DEPARTMENT OF ECONOMICS Uppsala University
Economics C/Thesis Work, 15c
Authors: Lina Björkman & Maria Ullgren Supervisor: Henrik Jordahl
Spring semester 2018
Political Trade
A study of the relationship between bilateral trade flows and regime type
Abstract
In this study we examine how bilateral trade flows are affected by regime type. Previous research with the aim to investigate the effect of regime type on bilateral trade has primarily used a binary definition of democracy, and findings have indicated that democracies trade more extensively with other democracies. In this thesis, we add to the existing literature by disaggregating regime types, defining liberal democracies, electoral democracies and authoritarian regimes, and adopt the well-proven methodology of the gravity model of trade in OLS regressions. We use trade data adopted from IMF in the years 2004 and 2014 covering 162 countries. Data on the main explanatory variable, regime type, is defined in accordance with democracy indicator Freedom House’s methodology. Results show that pairs consisting of two liberal democracies have the highest values of trade flows, compared to other combinations of trading regime type-pairs. However, due to methodological discrepancies and ambiguous estimates, the relationship between regime type and trade flows remains uncertain.
Acknowledgement: We would like to thank our supervisor Henrik Jordahl for providing us with wise insights along the way of the work with this thesis.
Key words: Trade, Regime type, Gravity model, Democracy
Table of Contents
1. Introduction 4
2. Background 6
2.1 Conceptualisation and definitions of regime types 6
3. Theoretical framework 7
3.1 Mansfield et al.’s Model of Bilateral Trade Actors and Their Preferences 8
3.2 Gravity Model of Trade 10
3.2.1 Theoretical predictions on the effect of regime type on bilateral trade flows 12
3.3 Literature review 13
4. Data 17
4.1 Baldwin and Taglioni’s Gold-, Silver- and Bronze-medal mistakes of the gravity model 17
4.2 Data on the dependent variable - Trade flows 18
4.2.1 Evaluation of data for control variables 18
4.2.2 Assessment of missing values 19
4.3 Data on the main explanatory variable - Regime type 20
4.3.1 Evaluation of regime type indicators 20
4.3.2 Assessment on the problematics in quantifying regime type 25
5. Empirical method 25
5.1 Year and country pair specific-fixed effects 26
5.2 Evaluation of additional control variables and methodological adaptations 26
5.3 Introduction of regime type-paired dummy variables 28
5.4 Assessment of internal and external validity 30
6. Estimation and results 32
7. Discussion and concluding remarks 35
7.1 Concluding remarks 39
References 40
Appendix I 43
Appendix II 45
List of tables and figures:
Table I: “Summary of previous literature” 13
Table II: “Descriptive statistics” 19
Table III: “Matrix of aggregated number of trade flow observations according to regime type” 24
Table IV: “Regression results: (1) Base model” 32
Table V: “Regression results: (2) Extended model with regime type-dummies” 34
Table AI: “Listing of civil liberties scorings (CLS) and criteria in accordance with the FH methodology” 44 Table AII: “Listing of political rights scorings (PRS) and criteria in accordance with the FH methodology” 44
Table AIII: “VIF I: (1) Base model” 45
Table AIV: “VIF II: (2) Extended base model with regime type-dummies” 45
Figure I: “Number of Free-, Partly Free- and Not Free countries according to Freedom House, 2003-2014” 23
Figure AI: “Map of Freedom, 2004: Countries categorised in accordance with FH status in 2004” 43 Figure AII: “Map of Freedom, 2014: Countries categorised in accordance with FH status in 2014” 43
Abbreviations
DOTS Direction of Trade Statistics
FH Freedom House
GDP Gross Domestic Product
GNP Gross National Product
IMF International Monetary Fund
OLS Ordinary Least Squares
OVB Omitted Variable Bias
UN United Nations
VIF Variance Inflation Factor
WDI World Development Indicators
1. Introduction
“It has been said that arguing against globalization is like arguing against the laws of gravity.”1 Kofi Annan, former Secretary-General of the United Nations.
For the first time in history, regime type indicators show that democracy is the prevalent regime type among the world’s nations (Mukand & Rodrik, 2017). Simultaneously, international trade is more extensive than ever (World Bank, 2018a). Economies worldwide rely hard on their imports and exports, but who do they chose to trade with? Empirical findings from previous studies provide evidence in favour for the so-called “hypothesis of democratic trade”, that is: democracies trade more with other democracies than they do with other regimes (see for example Mansfield et al., 2000; Milner & Kubota, 2005; Yu, 2010).
However, China – one of the world’s least democratic countries (Freedom House, 2014a) – is also the world’s largest trading partner, and democracies like Sweden and the U.S. trade extensively with authoritarian states such as Saudi Arabia and Russia (World Bank, 2014).
This raises the question - what is the relationship between regime type and trade?
Albeit the great importance of trade in the modern globalised society there is a lack of studies made on how regime type affect commerce, leaving ample room for further research in the field. Previous research has mainly focused on the variation between democratic and authoritarian states, not taking differences within regime types into account (see Section 3.3).
If we conclude that all democracies and authoritarian states seek out trading partners in a similar pattern, and leave out modern variations in regime types, we will not be able to analyse reasons for this in a as credible way as we could if we are aware of potential heterogeneity. Previous findings have indicated rather unanimously that there is a relationship between bilateral trade flows and regime type. Recent studies argue that the uncertainty of the effects of economic growth, a closely related study field, on regime type may be based on the need to distinguish between electoral and liberal democracy (Mukand & Rodrik, 2017). This distinction between regime types has not yet been made in literature on international trade.
Using the distinction between liberal and electoral regimes allow us to analyse differences in regime type reflecting modern society, thus a more precise and correct analysis of how bilateral trade is affected by regime types can be achieved. Considering trade and its influence on the global economy, it is interesting to investigate how regime type affect trade
1 Kofi Annan in his opening address to the 53rd UN Department of Public Information/Nongovernmental Organizations
flows when we introduce the disaggregated definition of democracy. The aim of this study is thus to investigate how trade flows differ between regime type-pairs, observing the years of 2004 and 2014. We do this by analysing how differences in regime type can explain variation in trade flows between liberal democratic, electoral democratic and authoritarian states.
Using trade level data provided by the International Monetary Fund (IMF, 2018), this thesis will perform an empirical study of a set of 162 countries in the years 2004 and 2014 respectively. Thereby, we add to the existing literature by examining more present data. In line with previous literature, we then adopt the methodological framework of the gravity model of trade, which describes bilateral trade as positively related to the economic masses of the trading countries and negatively related to the distance between them (Tinbergen, 1962;
Anderson, 1979; Green et al., 2001; Anderson & Marcouiller, 2002; Yu, 2010). Anderson and Marcouiller (2002) describes the gravity model as one of the most successful within the field of international trade, largely due to its effectiveness in connecting trade flows to GDP, trade barriers and measurements of distance. We find that country pairs consisting of two liberal democracies trade the most extensively, in line with previous research and the hypothesis put forward (as referred to in Section 3.3). However, we do not acquire any significant results implying pairs consisting of either solely electoral democracies or authoritarian regimes having higher trade flows than mixed country-pair combinations.
Results are dubious and no indesputible conclusions can be drawn.
The outline of this thesis is as follows: Further explanation of the classification of the three regime types adopted in this study will follow in Section 2. We provide a theoretical framework and evaluate previous literature, as well as outline previous findings in research regarding the connection between trade and regime types in Section 3. In Section 4, data for the dependent variable as well as for the independent variables is presented along with descriptive figures and statistics. We introduce the empirical method in Section 5 along with the threats to internal and external validity of the estimations as well as our model specifications. In Section 6 we present results and estimates of the empirical models. Finally, a discussion of our findings and concluding remarks are presented in Section 7.
2. Background
In this section we provide a discussion of definitions of regime types, granting us the foundation of the main explanatory variable presented in this thesis. Studying the relationship between bilateral trade and politics is neither new nor controversial. Using trade as tool for internationalisation and development has been prevalent since ancient times (Stearns &
Langer, 2001), and trade plays an important role in economic politics as well as foreign policy among several policy fields (Milner, 1999). In 2016, trade made up on average 56 per cent of world countries’ GDP, according to national accounts data (World Bank, 2018a), implying the great importance of imports and exports for countries worldwide. However, countries differ to what extent they trade with each other and many studies has been dedicated investigating potential explanations to this, regime type being one (see Mansfield et al., 2000; Milner & Kubota, 2005; Yu, 2010).
2.1 Conceptualisation and definitions of regime types
Although there are no objective definitions of neither democracy nor authoritarianism, but it is widely accepted among political scientists that there are gradient differences within regime types (Dahl, 1989; Diamond; 1997). Political theorist Dahl (1989) has assessed what democracy (or ‘polyarchy’ as he refers to in terms of imperfect democracy) is by examining the most elemental characteristics of democracy and tested them against arguments raised by its critics. Dahl finds two key concepts for democracy. First, participation of citizens in the political system is essential. Citizens right to vote in fair elections as well as their extent to legal rights are fundamental for democracy. This is sometimes referred to as a minimalist definition of democracy in political science literature (Diamond, 1997). The second key concept of democracy is contestation or opposition, which is present in what is referred to as liberal democracies. In order to be a liberal democracy, civilians must have the right to arrange in political parties and independent associations and organisations. Furthermore, citizens must have the freedom to speak and publish opinions without risking being censored by the government. Leading contemporary scholar in democratic studies Diamond (1997), was among the first to elaborate a new definition of the democracy term with two distinctions. Diamond distinguishes between liberal and electoral democracies, the former being more restrictive in its definition and the latter being minimal, less restrictive, in its definition of democracy.
Electoral democracy is a formal conception of democracy. It is constitutional in its characterisation, using institutionalisation of elections instead of focus on the rule of law or the rights of minority groups as determinants in its definition (Diamond, 1997). Liberal democracy however, extend beyond the minimal definition democracy and comprise civil liberties and freedom for both individuals and groups (Diamond, 1997). Attaining the status of liberal democracy is possible only for political democratic systems in which the governments are liable and represent and its people. A full set of stable and legitimate political institutions must be in place for a country to acquire the political status of liberal democracy. In liberal democracies, property rights protect asset holders against expropriation from the government or other groups holding political power (Diamond, 1997). Regimes with no democratic institutions in place are consequently defined as authoritarian regimes.
Authoritarian regimes are associated with a lack of both political rights and civil liberties, however there are variations also in authoritarian regimes. Some authoritarian states might hold elections or have some civil society and autonomous judiciary, while others are totalitarian and regressive towards its population. The most prominent thing they have in common is that they do not allow for a legal and independent opposition (Diamond, 1997;
Freedom House, 2014b).
In a study on democracy and its determinants, Barro (1999) concludes that highly democratised regimes are associated with sufficient property rights protection and maintenance of the rule of law, which corresponds with the definition of liberal democracies.
However, electoral democracies do not provide institutions of comparable quality, and in the case of authoritarian systems, the executive power is generally synonym with the judiciary and the government is often more or less an absolute property holder. Barro argues that property rights protection and an independent judiciary are two factors that help to establish and provide a proper foundation for a fair market.
3. Theoretical framework
First, we present a simple model of bilateral trade actors and how they determine trade barriers according to their respective utility functions, which was introduced by Mansfield et al. in 2000. With the derived utility functions, we make the generally accepted assumption that the lower trade barriers in a trading country pair, the more extensive trade will be performed and thus leading to higher trade flows, which corresponds with the theory of
comparative advantages (IMF, 2001; Salvatore, 2004) In addition, we propose the gravity model of trade as a suitable theoretical framework that allow us to predict how trade flows are related to the regime type of the trading partners empirically. In relation to this, we state our central hypothesis. Then, we present previous literature made in the field, and finally the predictions stated in this thesis are put in relation to prior works.
3.1 Mansfield et al.’s Model of Bilateral Trade Actors and Their Preferences
Mansfield et al.’s (2000) model of bilateral trade actors and their preferences state that all countries have their individual utility functions in terms of trade, depending on their respective trade preferences. In the following, this will lead to different levels of trade flows depending on the chosen policy. The model makes three assumptions; 1. Trade barriers can be lowered by negotiations, and thus the two countries’ aim to negotiate barriers that lead to their optimal utility functions respectively; 2. The trading countries have no power over the counterpart’s exporting function (e.g. through subsidies on exports); 3. Only two countries are involved in the trade, giving no attention to multilateral impact. Democratic and authoritarian regimes differ in their institutional characteristics, which this model takes into consideration and applies to trade barrier negotiations.
One of the prevalent differences between a democracy, either electoral or liberal, and authoritarian regimes is that the former is ascribed a legislature, which has de facto power of authorisations of proposals put forward by the chief executive. In authoritarian regimes, no such institution exist making it an arbitrary actor; it lacks an independent legislature and does therefore only consist of an executive, which we label A in the following utility equations (see page 10). Democracies, electoral and liberal, are contrariwise distinguished by two actors; the executive (denoted P) and a legislature, who is represented by the median legislator (denoted C) in accordance with the median voter theorem. The median voter theorem declares that in a majority voting system (the theorem assumes direct democracy, but is usually applied to representative democracy as we do here), the outcome supported by the median voter will be selected (Black, 1948). Here, this would be the voter with the median value for level of domestic trade barriers, and thereby also represent the outcome of the legislative. Again, in an authoritarian regime system, the legislator is synonym to the chief executive. Considering bilateral trade includes two countries, we distinguish between the two by referring to them as the “importing country” (denoted country i) and the “exporting
country” (denoted country j). We assign the country i the labels A, P and C, whereas we refer to country j with A*, P* and C*.
The two trading countries both wants to maximise their utility by obtaining as much political support as possible. To enable for this, the actors must chose the most favourable trading policy, that is the one that will maximise political support. For a democratic country, that is the ideal point of the median voter. However, there are several groups with potentially conflicting preferences that needs to be considered. That could be consumers or exporters, who prefer low trade barriers, or import-competing firms who would prefer higher trade barriers. Albeit these conflicting preferences, there are common grounds; the different groups all prefer the trading counterpart to apply as low trade barriers as possible, and the home country, i.e. the importing country, to employ trade policies that maximise consumer surplus and firms’ profits. The policymakers who are able to negotiate the lowest trade barriers are thus the ones prefered by the public, ceteris paribus. We refer to country i’s trade barriers as t and those of country j as t*. The ideal level of trade barriers are denoted ti and ti*. In a democracy, the ideal domestic levels of trade barriers (i.e. tp and tc) are prone to differ. This is due to the legislative, C, (representing the median legislator) and the executive might have different preferences in terms of consumer surplus and firms’ profit (Mansfield et al., 2000).
Mansfield and Busch (1995) and Rogowski (1987) argue that legislators are more prone to emphasise special interest for smaller constituencies. The chief executive, P, on the other hand is assumed to value consumer surplus higher. Lohmann and O’Halloran (1994) and Rosendorff (1996) makes the assumption that this indicates that the median legislator hold more protectionist preferences than the executive counterpart, hence preferring higher domestic trade barriers. From this, we assume that the ideal domestic trade barriers for country i and country j (*) are tp < tc and tp* < tc* respectively.
The function representing political support, Ui and Uj, is a loss function. The two actors maximise their respective utility by minimising the difference between the levels of trade barriers that it prefered by its own country, and the level that it obtains when negotiating with the counterpart.
The utility function that country i aims to maximise is thus Ui(t, t*) = −(t − ti)2− t*2 for i = A, P, C.
whereas the counterpart, country j (*), wants to maximise U,(t, t*) = −t2− (t* − ti*)2 for i = A*, P*, C*.
Moreover, Mansfield et al.’s (2000) model implies that regardless regime type, democratic or authoritarian, there will always be incentives to trade. For democratic regimes, this may not sound controversial. Politicians will be motivated to be reelected and thus aim to make policy in accordance with the public opinion to obtain political support. However, political support is important also for politicians in authoritarian regimes, but the extent of support depend to a greater extent on economic conditions. When maximising consumer surplus and firms’ profit, this will lead to higher incomes (i.e. a better state for the overall economy, which is more important for the authoritarian leader?), and thus increase political support for the leader and reduce public incentives to riot.
As previously mentioned, for the home country, ceteris paribus, foreign trade barriers are prefered to be as low as possible, disregarding regime type. However, regarding domestic trade barriers, the ideal levels vary depending on preferences. This will consequently lead to trade flows differing between country pairs.
3.2 Gravity Model of Trade
When conducting research on bilateral trade and which factors that affect trade flows, the gravity model of trade has proven to be successful (Baldwin & Taglioni, 2006; Santos Silva
& Tenreyro, 2006). The original, intuitive way of using the gravity model of bilateral trade is similar to the logics of Newton’s gravity model of physics. That is, the gravity force between two objects depend on their respective masses and is inversely proportional to the distance between the two. The analogy of the gravity model of bilateral trade implies that trade flows between two economies depend on the size of their economic masses and is negatively correlated to trade costs in between them. Economist Tinbergen first introduced the theoretical framework of the gravity model in 1962 in an attempt to explain the pattern of international trade flows, where the importance of each of the trading countries’ respective gross national product (GNP) and the geographical distance between the two were emphasised as explanatory factors. Tinbergen’s specification has since been widely used as a
result of being a good fit for most datasets on international trade flows (Helpman et al., 2008).
The traditional equation of the gravity model captures the size of an economy and how it affects the production - and therefore export possibilities - as well as the domestic market and import possibilities of any country. Thus, the GNP of the two countries in a trading pair are positively correlated to the size of the trade flow between them. In most literature using the gravity model, the gross domestic product (GDP) – not GNP – is the prevalent measure of the countries’ economic masses (e.g., Mansfield et al., 2000; Anderson & Marcouiller, 2002;
Milner & Kubota, 2005; Yu, 2010) In contrast, distance is negatively correlated with the trade flow between countries, due to its relationship with transportation costs. More generally, the simplest gravity model is thus:
𝑌!" = 𝛼!𝐺!!!𝐺!!!𝐷!"!!
where Yij is the trade flow from country i to country j and is proportional to the product of the two countries respective GDPs (denoted by Gi and Gj), and the variable for distance, Dij, which is inversely proportional to the trade flow (Santos Silva & Tenreyro, 2016). α0 is the constant and α1, α2 and α3 are unknown parameters. Distance here can be interpreted as describing all factors impeding on trade, but as addressed above, the distance variable refers to the geographical distance between the two countries as in the original gravity model (Tinbergen, 1962; Anderson, 1979).
De Benedictis and Taglioni (2011) argue that the research that has been conducted since Tinbergen presented his original formulation of the gravity model of bilateral trade in 1962, has led to more a productive approach to many of the issues Tinbergen first emphasised.
Additionally, it has been established through more recent research that there are more things that goes into trade costs than solely distance. The concept of the distance variable has thus been interpreted more freely, some of which will be further elaborated in Section 3.3, page 13.
In conclusion, the gravity model of trade explains the size of bilateral trade flows as positively linked to the size of the trading countries’ GDP and negatively related to the
distance between them. Because trade makes up a large part of a country’s GDP (World Bank, 2018a), it appears evident that a large value of GDP should correlate with active participation in international trade. Hence, the assumption that the size of the trading countries’ GDP will have a positive effect on the bilateral trade flows appears to be reasonable.
3.2.1 Theoretical predictions on the effect of regime type on bilateral trade flows
As proposed by Mansfield et al. (2000), both democratic regimes and authoritarian regimes have incentives to trade, but they will implement different levels of trade barriers depending on their respective preferences. Democratic regimes, liberal and electoral, have to take the legislative as well as the chief executive’s preferences into account, whiles for authoritarian regimes the legislative equals the chief executive, hence there are no conflicting preferences domestically. All three regime types do, however, prefer the counterpart trading country to keep their trade barriers as low as possible. The countries’ domestic trade barriers will differ depending on the outcome of negotiations and public opinion. Consequently, there will be divergence in trade flows.
In accordance with the theoretical framework made up by Mansfield et al.’s (2000) model and the gravity model of trade, we predict that using regime type as a measure of distance between trading partners will lead to a inversive relationship with trade flows. Two regimes of the same type will apply lower trade barriers, hence trade more extensively which will lead to a positive effect on trade flows. Trade flows between mixed regime type-pairs types will correspondingly be lower due to higher trade barriers, as a result from the distance between the regime types being bigger. Hence, trade flows will be affected negatively if the trading partners are of different regime types. Our central hypothesis is thus:
H1: Country pairs of the same regime type will have higher trade flows than regime type- pairs consisting of mixed regimes.
This prediction applies to all three of the investigated regime types. Furthermore, the positive proportional relationship between trade flows and trading partners’ GDP implies that larger economies will trade more extensively and thus have higher trade flow values.
3.3 Literature review
In this section we present previous studies on the relationship between bilateral trade and regime type and what their findings. This thesis is primarily adjusted to test the results and conclusions made in the study by the aforementioned Mansfield et al. in 2000, using the same theoretical assumptions and the gravity model of trade to predict the relationship between regime type and trade flows. More recent studies on the effect of regime types on trade flows are also presented: a contribution made by Yu in 2010 using similar methodology to the one in our thesis, and two studies discussing the impact of different political institutions and institutional quality on trade, which allows us to distinguish between liberal and electoral regimes. Finally we present a recent study discussing the importance of disaggregation of regime types in economic research.
Table I
Summary of previous studies
Authors Geography and
studied time period
Studied variables (dependent
variable, main explanatory variable)
Empirical findings2
Mansfield, E., Milner, H., &
Rosendorff, P. (2000)
All countries in the interstate system, according to Correlates of War, 1960-1990
Trade barriers and regime type
Positive relationship between lower trade barriers and democratic regimes
Anderson, J., & Marcouiller, D. (2002)
58 countries worldwide, 1997
Trade levels and corruption (i.e. stability of political institutions)
Negative relationship
Milner, H., & Kubota, K.
(2005)
179 countries worldwide, 1970-1999
Trade liberalisation and democratisation
Positive relationship
Yu, M. (2010) 157 countries worldwide, 1962-1998
Trade flows and democratisation
Positive relationship.
Impact of regime type differs between importer and exporter
Mukand, S., & Rodrik, R.
(2017)
- Economic growth and
regime type
-
Note: The variables above are often only examples of examination variables used in the studies respectively. The article by Mukand & Rodrik (2017) does not provide an empirical study on the relationship between trade and regime type, but discuss disaggregation of regimes and its impact on results in economic research.
2 In “Empirical findings”, only statistically significant results of the studies are reported. Further description of results of the previous literature will follow.
Studies on international trade och regime type primarily target democracy or democratisation as a potential explanatory factor for variations in bilateral trade. Mansfield et al. (2000) were among the first to study the effect of regime type on trade using the gravity model, providing empirical evidence for the theoretically derived “hypothesis of democratic trade” – that democracies engage in more trade with other democratic regimes than with other regime types – consequently leading to higher trade flows. In their article, Mansfield et al. use trade barriers as the dependent variable. A key assumption made by the authors is that democratic states and authoritarian states share the same incentives to trade internationally, as we derived in Section 3.2. Regardless regime type, there are benefits of trade to gain and sustain political support. Data used in the study consisted of cross-sectional data for all dyads listed as members of the interstate system according to the Correlates of War Project between the years 1960-1990. Controlling for possible effects of the bipolar world politics during the Cold War, the study generated statistically significant estimates that agreed with the theoretically generated proposition. That is, aggregate trade barriers between the observed democratic pairs were lower than between mixed regime-pairs. This implied more extensive trade between two democracies. The extent to which democratic country-pairs trade more with each other also experienced an upward trend during the studied time period. In the 1990’s, pairs composed of two democratic regimes traded on average 40 per cent more than dyads composed of a democracy and an authoritarian regime.
In a more recent study performed by Yu (2010), robust evidence of democratisation being a boosting factor for trade is presented. Applying democratisation to the gravity model using a dataset consisting of cross-sectional data covering 157 countries between the years 1962- 1998, estimations implied that democratisation contributed a 3-4 per cent increase in global bilateral trade flows after controlling for endogenous effects. Moreover, Yu’s (2010) findings denote that democracy’s impact on trade flows differs between the importer and the exporter.
Democracy within the importing country lead to lower tariffs thus promotes trade. This is in line with the conclusion made by the aforementioned Mansfield et al. (2000). The impact of the exporter’s level of democracy on trade depends on several factors, institutional quality being one. Being a highly democratised country will be an asset in international trade as good institutional quality is an indicator of being trustworthy (Yu, 2010).
Furthermore, Anderson and Marcouiller (2002) concluded that unstable political institutions,
considerable decrease in trade levels. Their empirical model is based on data over 58 countries collected by the World Economic Forum in 1997, which is applied to the gravity model of trade. In accordance with the predictions of the gravity model, Anderson and Marcouiller (2002) showed that lowering transaction costs will consequently increase trade.
Transaction costs in this case refer to costs related to security issues and trustworthiness in the exporting part of the trading dyad. Results show that corruption and other insecurities across borders, primary authoritarian features, will constrain commerce in a way similar to tariffs. Thus, lack of institutional quality acts as a hidden tax on international trade. In sum, when trying to explain trade flows, costs associated with unstable institutions should be considered.
In the previously mentioned study by Yu (2010), it is argued that a fair market tends to care about consumer rights, product and labour regulations as well as legal constraints, favouring product quality and thus exports. Hence, highly democratised exporters have the possibility to enjoy reduced trade costs and potential to increase trade in the international market. This follows the argument that liberal democratic-pairs are able to apply lower tariffs as a result of stable and trustworthy political institutions and good product quality, made by Yu (2005). On the contrary, electoral democracies and authoritarian regimes correspond with uncertainty and higher trade costs as a result of poor political institutions, thus leading to lower bilateral trade flows.
How can disaggregation of democracy provide further explanation of the relationship between bilateral trade levels and regime type? Disaggregating regime types is common in the political science literature and often used to distinguish differences between democracy and authoritarianism. In economics however, it is still prevalent to use the deficient classification that is insufficient by the terms of reflecting the political state of the modern interstate system. The aforementioned Diamond (1997) argues that ignoring the gap between liberal and electoral democracies should be of great concern within policy making and comparative analysis. Mukand and Rodrik (2017) present some evidence to that liberal democratic regimes can only develop under quite special political economic circumstances, and thereby use and argue for the distinction between liberal and electoral democracy. In their framework, Mukand and Rodrik distinguishes between electoral and liberal democracy based on the fact that the elected leaders in a liberal democracy can not discriminate against individuals in terms of civil liberties. Thereby institutions such as the judiciary as well as
freedom of speech and media are not interfered by individual leaders’ interest in gaining or sustaining political power, that could impact national economic policies such as trade policy.
When analysing the effect of regime type on trade flows, internal, gradient differences in regime types could possibly capture determinants of bilateral trade levels more precisely than if differences are omitted. Using a time-series cross-section dataset including 179 developing countries, territories and dependencies between the years 1970-1999, Milner and Kubota (2005) found that democratisation of the political system is associated with trade liberalisation. Hence, their findings support Mansfield et al.’s “Model of Bilateral Trade Actors and Their Preferences”, implying that the political power shifts from an elite to the people, trading liberalisation will be promoted. Countries that encountered liberalisation in trade policy during the observed years had either been preceded by a democratisation or experienced the political transition simultaneously. Data indicated that the likelihood of reform in trade policies, that is a movement towards either a more liberal or protectionist approach on trade issues, differed depending on the type of established democratic institutions. By disaggregating regime type further, this could potentially be controlled for by examining the impact of different political institutions.
In sum, there is a consensus amongst economists that there is a significant relation between bilateral trade flows and regime type (Mansfield et al. 2000; Milner & Kubota, 2005; Yu, 2010). Discrepancy between pairs from being democratic and authoritarian inhibit trade by higher tariffs and problems with mis-trust based on poor institutional quality. By introducing a distinction between liberal and electoral democracies in accordance with contemporary literature on democracy, we enable for an analysis reflecting the modern interstate system using a distinction between liberal and electoral democracies and authoritarian regime systems (Diamond, 1997; Milner & Kubota, 2005; Mukand & Rodrik, 2017).
4. Data
We use data for trade flows and regime type indicators for the years of 2004 and 2014. By examining two years, we obtain a great number of observations of country pairs and trade flows (further described in Table II “Descriptive statistics”). Also, by using data for the years of 2004 and 2014, we use more recent data than previous research, thus update the existing literature (e.g. Mansfield et al., 2000; Milner & Kubota, 2005; Yu, 2010).
In this section we present the data and their sources that are used in the estimations of the empirical model. The included sources are discussed and put in comparison to alternative ones. Two issues concerning data are discussed more in depth; missing values for trade flows and the problematics in quantifying regime type. We begin by assessing common errors made in previous studies and refer to literature that has aimed to problematise these.
4.1 Baldwin and Taglioni’s Gold-, Silver- and Bronze-medal Mistakes of the Gravity model
Baldwin and Taglioni (2006) identifies three frequently made mistakes in earlier studies that has used the methodology of the gravity model, two of which are closely related to the selection of data. These errors are referred to as the gold-, silver- and bronze medal-mistakes of the gravity model. The “gold medal-mistake” is the most apparent of the three and occur when there is a correlation between the omitted variables and the term specifying the cost of trade, i.e. omitted variable bias (OVB). The consequences of OVB is a methodological issue, and will be further conceptualised in Section 5.4.
The most common mistake when applying the gravity model methodology materialises when researchers overestimate the bilateral trade flow data they are using. This is often referred to as the “silver medal-mistake”. Due to the available raw data for trade flows being unbalanced (further assessed in Section 4.2.2.), some studies use an average of the four reported trade values to get one average value of trade flow. However this often leads to aggregation bias in the results, especially when the two trading economies differ vastly in size, and should therefore be avoided (Baldwin and Taglioni, 2006; Erikson et al., 2009).
The third mistake, the so-called the “bronze medal-mistake”, emerge when nominal trade values are incorrectly deflated according to the U.S. aggregate price index. Using deflated
U.S. dollars will lead to biased estimations as a result from global trends in inflation rates leading to counterfeit correlations. By applying time fixed effects, this potential error can be offset. The time fixed effects corrects for deflation since the reported trade flows are divided by the same price index. However, the time fixed effects will also identify other idiosyncratic time-specific changes, so not only conversion factors will be recognised (Baldwin and Taglioni, 2006).
4.2 Data on the dependent variable - Trade flows
The source used for bilateral trade data is International Monetary Fund’s (IMF, 2018) publicly accessible Direction of Trade Statistics (DOTS), which is commonly used in bilateral trade research (Baldwin and Taglioni, 2006). DOTS include reported import and export values in current U.S. dollars, leading to a total of four observations of trade for each country pair. For estimations, we use the directional values of import. This is in line with most studies on bilateral trade, and is based on the premise that countries allocate more resources to measuring their imports than exports in order to avoid tariff fraud. By doing so, we also avoid to be exploited to the problematics of Baldwin and Taglioni’s silver medal- mistake (Baldwin and Taglioni, 2006).
The data used in this thesis includes UN member states with population size over 500,000.
The omittance of countries with populations under 500,000 are expected to have insignificant impact on our estimations due to their neglectable geographical and economic size, as concluded by previous studies within the field of international trade (see for example Mansfield et al. 2000).
4.2.1 Evaluation of data for control variables
This thesis include four categories of control variables; population, GDP, Freedom House status and country pair-common currency for the relevant years. Data on population size and GDP (in current U.S. dollars) has been adopted from the World Bank’s World Development Indicators (WDI, 2018b). The variable for Freedom House status (further on referred to as FH status) for the two trading countries i and j respectively, has been obtained from FH’s Freedom in the World Report (2004;2014a;2016), stating liberal democracies (taking on the value 3), electoral democracies (=2) and authoritarian regimes (=1). Taking the FH status of both countries in each country pair into account, six dummy variables are created to represent
all possible combinations of the three regime types. The variable indicating common currency has been created by adopting data from the Swiss Association of Standardization (2018), and has then been edited in order to get two sets of data with the correct currencies, matching the currencies for the countries in the 2004 and 2014 trade flow data respectively.
From this, a dummy variable, representing common currency has been created which takes on the value 1, if the two countries in the trading country pair charge the same currency (=0 if they do not).
Table II Descriptive statistics
Variable N Mean St. dev. Minimum Maximum
Trade flowij, t 37,925 6.76e+08 6.29e+09 1 4.67e+11
Freedom House status, country it-1 37,620 2.196863 0.8029166 1 3 Freedom House status, country jt-1 37,670 2.190948 0.8082756 1 3
Population, country it-1 37,597 5.05e+07 1.66e+08 520106 1.36e+09
Population, country jt-1 37,669 5.21e+07 1.66e+08 520106 1.36e+09
GDP, country it-1 36,883 4.76e+11 1.59e+12 3.18e+08 1.67e+13
GDP, country jt-1 36,771 4.82e+11 1.60e+12 3.18e+08 1.67e+13
Note: Data adopted from IMF; Freedom House; The World Bank. Note that all variables are not included in all models and that the included dummy variables (regime type-pairs, common currency) are not included in the table.
In Table II above, descriptive statistics are presented. The dependent variable trade flow represent the imports from one country to another. For the years 2004 and 2014, we obtain values for 162 countries in both years, making up a total of 37,925 trade flows. Note that most country pairs are accounted for twice, since the countries in the trading pair report trade flow values in the two directions separately. Moreover, only values of de facto reported trade has been included in the data, hence the minimum value of trade flow is 1 U.S. dollar. An assessment regarding missing values for the dependent variable as well as for the control variables will follow in the following section.
4.2.2 Assessment of missing values
An important issue to address is the amount of missing values in the data regarding the dependent variable, trade flow, as well as for the variables controlling for population and GDP. Coming across completely balanced data on trade flows is unfortunately difficult and is
a persistent problem for studies on international trade (Baldwin & Taglioni, 2006). The most established data used for trade statistics is DOTS by IMF which is used by many studies on international trade (Baldwin & Taglioni, 2006) and is thus the one used in this study.
Although noteworthy is that the missing data in DOTS is somewhat randomly scattered throughout the dataset, but missing trade flow values seem to be more common for warfare or less developed states, many who are authoritarian regimes. Hence, observations of electoral or liberal democracies are more frequent and well-represented in our data. As we have previously noted, not all world countries engage in trade with each other. Country pairs that has no reported values of trade are therefore not included in the data. This study estimates the log values of trade flow, and using values of trade flows reported as zero would lead to further problems when performing regressions (Santos Silva & Tenreyro, 2006). Previous studies has applied different methods of dealing with this issue, but the approach that has followed and been embraced by most studies using log trade values is to simply drop pair with zero trade from the data set (Santos Silva & Tenreyro, 2006). Consequently, we expect a positive bias regarding the trade flow of liberal democracies and a negative bias regarding the trade flow of authoritarian regimes due to these missing values. Although the missing values make it difficult to declare any results obtained in this thesis as unambiguously trustworthy, but we are still able get an indication of the direction of the explanatory variables’ and an approximate proportion of their corresponding effects.
4.3 Data on the main explanatory variable - Regime type
There is no single true measure of regime type due to its complexity and the fact that there are several definitions of democracy and authoritarianism, as well as being associated with reliability and subjectivity issues. Traditionally, measuring political freedom and regime types has been based on political institutions and procedures, in accordance with the minimal definition of democracy. In this section we present the two most frequently used democracy indicators in academia and discuss the overall problematics with quantifying regime types.
Finally, we present the chosen indicator and argue for why it is the most suitable one for the methodology of this thesis.
4.3.1 Evaluation of regime type indicators
Using the minimal definition of democracy is common when quantifying political freedom and regime types in literature on international relations (Mansfield et al., 2000), and is the
foundations of the academic standard measure when grading democracy, the Polity IV Project Index. The index is an extensive, annual measure of most countries in the interstate system, and cover the years 1800-2016 (Polity IV, 2016). Polity IV’s definition of democracy reflect three main elements; 1. Existence of institutions and procedures where citizens can express preferences about political parties, leaders and policies, i.e. participate in fair elections; 2. Citizens ability to constrain the executive and; 3. Guaranteed civil liberties.
However, civil liberties are not actually measured thus makes it difficult to distinguish between electoral and liberal democracies according to the previously presented regime type definitions. Although the Polity IV indicator has proven successful in previous research, its methodology does thus not seem to correspond with the information necessary to investigate the proposition made in our thesis.
As previously declared, levels of democracy and authoritarianism can be put on a gradient scale, reflecting differences within the regime types. It is thus not possible to say that regime is democratic or not democratic - or authoritarian or not authoritarian for that matter - but further distinction must be made in order to properly capture the effects of regime type on trade. The presence of civil liberties is thus important in order to distinguish liberal democracies from electoral democracies (and authoritarian regimes) in accordance with the literature (Diamond, 1997). In order to capture the effects of the dissimilarities between liberal and electoral democracies, an indicator incorporating the additional element that distinguishes between the two types of democracy – the measure of civil liberties – must be used.
Since 1972, independent watchdog organisation Freedom House (FH, 2018) has provided the annual report “Freedom in the World”. Together with the Polity IV measure of democracy, FH is the most commonly used indicator of diversification in regime type in academia. By employing an additional checklist of civil liberties, the report provides a continuous measure of countries’ substance of democratic regime. Further criteria include: free media, academic freedom, freedom of religion, independent judiciary and prevail for the rule of law, protection from political terror, social freedom, civilians right to property and freedom of assembly. By differentiating between political rights and civil liberties, the methodology employed by FH allows us to analyse differences within regime types.
Countries are scored on separate scales for civil liberties and political rights respectively, ranging from 1-7 (where 1 is most free and 7 least free) according to how they reflect the political rights and civil liberties criteria. The classifications “Free”, “Not Free” and “Partly Free” are then designated to all countries’ regimes according to the aggregate score of the two measures.3 In general, countries assigned the status “Free” are regarded as liberal democracies. Authoritarian regimes are unambiguously categorised as “Not Free”. Countries that fulfill the criteria for the minimal definition of democracy and also supply a basic level of political rights are designated the label electoral democracies. More specifically, to qualify for the designation electoral democracy, the regime must: 1. Be a competitive multiparty political system allowing; 2. Regular, fairly conducted elections; 3. Allowing for public access to political parties through medias and open campaigning; 4. Provide universal legal- rights to for all adult citizens. Electoral democracies tend to qualify as “Partly Free”
according to FH, hence these are going to be regarded equally in our study (Freedom House, 2014b). A full list of civil liberties and political rights adopted in the FH methodology can be found in Appendix I.
In 2014, 45 per cent of the world’s countries were classified as “Free” and thus liberal democracies, including countries such as the United States, India and all EU member states (Freedom House, 2014a). Examples of countries labeled as “Partly Free” by FH in 2014 were Indonesia, Mexico, Turkey and Ukraine. Noteworthy is that the vast majority of electoral regimes in 2014 were recently transitioned democracies (Freedom House, 2014a). Examples of countries listed as “Not Free” by FH in 2014 were China, Russia, Sudan and Cuba (Freedom House, 2014a). China and Russia, along with authoritarian state Saudi Arabia are consequently to be found on the list of the world’s twenty biggest exporters (World Bank, 2014).
In conclusion, in order to properly capture the effect of regime type on trade it is not sufficient to use a binary definition when differentiating between democracy and authoritarianism, taking into account that this does not reflect the modern interstate system.
By disaggregating democracy using the FH methodology distinguishing between political
3 “The average of a country’s or territory’s political rights and civil liberties ratings is called the Freedom Rating, and it is this figure that determines the status of Free (1.0 to 2.5), Partly Free (3.0 to 5.0), or Not Free (5.5 to 7.0)”, (Freedom House,
rights and civil liberties, we enable further and more precise analysis of the effect of regime type on trade.
Figure I
Number of liberal democracies, electoral democracies and authoritarian regimes according to Freedom House, 2003-2014
Note: Data adopted from Freedom House, 2016.
As pictured above in Figure I, the number of liberal democracies has been a fairly stable between the years 2003 and 2014, with few examples of de-democratisation.4 The interpretation of the number of electoral democracies and authoritarian regimes is somewhat more complex. During the studied time period there seem to be an inverse relationship between authoritarian regimes and electoral democracies, with countries pending between being classified as electoral democracies and authoritarian states. Levitsky and Way (2002, p. 52) present a potential explanation: “Although scholars have characterized many of these regimes [i.e. electoral democracies] as partial or ‘diminished’ forms of democracy, we agree with Juan Linz that they may be better described as a (diminished) form of authoritarianism.”.
The gap between electoral democracies and authoritarian regimes is thus smaller than the one between liberal democracy and electoral democracy. With these types of regimes generally
4 To get an overlook of the distribution of regime types globally for the two studied years, world maps indicating FH classifications are presented in Appendix I, Figure AI (2004) and Figure AII (2014).
30 40 50 60 70 80 90 100
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Number of Countries
Liberal democracy Electoral democracy Authoritarian regime
being associated with less developed political institutions (Mukand & Rodrik, 2017), instability could lead to classifications being more volatile than for liberal democracies thus explain the inverse relationship.
Table III
Matrix of aggregated number of trade flow observations according to regime type-pairs
Regime type, % of N Exporting country, j Regime type, % of N
Importing country, i
Authoritarian regime
Electoral democracy
Liberal
democracy Total, % of N
Authoritarian regime 5.6 (2,111) 7.2 (2,687) 11.5 (4,301) 24.4 (9,099)
Electoral democracy 7.8 (2,913) 9.6 (3,582) 14.3 (5,342) 31.7 (11,837)
Liberal democracy 11.5 (4,302) 14.2 (5,324) 18.2 (6,804) 43.8 (16,430)
Total, % of N 24.9 (9,326) 31.0 (11,593) 44.0 (16,447) N = 37,366
Note: Reported unit in the matrix is relative frequency (%) of regime type observations in relation to reported trade flow values. Total number of observations given in parenthesis. Regime type is stated in accordance with Freedom House status for the years 2004 and 2014 (Freedom House, 2016).
Table III above displays regime type frequencies in trade flow observations. The total number of trade flow observations (N) for the two observed years, 2004 and 2014, is 37,366. In the vertical column the importing country i’s regime type is determined, while as the counterpart, the exporting country j in the trading pair, is displayed horizontally. The table shows that the most common combination of regime type-pairs are ones where both trading partners are liberal democracies (18.2 per cent), which is in line with our predictions and the fact that liberal democracy is the prevalent regime type worldwide, as seen in Figure I, page 23.
Authoritarian countries, however, are engaged in trade with more democratic countries than with other authoritarian regimes (5.6 per cent for authoritarian regime-pairs). This could potentially be explained by authoritarian regimes being the least frequent regime type in the observed years, leading up to fewer observations as noted in Figure 1, page 23.
4.3.2 Assessment on the problematics in quantifying regime type
As previously mentioned, there is not one accepted way of distinguishing between regime types, hence there are straightforward issues when quantifying regime type, being a categorical variable. In order to enable for estimations, categorical variables (such as regime type) must the quantified. Interpretation of categorical variables should be done carefully, especially in the case of a variable such as regime type that is associated with ambiguity and lacks completely accepted definitions. However, by comparing democracy indicators we get an indication of the robustness of the regime type indicator. If countries’ regimes are systematically categorised in a similar pattern according to more than one indicator, results could potentially be interpreted with more confidence. Polity IV Index and Freedom House measures of democracy are strongly correlated (Högström, 2013). As previously noted however, Polity IV Index do not apply the distinction between liberal and electoral democracy. Thus, in accordance with our methodology, we effectively adopt the distinction between liberal democracies, electoral democracies and authoritarian regimes according to FHs’ classifications. Thereby, no discretionary distinctions have been made. In conclusion, to make up for the problematics in quantifying regime type, several measures have been adopted in order to test for these. However, as we have previously emphasised: since there is not one fully accepted way to neither define nor quantify regime type, all estimates and results should be interpreted with caution.
5. Empirical method
In the following section we will discuss and present the empirical methodology used in order to test the presented hypotheses. Methodologically, this thesis applies the same functions as most literature using the gravity model of trade (see for example Mansfield et al., 2000;
Anderson & Marcouiller, 2002; Milner & Kubota, 2005; Yu, 2010). To obtain estimates for parameters that could explain differences in trade flows and different combination of trade partner regimes we will perform conventional ordinary least-squares regression analysis (OLS). OLS is the dominating regression method in the gravity model literature thanks to its simplicity and intuitivity, as well as having provided successful results in previous research (e.g., Mansfield et al., 2000; Anderson & Marcouiller, 2002; Milner & Kubota, 2005; Yu, 2010). In the regressions we use robust clustered standard errors, which allows for correlation within a cluster. In this case, we have data on trade flows between trading country pairs,
hence we cluster the country pairs. If we apply default standard errors to such setting, we obtain a risk of overstating estimate precision (Stock & Watson, 2015).
5.1 Year- and country pair-specific fixed effects
In order to detect unobserved time- and country pair-specific effects, we use fixed effects. We use year-specific fixed effects, since it makes it possible to control for unobserved variables that vary over time, but not between states. Country pair-specific fixed effects are applied in order to control for unobserved variables that vary between states, but not over time (Stock &
Watson, 2015). A range of examples of variables that have been included as independent variables in previous studies, but do not add any interpretation value when using country pair-specific fixed effects are: common border, common language, historical colonial relationship, and geographical distance (Baldwin & Taglioni, 2006) among other characteristics. A possible limitation when using fixed effects is that there needs to be ample variation in the explanatory variable over time, within the country pairs, which should be the case in our data, due to its size and construction. An alternative method would be to perform pooled regression and include year dummies to control for time-specific effects. However, findings have shown that performing pooled regression lead to systematically biased estimates when adopting the methodology of the gravity model and is therefore not recommended (Green et al., 2001).
5.2 Evaluation of additional control variables and methodological adaptations
In accordance with the gravity model of trade, we include variables controlling for country pairs’ GDP (denoted ln gdpi and ln gdpj) and population (denoted ln popi and ln popj), and expect estimates to be positive following the predictions of the gravity model. In order to facilitate the interpretation of the variable estimates in relation to the dependent variable, natural logarithms are used. When interpreting logged variable estimates, the parameters of the independent variables are equivalent to the elasticity of trade flow in a change in the control variables respectively. This adaption corresponds with the empirical methodology of previous literature (Santos Silva & Tenreyro, 2006).
Furthermore we use time lags in order to capture the effect of the explanatory variables on the dependent variable, trade flow, at later point in time. That is, we can observe the stability of the effect regime type has on trade flows (Stock & Watson, 2015). Time lags of one year (t-1)
are adopted to control for delayed effects of the included variables. Since this thesis do not aim to investigate changes in regime type and trade over time, e.g. the effects of a regime transition on trade or the effect of democratic consolidation after democratisation, we do not apply any additional time lags to control for effects deriving from further back in time.
Finally, we include a dummy variable to indicate common currency in both models (denoted comcur). Although it can be argued that there is not much variation in common currency in general, hence it could be potentially be incorporated in either time specific- or country pair- specific fixed effects, but we note that 18 countries had a change in currency between the two years observed in our study. Four of the countries experiencing a change in currency were European countries adopting the Euro in favour of their national currency, the remaining countries were non-European countries whose currency change neither brought them in nor out of a currency union. Due to this, the effect of common currency can not be regarded as fixed in the observed years. Rose and van Wincoop (2001) find national currencies to have a considerably negative effect on trade, whereas currency unions seems to boost international trade. As for the previously mentioned problem with institutional quality and its impact on trade barriers and thus trade flows, currency union may be an efficient tool to combat trust issues and should therefore increase trade flows (Barro & Alesina, 2000). The variable indicating common currency is adopted with a time lag of one year (t-1) to control for delayed time effects, in accordance with the other included control variables.
