the interdependence in a network, the most common factors typically found in social networks are those of reciprocity and network closure. Reci-procity describes the likelihood of node j to form a tie to node i if i already have a tie to j, as illustrated in figure 3.1. In effect, reciprocity means that ties tend to be mutual. For our purposes, we can view FDI as a business relationship and it is likely that an existing business relation lowers the barriers of investment, increasing the likelihood of reciprocating FDI.

Figure 3.1: Reciprocity

Second, network closure, or transitivity, refers to the likelihood of forming a triad and is measured as the proportion of closed triads among all possible triad configurations in the network. i.e. if node i is connected to node j and node k, then j and k are more likely to also form a connection, as illustrated in figure 3.2. Thus, the relationship between j and k is influenced by i. There are several different types of triad closure, depending on the directionality of the ties in the triad. Again, this relates to business relationships, actors with mutual business partners are more likely to have similar interests and connections, therefore the likelihood of meeting and also investing in each other might be larger.

Figure 3.2: Transitivity

Another network-specific measure that is of great interest for this thesis is homophily, the tendency of nodes with similar characteristics to also be tied together. As discussed by McPherson et al. (2001), this leads to homogeneity in ego-networks and has the possibility

(23)

of creating divides between groups. A central point in this thesis is the issue of homogeneity bias in institutions, not only through homophily (the selection effect) but also through insti-tutional diffusion (the influence effect). We will investigate this relationship by separating the selection and influence effects, similarly to how Steglich et al. (2010) investigate homophily in substance use among teenagers. Selection and influence effects in a SAOM are not mutually exclusive and can be modelled without raising concerns of endogeneity when including both in a model. They can further be utilised to explain what is the main contributory effect to the possible homogeneity bias in the network.

3.2 Applications to FDI

A network considers nodes of individual actors and the ties they have to other nodes. While it is intuitive to picture such a network as a network of friends, a network can just as well be made up of organisations and business transactions. When thinking about a global network of FDI, we see individual countries as nodes and FDI positions represent directional ties between countries. However, if we cannot assume interdependence between the countries FDI flows, the network approach is not necessary. This interdependence can likely be explained by several factors but at the heart of international relations lies politics. Countries strike trade and investment treaties, form alliances, condemn each other, praise each other, and even bomb each other. Such relations are likely to affect international investors in their decisions and possibilities in far more ways than could ever be incorporated into a statistical model. Koskinen et al. (2015) and Schoeneman et al. (2020) show, by exponential random graph models, that the global FDI network formation is highly influenced by processes of reciprocity and transitivity. Thus, it seems that FDI is indeed subject to indirect dependence and that a network approach to modelling FDI is warranted.

(24)

4. Literature review

The positive economic consequences of a rigid and efficient institutional environment stems partly from the works of Barro (1996), North (1990, 1997), and Acemoglu et al. (2001). Knack and Keefer (1995) and Mauro (1995) were among the first to empirically show that the institutional environment (property rights, contract enforcement and corruption) matter for the magnitude of private investment in an economy. Since then, a growing body of literature has emerged that empirically show that host country institutions are an important determinant for attracting foreign investment. Wei (2000) concludes from the estimates of a tobit model that ”An increase in the corruption level from that of Singapore to that of Mexico would have the same negative effect on inward FDI as raising the tax rate by eighteen to fifty percentage points, depending on the specification”. Similar results can be found in Campos and Kinoshita (2003) and Bénassy-Quéré et al. (2007) where low quality of government and poor legal institutions are found to be a deterrent of foreign investors. Another study has also found statistically significant effects on FDI flows when aggregating the institutional variables into a single variable (Buchanan et al., 2012). In a meta-analysis by Bailey (2018) of 97 previous studies, the author finds evidence that democratic institutions, rule of law, and political stability attract FDI while corruption, tax rate and cultural distance deter it. Critics of the view that good institutions are important for FDI inflows often point to the case of China, the largest recipient of FDI in the world even though the country is known for its weak democratic and economic institutions. However, Fan et al. (2009) argue against this view, showing that the amount of FDI China receives is not extraordinary given their development level and institutional quality. In sum, one who wishes to make novel findings on the general causality regarding if host country institutions can affect FDI inflow face a burdensome quest. However, Trevino et al. (2008) nuance this view a bit, arguing that institutionalisation works through three pillars (cognitive, normative, and regulative) and that FDI is mainly attracted by cognitive and normative institutional change and not regulatory institutional change. While the importance of institutions for attracting FDI has been quite well established, studies of the opposite direction are generally inconclusive. Most previous literature on this topic focuses on the general effect FDI has on institutions and not on how FDI spread institutions. However, we still deem studies on the general effect relevant and we are confident

(25)

that the two phenomenons are closely related and overlapping to a large extent. The two effects are very difficult to disentangle theoretically. Perhaps the general effect can be argued as being within the institutional diffusion effect, and that how a country affects another country’s institutions via FDI depends on a multitude of factors where the general impact of FDI and spillovers can be one of several. Thus, the general effect can be seen as a relevant proxy for the institutional diffusion effect.

The empirical literature on FDI’s impact on host country institutions emerged in the early ’00s. A study by Hellman et al. (2002) finds evidence that FDI firms are more likely than domestic firms to engage in corrupt action in already misgoverned settings, further magnifying problems of corruption. Contrary to this, another early study by Kwok and Tadesse (2006) examines the effect of the presence of multinational corporations and inflows of FDI on corruption. They find that countries that have received high levels of FDI in the past exhibit significantly lower levels of corruption. Hyun (2006) find results similar to Kwok and Tadasse, when adopting a bidirectional approach to the relationship of FDI and institutional quality. Hyun investigates a more general institutional variable that includes rule of law, bureaucratic quality, and corruption, and find long-run causality between institutions and FDI.

Following the early literature of FDI impact on host country institutions, several efforts to improve the econometric methods and deal with the implied simultaneous causality followed. Earlier studies such as Kwok and Tadesse (2006) and Hyun (2006) rely on panel estimations with lagged explanatory variables to investigate the directional causality. But, as Bellemare et al. (2017) discuss, using lagged variables to counter reverse causality in a panel model ”merely moves the channel through which endogeneity biases causal estimates”. Malesky (2009) deploy an instrumental variable approach instead when investigating the effects of FDI on the economic reformation process in a sample of transition economies. Malesky find that the type of FDI affects the influence it has on economic reform. To gain thoroughly robust results, Ali et al. (2011) use a variety of different types of models, ranging from random effects to system GMM specifications. They find robust results of FDI inflows affecting property rights positively.

The growing pattern of positive institutional effects from FDI falls short of being conclusive. Most empirical studies on the causal effect from FDI on institutions (Dang, 2013; Long et al., 2015; Zhang et al., 2019) as well as those studies looking at bidirectional causality (Fukumi & Nishijima, 2010; Shah et al., 2016; Huynh et al., 2020) are limited to specific countries or regions and do not allow for generalisation of the results. Other studies are too narrow in their focus, such as Prakash and Potoski (2007) that only look at the adoption of voluntary environmental regulations and not general institutional change.

Other studies find negative or no impact from FDI on host country institutions. Desbordes and Vauday (2007) investigate the political influence of multinational corporations through a

(26)

probit regression, finding that foreign firms are able to use their bargaining power to acquire regulatory advantages over domestic firms. Similarly, Ginsburg (2005) find that bilateral in-vestment treaties (BIT) have ambiguous effects on governance quality, concluding that they might even be harmful to local institutions under certain conditions. Ginsburg (2005) discuss that this could be caused by BIT’s offsetting previous regulations, giving competitive advan-tages to foreign investors. It is however important to note that even though BIT’s promote FDI inflows (Busse et al., 2010), they are effectively institutions in themselves and likely to affect different mechanisms of institutional change than those we are interested in. FDI may also hinder and slow down reformation processes, Bak and Moon (2016) find evidence that FDI may fund authoritarian leaders and help them retain their power. Vanderhill et al. (2019) reach a similar conclusion when examining the countries of Kazakhstan, Azerbaijan and Kyrgyzstan, arguing that institutional reforms can be suppressed by FDI in countries rich on natural resources. Finally, Demir (2016) find no evidence that FDI affects institu-tional development gaps between countries in a general setting, but in natural resource-rich countries, FDI can have both converging and diverging effects on the institutional distance between countries depending on the host and home country. The results from Demir (2016) are particularly interesting as it is one of the few studies that investigate the effect FDI can have on spreading institutions.

4.1 Specifics of the relationship

Since the evidence for institutions affecting inwards FDI is convincing, but the evidence for FDI’s impact on institutions is not, it remains ambiguous if FDI can help to spread institutions. The effects also seem governed by several moderating factors. These moderating factors can help us to partly understand the underlying mechanisms of the relationship. The meta-analysis conducted by Bailey (2018) show that the effect institutions have on FDI seems to be significantly moderated by trade relations, host country wage level, geographical distance, infrastructure, common language and two of the following: GDP, GDP per capita and population. Also, Pajunen (2008) show in estimations that the different institutional effects on FDI can have different outcomes depending on which region or country that is examined. There is also evidence that cultural distance has a moderating effect on the relationship as pointed out by Du et al. (2012).

Even as the evidence for an institutional diffusion effect remains inconclusive, there appears that some moderators are important for this direction. FDI in natural resource-rich countries may have both positive and negative effects on institutional similarity, depending on the host and home country attributes (Demir, 2016). As argued by Vanderhill et al. (2019) natural

(27)

resources might create rentier states that are incapable of developing their institutions. Con-sidering the results of Demir (2016) this seems to be true if the investor is from a developing country and the host country also is a developing country, then FDI may actually increase institutional distance. However, if the investor is from a developed country, they may in-stead decrease the institutional distance when investing in natural resource-rich developing countries. This contrasts with Vanderhill et al. (2019) and the natural resource curse theory in general. Malesky (2009) also find evidence that supports the fact that investment in the natural resource sector slow institutional reform, while FDI in the manufacturing and service sector helps to promote institutional reform. Similar to this, Shah et al. (2016) find that long and short-run positive bidirectional causality primarily exists for investment into the man-ufacturing sector in Pakistan. Also, some studies indicate that initial institutional quality (Dang, 2013) as well as previous FDI concentration (Long et al., 2015) can have moderating effects on the FDI to institutions relation.

Another interesting dynamic of the relationship between FDI and institutions that we need to consider is the time dimension. The bidirectional studies indicate that the relationship exhibits a long-run relationship rather than a short run (Hyun, 2006). Studies on FDI’s effect on institutions usually lag their explanatory variables, both because they want to deal with the potential reverse causality, but also because the process of institutional change is naturally slow. Kwok and Tadesse (2006) test both 30-, 20-, and 10-year-old FDI inflows on institutions and find that FDI inflows in all periods have significant effects on corruption. Ali et al. (2011) lag their endogenous variables by five years and find significant effects on property rights. Zhang et al. (2019) who employ impulse response analysis find significant effects on institutions after four years. What is interesting is that Demir (2016), who find no significant effects other than when examining natural resource-rich countries, only look at past FDI flows to a maximum of two years back in time. To the contrary Malesky (2009) find a significant effect on institutional reform when only lagging the explanatory variable with one lag.

An important aspect when it comes to the time dimension is also if the studies examine FDI stocks or FDI flows. Stock and flows are highly correlated but since FDI stock is the accumulation of FDI flow over time minus the depreciation it can be argued that FDI stock is a better measurement at capturing long term effects than the more responsive and volatile FDI flows (Wacker, 2013). Kwok and Tadesse (2006), Demir (2016), and Zhang et al. (2019) examines FDI flow, Hyun (2006) and Malesky (2009) examines FDI stock and Ali et al. (2011) examines both measures. What is distinguishable is that studies examining FDI flows seem quite sensitive to lag length while studies examining FDI stock are a bit more ambiguous regarding this.

(28)

unreasonable to theorise that FDI, in general, has a long-term effect on institutions. When modelling FDI flows the effect seems to be present on a minimum four to five-year horizon, when modelling FDI stock the effect might be more instant since a one-year lag length is not debunked by Malesky (2009) probably due to the accumulation aspect. Even though this evidence is not entirely conclusive and should be interpreted conservatively it is an important indication for modelling. A model that ignores this time dynamic might miss this effect entirely.

4.2 Network literature on FDI and institutions

To our knowledge, no studies are examining the relationship between FDI and institutions in a network setting, there are however a few that focus on each concept separately. Goddard (2018) examine the role of network positions in an institutional network and conclude that revisionist states are limited by and must adapt their reform agenda according to their network setting. Arif et al. (2021) study both FDI network formation and possible influence factors of FDI. They find that global FDI networks have become denser, less independent and that the global FDI network is largely controlled by a few influential actors. They also find that national competitiveness levels and FDI are closely linked, which, for our interests, could indicate that a relationship between FDI and economic institutions is observable in a network setting as well. Lima et al. (2020) use a global FDI network to analyse the network position of Portugal and discuss what the network can say about country relationships and international paths of investment. Finally, both Koskinen et al. (2015) and Schoeneman et al. (2020) find that global FDI flows are formed by processes of reciprocity and transitivity. Thus, we can conclude that FDI network formation appears to be influenced by interdepen-dent processes and that there is a gap in the literature on FDI networks and institutional change. This, in addition to the inconclusive results with regard to institutional diffusion through FDI, puts forward a research agenda where it would be possible to bridge the gap between the literature on networks and that on institutions and FDI.

(29)

5. Hypotheses

For the purpose of our study, we have outlined two hypotheses. The hypotheses are derived from theory and previous findings and have been constructed to be testable in a Stochastic Actor-Oriented Model.

As discussed in the chapter 2, the relationship between FDI and institutions can be divided into a selection and an influence effect. Since the fact that institutions attract FDI is quite well established in the literature (Wei, 2000; Campos & Kinoshita, 2003; Bailey, 2018), it should be present in the model as the selection effect. We hypothesise that the network formation to a large extent depend on homophily, which needs to be tested as a formality to validate that the generalisation that institutions attracting FDI holds in a network setting. Given this, the first hypothesis will be formulated as the following:

H1: There is a selection process present in the model

Furthermore, as outlined in the introduction, we want to investigate whether FDI can con-tribute to institutional diffusion. Something that is ambiguous in the literature today (Kwok & Tadesse, 2006; Ali et al., 2011; Demir, 2016). As expanded upon in chapter 2 it can be interpreted as the influence effect. Based on transaction cost theory and the four transmis-sion channels we hypothesise that countries spread institutions similar to their own through foreign investments (DiMaggio & Powell, 1983; Williamson, 1989; Beckert, 2010). The hy-pothesis will be formulated as the following:

H2: There is an influence process present in the model

Another aspect we are curious about is what the results from our models can tell us about the theoretical questions in play regarding institutional convergence and the globalisation process. What we are primarily interested in is the Fukuyama (1989) theory of institutional convergence. We will not test if the theory holds or not, since we have no proper way of testing this with our methodology, but we will look closer at if FDI and the influence effect can be a contributory cause for this theorised process. If the influence effect for the ”liberal market economy” institutional variable is positive, it is probable that FDI matter for institutional convergence and that institutional convergence is driven by influence. This does not mean

(30)

that institutional convergence is happening nor that it cannot be driven by other factors as well, only that influence processes stemming from FDI are contributing factors towards global institutional convergence.

(31)

6. Method

Early literature investigating the link from FDI to institutions with econometric methods relied on lagged dependent variables to counter reverse causality (Hyun, 2006; Demir, 2016), the rationale for this being that causality cannot run backwards in time. Since this is also in line with the idea that institutions are not affected immediately but rather by past values of FDI, such a specification appears tempting.

However, as Bellemare et al. (2017) discuss, lagged explanatory variables are still subject to reverse causality bias and ”merely moves the channel through which endogeneity biases causal estimates”. In our case, past values of FDI are still affected by past institutional settings while at the same time, following the notion that institutions are highly path-dependent, past institutional settings is a determinant for the current institutional setting. Thus, the lagged explanatory variable is still subject to reverse causality and the estimates cannot be trusted. This logic extends to the use of system GMM models (as used by Ali et al. (2011)), which rely on lagged values as instruments.

Instead of lagging the endogenous explanatory variable, it is often advised to use instrumental variable regression to isolate the exogenous variation in the endogenous explanatory variable. Finding a good instrument that is uncorrelated with the error term and correlated with the endogenous explanatory variable can however be quite a challenge. The main instrumental variables used in the literature to isolate the exogenous variation in FDI are exchange rates and infrastructure (Kwok & Tadesse, 2006; Malesky, 2009; Ali et al., 2011). There are however issues with both these variables. Exchange rates are, as argued by Malesky (2009), prone to domestic policy manipulation and can thus be correlated with the institutional dependent variable. This with the combination that the data availability on exchange rates for our sample is far from ideal, shifts our focus to infrastructure.

Data availability on infrastructure is seemingly better, but there remains a problem with exogeneity. As seen in table 6.2 our infrastructure variable of the percentage of the population with access to electricity, is moderately correlated with the institutional variables. It might not be an issue if one could introduce a relevant control variable that isolates the instrumental variable from correlating with the institutional variables. However, identifying such control

(32)

variables is challenging, and there is no sure way of testing if they can help isolate the instrumental variable.

If the instrumental variable is correlated with the error term the model might contain incon-sistent estimators. This ultimately becomes a trade-off situation where we have to balance possibly inconsistent estimators from endogenous instrumental variables with inconsistent estimators from simultaneous causality bias. Thus, we still expect that the regression results would be unreliable. In part due to inconsistency stemming from the instrument’s correla-tion with the error term and not adequately controlled for, which could result in a positive bias on the coefficient for FDI stock. And in part due to the theorised confounding factors of interdependence still not being solved, this is likely to further bias the estimates upward since reciprocal effects might be interpreted as influence effects in the regression.

Since we have a theoretical understanding of the complexities of the relationship between FDI and institutions, we suggest that a network approach is warranted to give a better and unbiased interpretation of the dynamics in this relationship. The network approach that will be used is a Stochastic actor-oriented model (SAOM) which is used to investigate social network formation over time (Snijders et al., 2010).

The SAOM approach to the relationship of FDI and institutions lets us circumvent endo-geneity and interdependence issues. It allows us to examine the interactive mechanisms of the relationship on an aggregate level in a way that is simply not possible with econometric modelling. The causal interpretation of an econometric model does not only risk being biased in this case but also risk invalidating any conclusion drawn due to the complex time order of cause and effect in this case. Furthermore, this allows us to model a theoretically more reasonable similarity effect rather than the general effect of FDI.

However, the results obtained cannot straight up be interpreted as causal, but rather as insights to the processes in action that can help us understand the dynamics in play. As dis-cussed in the theoretical section of this thesis, some of the influence channels are contingent on selection first. With this perception of the order of events, the basic concept in causality that cause must precede the effect is put into question. How can we examine the effect (insti-tutional change) when it precedes the cause (FDI) in some examples? This complicates the study at hand since our aim is to disentangle the relationship between FDI and institutions and model it correctly. However, few models come without caveats and many times it comes down to individual subjective judgements of what is the best approximation of reality. We cannot ignore that the economic models might ”be pushed to their limits” not only when it comes to embeddedness, but also to the causal interpretation when it comes to this rela-tionship. We remain open to the fact that our study can be restricted to finding a plausible economic narrative instead of establishing causality. To quote the famous statistician George E.P Box: ”All models are wrong, but some are useful”.

(33)

SAOMs are agent-based models that deal with longitudinal network data sequenced into two or more ”waves”, where a wave is the period between each observation. The changes between the waves are decomposed into ”mini-steps”, where each mini-step constitutes a decision for a single node in the network. The node is then allowed to either modify its outgoing ties, its behaviour or do nothing. Ties are modified by sending a new tie or removing a tie, behaviour is modified by changing the dependent attribute variable. Each decision is determined by a set of probabilities estimated by the model. Parameter effects are specified to determine what probabilities, and thus decision rules, that are evaluated in each mini-step. The waves are then simulated thousands of times in order to obtain effect probabilities that lead to a simulated network matching the real network as well as possible.

The certain specification for the model we will use is a directed one-mode network which will allow us to study the co-evolution of the network structure and the actors’ behaviour in the network. As discussed before, FDI stock will be used to form the ties in the network, what we have not touched upon is the behavioural aspect that the model allows for, which is key when it comes to the selection and influence effect. The SAOM can include behavioural data of the nodes, which in our case will be the institutional variables of the countries. The behavioural data can be thought of simply as an attribute that is attached to the actor. The effect of how the behavioural variable affects network structure can now be referred to as the selection effect and the reverse proposition can be referred to as the influence effect.

SAOMs cannot handle continuous weighted ties, but they can handle discrete weighted ties through ordered SAOMs (Snijders et al., 2010; Elmer et al., 2017; Ripley et al., 2020). How-ever, our initial tests with ordered SAOMs resulted in excessively complex models with many issues. Instead, the FDI network data is separated into a general FDI network which repre-sents all FDI ties, and a high FDI network which reprerepre-sents only the largest FDI positions. We run analyses on both networks for each dependent variable.

It is also possible to add exogenous covariates. The covariates can be modelled to have a certain effect on network formation or behavioural change. This kind of modelling is necessary for a SAOM to properly find the best model fit but also to test different hypothesises. We can model the SAOM with different effects stemming from the covariates, behavioural change, network formation or different interactions of these entities. The effects are used to construct a mosaic of network dynamics that explain why the network and behaviour change the way it does between the waves. The effects can be thought of as the explanatory variables of the network and are the study object of interest. Some covariates need to be added as controls to yield unbiased estimates in order to avoid that any unobserved factors influence the results. The resulting output can be interpreted similarly to a logistic regression, and each parameter represents non-standardized contributions to log-probabilities (Ripley et al., 2020).

Institutional Dynamics in the Global FDI Network : Examining The Co-evolution of Institutions and FDI with Stochastic Actor-Oriented Modelling