J
Ö N K Ö P I N GI
N T E R N A T I O N A LB
U S I N E S SS
C H O O L JÖNKÖPI NG UNIVER SITYTrade Restrictiveness or Trade Openness?
The Effects on Corruption
A panel data study of the relation between trade restrictiveness and corruption in
Europe
Paper within Economics
Author: Alisa Senderovic, 860819-9025 Tutor: Börje Johansson
Tobias Dahlström Jönköping June 2009
i
Bachelor Thesis within Economics
Title: Trade Restrictiveness or Trade Openness? The Effects on Corruption.
- A panel-data study of the relation between trade restrictiveness and corruption in Europe
Authors: Alisa Senderovic Supervisors: Börje Johansson Tobias Dahlström
JEL-classifications: F10, D73, D40, D43, D72 Date: 2009-05-29
Keywords: Trade, corruption, competition, regulations, rent-seeking
Abstract
This thesis analyzes the relationship between trade restrictiveness and corruption levels in Europe and tests the robustness of the results using two different measures of corruption; Corruption Perceptions Index constructed by Transparency International and World Governance Indicators; Control of Corruption constructed by the World Bank. The results show that the outcome differs among the two indices and that previous results found in studies may be subject to data choice and measurement errors. A majority of previous studies have used trade openness in the form of imports share in GDP, or trade openness indices as variables that explain corruption levels. This thesis focuses on trade restrictiveness. It also investigates the differences between restrictiveness and openness in their impacts on corruption. The author of this thesis finds a weak relation between trade restrictiveness and corruption on one hand and an even weaker relation between trade openness and corruption on the other hand when controlling for other variables that may have an impact on corruption. The limited time-frame could be an explanation for this result, i.e. trade policy effects. A longer time-frame would have probably resulted in a bigger difference between variables for trade openness and trade restrictiveness. It is found that the variables showing to have the largest impact on corruption levels in Europe are those associated with historical dimensions such as whether the country has had a tradition of the church being separated from the state or whether being a previously planned economy. Most importantly, the thesis presents suggestive evidence on the fact that not all government involvement increases corruption. Rather, it is regulations that limit internal competition have a large impact on corruption levels in Europe. It has also been found that not all government involvement in the economy implies higher corruption levels.
The findings outlined are in line with the common perception that corruption is highly dependent on previous values of corruption and that radical policy interventions are needed to curb corruption.
Table of Contents
1
Introduction ... 1
1.1 Purpose ... 2 1.2 Outline ... 22
Background ... 3
2.1 Corruption ... 32.2 Trade freedom or trade restriction? ... 6
3
Theoretical Framework ... 8
3.1 Theoretical treatment of corruption ... 8
3.1.1 Principal-agent-client problem ... 8
3.1.2 Rent-seeking ... 9
3.2 Corruption through limits on imports –the linkage ...10
3.2.1 Policy implications ...10
3.2.2 Critique of the theory ...11
4
Data and Methodology ... 11
5
Empirical findings ... 14
5.1 Import restrictions and corruption ...14
5.2 Regulations and corruption ...16
5.2.1 Fixed effects ...19
5.3 Country- and region specific effects and cultural factors ...20
6
Conclusions ... 24
7
References ... 25
8
Appendices... 28
8.1 Corruption levels in Europe ...28
8.2 Economic Freedom in Europe ...28
8.3 Correlation matrix between CPI and WGI Control of Corruption ...29
8.4 Correlation between explanatory variables ...29
8.5 The relationship between corruption and GDP/capita in 2007 ...29
8.6 Lagged corruption as explanatory variables ...30
8.7 Raw data on tariffs: panel vs. cross-section ...30
8.8 Fixed effects models using the log of GDP per capita ...31
8.9 Inclusion of Natural Resource Abundance in Fixed Effects models ...31
8.10 Final models with auto-correlation corrected coefficients ...32
iii
Figures
Figure 2.1. An ideal measure of trade restrictiveness...7 Figure 3.1. Principal-agent-client model...8
Tables
Table 2.1. Different types of corrupt government systems...3 Table 4.1. Correlation between variables measuring trade openness and policy..13 Table 5.1. Trade restrictions, GDP per capita and Corruption levels...15 Table 5.2. Regulations and the effects on Corruption...18 Table 5.3. Fixed effects models...20 Table 5.4. Final models –explanatory variables including dummies for country-specific effects...22
1 Introduction
Europe is a diverse region in various ways. Differing cultures, levels of natural resource abundance, political systems and stages of development are represented in the region of Europe. The level of corruption is a further factor that differs among the countries. Generally speaking, the countries of Western Europe, or member-countries of the European Union, show lower corruption levels than the countries that have not yet entered the European Union, or who did so just recently1. Meanwhile, the countries of the European Union all have an open market and are highly open to trade. In fact, they can be considered to be open to trade only internally within the EU, but since a majority of trade activities take place among EU member countries internally, the member countries of the European Union can be considered to be highly open to trade (EUROSTAT, 2009). In the meantime, countries outside EU all exhibit higher barriers to trade and more restrictive trade policies2. Could the fact that non-member countries of the European Union are more restrictive to trade be the explanation for why these countries all face higher corruption levels?
Free trade leaves little or no room for policy discretion and should therefore minimize corruption. A protectionist trade policy that limits imports creates rent-seeking behavior (Krueger, 1974). The definition of customs applying to different products becomes subject to political influence and thus gives officials substantial discretionary power. An anonymous Thai businessman said “If I was born again I would like to be a customs official” (Gatti, 1999) which illustrates the effect of trade restrictiveness on increased discretionary power and thus corruption. Increased exports can also decrease the corruption levels if export firms are located in the destination country and if their share is large relative to domestic firms and can thereby spread a different business culture (Johnson & Dahlström, 2007). However, this thesis will only deal with trade restrictiveness from an import-approach. Import restrictions on consumer products, in the form of tariffs, outside the EU, and non-tariff barriers, both in the EU and Europe as a whole, are still implemented and justifies further research on this subject. Plenty of real-world examples where limits on imports have lead to increased corruption may be brought up. One of them is a case study of Mali and Senegal where Stasavage and Daubrée (1998) find that high tariff levels in these countries have promoted customs fraud by increasing the incentive for private citizens to engage in illegal behaviour and for government officials to abuse their power for personal gain. This paper argues that the creation of rents through trade restrictions in the form of limits on imports is associated with higher levels of corruption. Therefore the natural result would be that countries that have less restrictive trade policies also are less corrupt, when controlling for differentials in GDP per capita levels and other possibly influential factors.
A majority of studies that examine how trade affects corruption use imports share in GDP as a measure of trade restrictiveness. Fewer studies measure the actual level of trade restriction. This paper is an attempt to narrow this gap in the literature. An analysis of trade restrictiveness and its effect on corruption is carried out, using two measures of corruption as a robustness check and indicators of trade freedom which uses hard data such as tariff and non-tariff barriers. Further explanatory variables, such as one capturing European specific effects, are included in the analysis in order to strengthen the analysis of the relation between trade restrictiveness and corruption levels in Europe.
1 See appendix 1.
2
The differing levels in Economic Freedom, which is a measure that incorporates both internal and external measures of regulations, are shown in Appendix 2.
1.1 Purpose
The purpose of this thesis is to empirically analyze the effect of trade restrictions on corruption in Europe as a whole, and to investigate whether there are any region-specific variables that apply to Europe which help explain the variance in corruption levels in Europe. This forms the research question: How does restrictiveness of the economy to trade affect the level of corruption? Many studies have used measures of trade openness such as imports share in GDP, but the purpose of this thesis is to examine the effects of trade restrictions on corruption.
Hypothesis (1):
Across all the European countries as well as across time, corruption is higher for those countries that are more restrictive to trade than for those that are more open to trade, when controlling for differences in GDP per capita.Hypothesis (1.1):
Trade restrictiveness, measured by tariff levels, is associated with higher corruption levels.Hypothesis (1.2):
Trade openness, measured by imports share in GDP, is associated with lower corruption levels.The results are then tested further to include other variables which can be regarded as regulations, according to previous research in order to deepen the analysis, and forms the second research hypothesis.
Hypothesis (2):
Regulations are associated with higher levels of corruption.Hypothesis (2.1):
There are country-specific effects that help explain the varying corruption levels in Europe.If the hypothesis 2 and 2.1 are not rejected, it is legitimate to expand the analysis to investigate what those fixed effects variables may be and if the results change when including these variables. This forms the third hypothesis.
Hypothesis (3):
Higher degree of regulations is associated with higher levels of corruption when controlling for region- and country- specific variables and cultural factors. The purpose of this thesis is not to investigate whether rents lead to corruption, although some comments on this aspect will be made under the theoretical framework. First of all, rents can be hard to measure, and second several studies have shown that the level of rents in an economy is associated with higher corruption (Luechinger et al. 2005). This paper assumes that rents lead to corruption, and will therefore use corruption levels as a proxy for rents. The purpose is thus to examine if more restrictive trade policies are associated with higher corruption levels, and if this shown not to be the case –are regulations in general an explanation for corruption, and are there any region specific variables that can explain corruption in Europe?1.2 Outline
The thesis is organized as follows; chapter two examines the concepts of corruption on one hand and trade restrictiveness versus trade freedom on the other hand. The third chapter considers theoretical aspects linked to corruption and trade restrictions with the main focus on the rent- seeking theories, and defines a theoretical perspective based on the import-restrictions theory. In chapter four, the methodology used to carry through the empirical analysis is explained. The empirical findings and results are presented and analyzed in chapter five. The relationship between corruption levels and trade restrictions as well as regulations is examined in a panel regression of the European countries. The analysis is then expanded into including more variables which may influence corruption levels according to previous studies. Finally, chapter five concludes the thesis and suggestions for further research are given.
2 Background
In this chapter the background and previous studies on corruption and trade will be presented. The first section deals with corruption while the second sections discusses trade restrictiveness vs. trade openness.
2.1 Corruption
Defining corruption goes beyond the scope of this paper, although many attempts have been done in defining and categorizing corruption. Bhagwati (1982) tried to categorize different forms of corruption calling them Directly Unproductive Profit-seeking (DUP) activities. Most would agree on the view that corruption is the abuse of public authority for private gain and can take the form of public officials who distort government resources for private gain3. The bureaucrat
can either steal government funds directly, or use his discretionary power to extract bribes from the citizens in exchange for favourable government services. The latter part usually comes about when individuals or groups attempt to increase profits at the expense of others without making any contributions to productivity, and is known in the economic literature as rent-seeking.
Furthermore, corruption can occur at different levels of government. Rose-Ackerman (1999) summarizes and categorizes four types of political corruption societies.
Table 2.1. Different types of corrupt government systems
Multiple bribers Few bribers Bribe recipients at top of
government
Kleptocracy Bilateral Monopoly
Multiple bribe recipients at low levels of government
Competitive bribery with a possibility of spirals
Mafia-dominated state
This paper will primarily deal with the category competitive bribery with a possibility of spirals, both theoretically and empirically. Corruption can occur wherever there are rents. Rents can occur for several reasons such as natural resource abundance (Leite and Weidmann, 1999), limits on competition (Ades and Di Tella, 1999), and import restrictions (Kruger, 1974 and Bhagwati (1982). Krueger (1974) points to the fact that imports were the main source of rents out of different factors. Rent-seeking can occur whenever there is possibility for competition for some existing value, whether it is competition for subsidies, ,the right to produce or sell products or the right to import products and thus gain a profit by buying the good cheap in a foreign market and selling in more expensive in the domestic market. However, not all rent-seeking needs to be corruption. As pointed out by Arthur Goldsmith (1999) for rent-rent-seeking to be classified as corruption there needs to be a violation of the law. Lobbying for special interests such as subsidies to those who are in need is considered as rent-seeking since the outcome of such activities may benefit some at the expense of others without contributing to productivity, if it is done in a legal way it does not need to be corruption. On the other hand, a health inspector who receives bribes from a restaurant manager in order to report the restaurant as clean when it is in fact hazardous according to the health standards stipulated by the law, is rent-seeking as well as illegal and is therefore classified as corruption, according to Goldsmith (1999).
3 This is similar to the definition used by the World Bank which states that corruption is the use of public power
Shleifer and Vishny (1993) distinguish between a monopoly providing all the goods and extorting bribes, several agents who extort bribes where the agents act independently of each other hence failing to take into account the level of bribes of other agents, and a competitive-market of provision of government goods where at least two agents can supply each and every government good or service. The first case can be ascribed to mafia-ruled societies such as pre-communist Russia where bribes were channeled through local Communist party offices where as the second case is most prevailing in post-communist Russia, some African countries and India. In many countries of the Western Balkans for example, starting a business requires bribing different authorities individually such as the fire authorities, the central ministry, tax authorities etc (Global competitiveness report, 2008). Refraining from paying these bribes may well result in the authorities using their discretionary power and denying the buyer the government license. In 2004, Swedish company IKEA was forced to close its newly opened department store in Moscow, because of issues with the local authorities (Realtid, 2004). The third case of corruption is present in countries with lower-level corruption where the level of bribes is relatively low because buyers of services can easily switch to another supplier of government services who will charge a lower bribe, such as drivers licenses, if the first agent tries to charge a high bribe. This will drive down the level of bribes.
The common perception now is that corruption has a negative effect on growth and trade45. An early opponent to the view of corruption as being advantageous to society was Pareto (1896). He argued that the market for political favours prevents productivity by using resources that could have been used to improve overall welfare through production, instead of just benefitting an individual or group of individuals. The view of corruption as being beneficial to society in reducing bureaucratic inefficiency is also nowadays dismissed. If political heads are rational individuals then they will seek more ways to maximise their income and create even more corruption (Rose-Ackerman, 1978). For example they may want to control lower level leaders to prevent them from receiving bribes to save resources for the top leaders. However, engaging in corruption may mean the government does not get re-elected next period, whereby the high-level official risks losing the future income if involving in corruption. In certain cases of the world, governments have been involved in this grand-type of corruption6. This implies more waste of resources to society.
Theoretically, the effect of the level of rents in an economy on corruption is somewhat ambiguous. Less competition induced through the restriction of imports means bureaucrats can exert more discretionary power and thus facilitate bribery. On the other hand, less competition means it is more valuable for the society to avoid corruption and thus it is more likely that the society tries to control the bureaucrat (Ades and Di Tella, 1999). However, Luechinger et. al find that the level of rents in an economy, measured by life satisfaction differentials between employees in the public sector and employees in the private sector, is closely correlated with most corruption indices (Luechinger et al. 2008).
Why then, is corruption so persistent in countries with high corruption? Mauro (2004) provides a model that explains why corruption is so persistent and suggests that corruption and political instability may be “two sides of the same coin”. He provides a game-theoretical approach to why corruption may come about, where individuals (politicians) need to take into account the fact that if they hurt the economy today, the citizens will not re-elect the government in the future which will deprive them of their ability to collect bribes in the future. The model
4 See for example Mauro (1995).
5 An influential, though now somewhat outdated, approach to corruption has been that corruption is advantageous
to growth. Advocates of this approach have argued that corruption serves as the “grease” of the economy, reducing red tape and bureaucratic inefficiency (Leff, 1964 and Huntington, 1968).
6 Several authors make the distinction between grand and petty corruption, see Bhagwati (1982) and Krueger
assumes that citizens are unaware of who is the corrupt one, but they are aware of that the government as a whole is corrupt. The resulting outcomes are that if individual A is very corrupt, individual B knows that the government will not be re-elected and therefore individual B becomes more corrupt in order to counteract the fact that he will not be able to collect bribes in the future. Hence, if one individual is not very corrupt neither will the other individual be. The implications of this is that countries left to their own devices may fall deeper and deeper into corruption if not drastic reforms are conducted to help them exit the vicious circle. This model is similar to that of Murphy, Shleifer and Vishny (1993) which comes to the conclusion that if many people steal, the probability of getting caught is low and therefore even more people will steal. Tirole (1996) analyzes the relationship of a group and its individuals where an individual’s behaviour cannot be identified in a group and therefore the behaviour of an individual is dependent of the group behaviour in the past to which the individual belongs. If the group is known to be corrupt, the individual will have strong incentives to be corrupt as well. This implies that countries‟ corruption levels depend on past corruption levels, and may persist even when the individuals who initialized the corruption are dead (Tirole, 1996).
These models have in common the fact that they (some implicitly and others explicitly) assume multiple equilibria. There is one “good” equilibrium towards which countries that have initial favourable conditions tend to move, and a “bad” one towards which countries who have initial unfavourable conditions move to. Mauro (2004) argues that the bad equilibrium is more likely to be stable when productivity is low and the public sector is large. This evidence is consistent with empirical evidence which shows that richer countries generally have lower levels of corruption, and as according to Tanzi (1998) large public sectors are generally associated with higher levels of corruption. Mauro argues that following the implications from the model, economic, institutional, political vari ables and even preferences can contribute to determining whether a good equilibrium can exist (Mauro, 2004).
The studies above deal with the link between corruption, trade and growth. However, there are a few known studies that examine the reverse link, namely the effect of trade on corruption. Tavares and Larrain (2004) empirically analyze trade and its effect on corruption but as a proxy for trade they use FDI. They find that FDI as a share of GDP is significantly associated with lower levels of corruption, regardless of import levels. They conclude that when including FDI/GDP in the regression import intensity is not negatively associated with corruption and that the case may be that previous studies on the link between imports and corruption suffer from omitted variable bias.
Ades and Di Tella (1999) analyze whether rents, induced by either natural rents such as oil or rents induced by lack of competition, are associated with higher corruption. They use share of imports in GDP but as a corruption measure they use Business International‟s corruption index and World Competitiveness Index which is directed more towards the business community and may thus influence the choice of countries covered since it may be biased towards countries that are more interesting from an investors point of view. In this article we are interested in policy discretion form of corruption –corruption of the political agent. Most studies on corruption recognize the endogeneity problem. For instance, it is plausible that trade freedom decreases corruption but it may be equally plausible that the level of trade freedom is the result of corruptive practices and protectionist policies. For the simultaneity bias to hold, the bureaucrat receiving the bribe, reflected in the corruption measure, must also be able to directly affect the level of trade restrictiveness through trade policy which would be reflected in the variable that measures trade restrictiveness and/or openness. Ades and DiTella (1999) investigate the possibility of simultaneity by using two stage least squares regression methods and find that the coefficient on import as a share of GDP is still significant and negative. However, there is a possibility that this result suffers from omitted variable bias as according to Tavares and Larrain (2004). Ades and Di Tella (1999) also find different results on the coefficient for GDP per capita. In some cases the coefficient is even significantly positive. This
may confirm the theory of Mauro (2004) that for countries with unfavourable conditions such as high inequality, an increase in GDP per capita may increase corruption whereas in other cases it leads to a decrease in corruption.
Kenneth S. Chan (2002) analyzes the effect of trade on bureaucratic efficiency. His main argument to why trade would improve bureaucratic efficiency is that small improvements in the performance of government services, many of which are inputs to the private sector will raise the productivity of export goods which will increase export earnings. Since the government can reap the benefits of these increased export earnings through taxes, there is an incentive for governments to improve bureaucratic efficiency, if the traded sector is large enough. However, as a measure of trade Chan uses both import and export share. It is not clear how, according to him, increased imports would lead to government efficiency since the reason for incentives existing for governments to improve bureaucratic efficiency was that they can reap some of the benefits of higher export earnings. For the government to be able to reap the benefits from imports, the level of tariffs would need to be high which in turn would decrease the level of imports. Furthermore, if corruption is present, higher levels imports may not lead to improved government efficiency since lower level-officials reap the benefits from the imports. Chan uses the average years of schooling, the sort of democratic institution and the log size of the population to serve as instrumental variables in his specification. However, the only of these variables that can be argued to be reasonably exogenous is the log size of the population. The other two are likely to be affected by bureaucratic efficiency. This analysis is one of the few analyses that considers the possibility that bureaucratic efficiency may be a function of previous degrees of bureaucratic efficiency, as according to Mauro (2004). He includes lagged variables of corruption as explanatory variables in the regression. Chan uses three data sets, none of which is Transparency International’s CPI which will be used in this thesis.
2.2 Trade freedom or trade restriction?
In the literature and empirical studies, the term trade openness is commonly used. There is a common understanding that trade openness refers to low barriers to trade, beyond this the term trade openness lacks clear definition. Thus far, no universally accepted definition for the term trade openness has been identified. Pritchett (1996) examines the correlations of different measures of trade openness and investigates “whether the economists’ intuitive notion of policy outward orientation can be captured empirically” (Pritchett, 1996, p. 308). He points out the fact that if the different measures for trade openness commonly used are correlated with each other they are capturing a common effect of countries trade policies. If not correlated, this means that empirical results using trade openness are subject to what measure of trade openness is used and in some cases may even be invalid.
Many empirical studies have tried to analyze the effect of trade openness on corruption by including either FDI share of GDP or share of imports in GDP (Treisman, 2000 and Knack & Azfar, 2003), sometimes without explaining the theory underlying the analysis. Using FDI/GDP as a measure of openness makes intuitively sense. The reason why it could lead to less corruption has been shown by (Larrain and Tavares, 2004). However, the use of imports share in GDP as an explanation for less corruption is ambiguous, if not implausible. No theories have been developed that suggest that the level of imports in a country relative to its GDP would lead to lower corruption. Pritchett (1996), points out that trade intensity measures, such as imports share in GDP or total trade share in GDP should and cannot be used as proxies for trade policies for the reason that trade intensity among countries differs for other reasons than differing policies. It is however plausible that some part of the policy measure is reflected in a measure of trade openness. Although it is plausible that imports share in GDP are not highly dependent on policies but rather on import demand and prices, it is incorrect to state that imports share in GDP is completely independent of trade policies. Higher tariff rates may lead to some part of imports share in GDP decreasing, all else equal. The existing theories
on the link between imports and corruption emphasize that it is the trade restriction itself that leads to lower corruption, that
would lead to higher corruption. Even though imports depend
countries’ lower import levels could merely be the reflection of lower import demand. To counteract the problem of endogeneity, imports share in GDP could be a more appropriate measure of trade restrictiveness. However, measu
better fulfil the purpose of analyzing how import restrictions can lead to rent corruption.
Ideally, a measure of trade restrictiveness would take into account the extent to which government policies affect the flow of goods
nominator measures the flow o
measures what the flow of goods and services would have been under free
Figure 2.1. An ideal measure of trade restrictiveness The obvious problem with such an index is that
would one know what the flow of goods and services under free trade policy was not present?
The measures of trade openness and/or trade restrictiveness considered in this thesis are
and physical measures of trade restrictiveness. Natural barriers to trade, such as trade distance and access to harbours are not covered in this thesis for the simple reason that it is difficult to find a theoretical direct link between these facto
The link between import restrictions and corruption may not always be clear to everyone. An illustrative example is Nigeria where in the early 1980
the IMF, but were never carried o
for profits (Rose-Ackerman, 1999). Import
discretionary power, and create opportunities for competition for some existing value namely the value that can be gained from buying a good cheaply in a foreign country and selling it more expensively in the domestic market. The rent
states that trade restrictions are competition
seeking7. For this to be true, i.e. protecting domestic firms from foreign competition, it must imply that countries both produce and import goods from the same industry, so called, intra industry trade. Studies have shown that this is very muc
countries8. Therefore, import-restricting policies in Europe would mean reducing competition in the domestic country with the possibility of creating rent
industry trade increases the cor
changes in net exports should be small which in turn makes this measure inappropriate for analysis.
Deregulation is, as proposed by for example Goldsmith (1999), a way to limit opportunities for corrupt officials to grant exceptions to the rules. Limiting competition through import restrictions means a higher degree of regulation, hence higher customs burden and thus
7 See especially Ades and Di Tella (1999) 8 See for example OECD Economic
high and stable or increasing levels o
on the link between imports and corruption emphasize that it is the trade restriction itself that leads to lower corruption, that is, if import demand is fixed, restricting imports would lead to higher corruption. Even though imports depend on trade restrictions,
lower import levels could merely be the reflection of lower import demand. lem of endogeneity, imports share in GDP could be a more appropriate measure of trade restrictiveness. However, measures that reflect trade policy, better fulfil the purpose of analyzing how import restrictions can lead to
rent-Ideally, a measure of trade restrictiveness would take into account the extent to which government policies affect the flow of goods and services (Suranovic, 1997
nominator measures the flow of goods and service with the restriction and the denominator measures what the flow of goods and services would have been under free-trade (Figure 2.1).
Figure 2.1. An ideal measure of trade restrictiveness
The obvious problem with such an index is that it is virtually impossible to construct. How would one know what the flow of goods and services under free-trade would have been if the
The measures of trade openness and/or trade restrictiveness considered in this thesis are
and physical measures of trade restrictiveness. Natural barriers to trade, such as trade distance and access to harbours are not covered in this thesis for the simple reason that it is difficult to find a theoretical direct link between these factors and the measures of corruption.
The link between import restrictions and corruption may not always be clear to everyone. An lustrative example is Nigeria where in the early 1980‟s free trade reforms were promoted by the IMF, but were never carried out, because the import licensing system was a major source Ackerman, 1999). Import-restrictions give the bureaucrat more tionary power, and create opportunities for competition for some existing value namely
be gained from buying a good cheaply in a foreign country and selling it more expensively in the domestic market. The rent-seeking theory that links trade to corruption states that trade restrictions are competition-reducing and therefore create incent
. For this to be true, i.e. protecting domestic firms from foreign competition, it must imply that countries both produce and import goods from the same industry, so called, intra industry trade. Studies have shown that this is very much the case for the sample of European
restricting policies in Europe would mean reducing competition in the domestic country with the possibility of creating rent-seeking. Furthermore, as intra
correlation between exports and imports increase, which implies changes in net exports should be small which in turn makes this measure inappropriate for
Deregulation is, as proposed by for example Goldsmith (1999), a way to limit opportunities for corrupt officials to grant exceptions to the rules. Limiting competition through import restrictions means a higher degree of regulation, hence higher customs burden and thus
la (1999)
mic Outlook 71 (2002). The report shows that 18 Europ of intra-industry trade as a share of total trade.
on the link between imports and corruption emphasize that it is the trade restriction itself is, if import demand is fixed, restricting imports on trade restrictions, lower import levels could merely be the reflection of lower import demand.
lem of endogeneity, imports share in GDP could be a more res that reflect trade policy,
-seeking and thus
Ideally, a measure of trade restrictiveness would take into account the extent to which and services (Suranovic, 1997), where the f goods and service with the restriction and the denominator
trade (Figure 2.1).
it is virtually impossible to construct. How trade would have been if the
The measures of trade openness and/or trade restrictiveness considered in this thesis are policy and physical measures of trade restrictiveness. Natural barriers to trade, such as trade distance and access to harbours are not covered in this thesis for the simple reason that it is difficult to
rs and the measures of corruption.
The link between import restrictions and corruption may not always be clear to everyone. An s free trade reforms were promoted by , because the import licensing system was a major source restrictions give the bureaucrat more tionary power, and create opportunities for competition for some existing value namely
be gained from buying a good cheaply in a foreign country and selling it seeking theory that links trade to corruption reducing and therefore create incentives for rent-. For this to be true, irent-.erent-. protecting domestic firms from foreign competition, it must imply that countries both produce and import goods from the same industry, so called,
intra-he sample of European restricting policies in Europe would mean reducing competition seeking. Furthermore, as intra-relation between exports and imports increase, which implies changes in net exports should be small which in turn makes this measure inappropriate for
Deregulation is, as proposed by for example Goldsmith (1999), a way to limit opportunities for corrupt officials to grant exceptions to the rules. Limiting competition through import- restrictions means a higher degree of regulation, hence higher customs burden and thus
opportunities for corrupt officials to extract bribes, as well as possibilities for companies to conceal parts of their operations and make it easier for them to cheat consumers. Furthermore, deregulations do not have to be limited to include only import restrictions. It is intuitively reasonable that all forms of regulations may lead to higher corruption since, as also pointed out by Goldsmith (1999), it provides the officials with more discretionary power.
3 Theoretical Framework
In this chapter the theoretical framework used as a point of departure for the empirical analysis will be presented and analyzed.
3.1 Theoretical treatment of corruption
3.1.1 Principal-agent-client problem
The principal-agent-client approach, which is often used to model bureaucracy and public institutions, is an extension of the standard principal-agent problem which was initially designed to analyze the relationship between private parties such as owners (representing the principal) and managers of a firm (agents) (Lambsdorff, 2007). The principal in this model is the government that entrusts the agent with power to implement the laws passed by the principal according to which the agent is supposed to serve the client. The agent in this case is usually represented by lower-level authorities such as tax or customs officials while the client is the regular citizen and taxpayer, hence both firms and individuals can enter the role of the client. The relationship between the government, authorities and taxpayers is described in figure 3.1 below.
Figure 3.1. Principal-agent-client model
Corruption enters the picture when the agent does not honor his contract towards the principal. Variants of this rule-breaching behavior are when the agent extorts his discretionary power towards the client and/or accepts bribes from the client (Lambsdorff, 2007). The client may be forced to pay an arbitrary amount of money to keep a license, even though there are no justifications for him losing the license otherwise. In other variants of corruption, the agent steals from the principal, also called embezzlement, and actively conceals information from the principal.
The conflict of interest between the principal and the agent arises when there are incentives for the agent to deter from the rules made by the principal and is made possible by the fact that both parties are dependent on each other. The principal is insufficiently skilled and suffers from time constraints and therefore delegates the agent to perform some of its sub-tasks. The agent on the other hand is dependent on the salary received by the principal. Moreover, the agent has informational advantage over the principal inasmuch the agent can hide his true behavior from the principal, i.e. the principal has no way of knowing if the agent acts honestly or corruptly.
The principal may in some cases try to anticipate the agent’s profit from corruption and thereby lower the agent’s salary in which case the principal would be considered to be involved in the deciding on the level of corruption (Lambsdorff, 2007). However, even if the principal tries to create disincentives for corrupt behavior, such as sanctions for those detected of being corrupt and incentives to create honesty, such incentives are costly. Therefore the principal must balance the benefits from more honest agents against the costs of obtaining more honest agents. Nevertheless, if the principal chooses this optimal level of corruption, he may have little influence on the overall level of corruption (Lambsdorff, 2007). The agent’s decision to use his discretionary power for private gains may be independent from the rules of the principal. In highly fractionalized societies, agents belonging to a certain nationality or group (or even family) may favor those clients that belong to the same group. Furthermore, due to the information advantage the agent has over the principal, if there are possibilities for individual agents to maximize their income, they will continue doing so irrespective of the principal’s rules. An example that illustrates this behavior is large multinationals who bribe their way through the administration in developing countries (Eigen, 2002). This leads to the conclusion that the level of corruption in an economy depends heavily on the agent and his incentives, and very little on the principal and his rules. This suggests that, theoretically, endogeneity in a model of trade and corruption should not be problematic.
3.2.1 Rent-seeking
It has been outlined above that corruption arises when the agent has incentives to refrain from the rule-bound behavior. Since the individual is assumed to be rational and profit-maximizing, a number of situations can occur where the agent has incentives to corrupt. Such situations can be a result of the mere existence of rents which can be a result of low wage of officials or of wage differentials (Hillman, 2004) and natural resources (Leite and Weidmann, 1999). In addition corruption can be a result of institutional reasons such as low probability of getting caught (Van Rijckeghem and Weder, 2002) or highly fractionalized society so that the official belonging to a certain social group favors that group over other (La Porta et al., 1999) but also wherever the agent sees the opportunity to extract extra benefits.
One must differentiate between profit-seeking and rent-seeking. The kind of “profit-maximizing behavior” corrupt individuals are involved in is referred to in the economic literature as rent- seeking. Rent-seeking does not involve offering improvements, such as productivity, in exchange for excess income9. If import restrictions are introduced, in the form of tariffs or quotas, this will, ceteris paribus, increase the profits for domestic firms through increased prices as a consequence of demand staying the same and supply of goods being restricted. This price differential will make it even more valuable to buy goods in the foreign market and sell it in the domestic market since the equivalent to a profit, the rent will be higher. The size of the rent will vary depending on if the trade restriction takes the form of a tariff or a quota. Since there is an opportunity to maximize income by importing goods, rent-seeking will take place.
The state has monopoly on allocating property rights, such as subsidies, tariffs or the right to import –quotas (but hires the agent to distribute these property rights). In deciding on the optimal level of rents in an economy the government must balance between the political benefit it would yield as a result of benefitting some interest groups, and the political cost it would mean as a result of not benefitting all (Lambsdorff, 2007). A variety of factors may impact on the size of the rent, though so far in the literature it has been ignored that competition may be one of the factors determining the size of the rent (Lambsdorff, 2007).
9 The distinction has been made by Buchanan (1980), amongst others, who argued that rent-seeking is not
How can one measure the level of rents in a country? Luechinger et al. (2008) introduce life satisfaction differential between public sector employees and private sector employees as a level of rents in an economy and find that it is closely associated with corruption indices. Previous approaches have used wage differentials as a source of rents, job queuing, etc (see Luechinger). Luechinger et al. (2008) find that their measure of the level of rents are highly associated with corruption indices.
3.2 Corruption through limits on imports –the linkage
We have seen how limiting competition through import-restrictions in the form of a tariff may lead to rent-seeking and corruption with the help of the principal-agent-client theory and the rent-seeking theory. A formal model will now be implemented to model the effect of limits on imports through quotas.
In an influential paper, Anne Krueger (1974) introduced a formal model where trade restrictions in the form of non-tariff barriers lead to competition for import licenses which in turn creates rent-seeking behavior and thus corruption. The major proposition of the model is that the welfare losses of a quota are the largest when comparing to those of a tariff-solution or free- trade solution. However, the model also explains how restricting imports leads to rent-seeking behavior.
The model assumes two sectors; food which is domestically produced and exported, and consumption goods which are imported. Labor is assumed to be the only factor of production. A further assumption of the model is that the country is reasonably small so that it cannot affect its international terms of trade. The level of distribution output, D is equal to the level of consumption goods, M, such that one unit of distributive services means exchanging one unit of imports of consumption goods for food in the domestic market and exporting the food in exchange for consumption goods on the international market.
Society’s demand for imports depends upon the domestic price of imports and total income generated from agriculture. Under free-trade, there is free entry into both agriculture and distribution. Consider now a restriction of imports such that the quantity imported is less than under free trade. Entry into distribution is now limited and this causes the labour employed in distribution to decrease, the labour employed in agriculture to increase and the wage in the agricultural sector to fall due to diminishing returns. Distributors will earn a rent in the sense that their wage will be higher than the wage of the agricultural laborers. Since in the case of a quota, there is a limited amount of import licenses, rent-seeking will occur until the wages in the two sectors equalize. It is this feature of the model that fosters corruption.
3.2.1 Policy implications
When an import-restriction in the form of an import quota is set by the government, it must be able to consider the consequences and effects of an import quota. The effects include, as pointed out above, a higher price of the good in the importing country as well as higher income for those able to obtain the right to purchase the good cheaply in the foreign market and sell it expensively in the domestic market. The impact of the choice of policy depends on how the government chooses to administer the quota. The government can choose from several ways to administer quotas (Suranovic, 1997).
The government could choose not to place any restrictions on imports until the quota is filled after which customs officials would prohibit entry of the good for the rest of the period under consideration. This, however, would cause the price of the good to fluctuate during the year, giving rise to rent- seeking as according to Krueger above. Customs officials would gain substantial discretionary power under the period when the good is to be restricted completely They could also auction quota rights. In practice, this means that the government sells quota tickets which allow the entry of one unit of the restricted good which may be presented to
customs officials. The price of the ticket can be determined through auction or by the price differential between good in the foreign market and the good in the domestic market. However, since quota tickets are limited, this outcome would also lead to competitive rent-seeking. Suranovic (1997) points out that if there are no transportation costs the quota holders will receive a quota rent which equals the price differential just mentioned. Selling the quota tickets through an auction would mean the government receive all the quota rents.
Another option is that the government can give away the quota rights by allocating the tickets to assigned individuals. Governments often decide on recipients of quota tickets based on past market share (of the importing companies). It is difficult to see how this option would lead to competitive rent-seeking and thus corruption. Since quota tickets are given away to assigned individuals, there is no possibility for individuals to try to capture the rents of the quotas. Another implication of this option is that since rents are so valuable, governments may use this to direct rents towards political supporters, in which case the level of corruption and trade policies would be considered endogenous.
3.2.2 Critique of the theory
The section above on policy implications also shows how the level and form of trade protection may depend on desired policy outcomes. What are the reasons for governments to implement import-restrictions? Several reasons are likely; to protect workforce or to protect domestic industries for example. In much of the rent- seeking literature the import levels are exogenous. However ,there is also a strand of literature that tries to endogenize import-levels to corruption. Lobbying for special interests by firms, organizations or groups can lead to rent-seeking. Brock and Magee (1978) explain the level of tariffs as a result of lobbying by special interest groups.
It is intuitively plausible that the level of trade restrictiveness may be a result of lobbying by special interest groups (i.e. corruption) and that trade policy in that case would be endogenous. This is the main critique against the theory used, and has gained support in some empirical studies (Lee and Azfar, 2000). Further empirical work remains to be done to establish whether the level of rents and trade policy are exogenously determined.
The theoretical framework used suggests that more restrictive trade-policies create rents and thus higher level of corruption which leads us to the three hypotheses used in testing the theory:
Hypothesis (1):
Corruption is higher for countries that are more restrictive to trade than for those that are more open to trade, when controlling for differences in GDP per capita.Hypothesis (2):
Regulations are associated with higher levels of corruption.Hypothesis (3):
higher degree of regulations is associated with higher levels of corruption when controlling for region- and country- specific variables and cultural factors.4 Data and Methodology
A panel-data sample of the European countries was chosen in order to investigate how trade restrictiveness affect corruption levels in Europe over the years 2002-2008 and to be able to examine whether there are any European-specific variables that affect the corruption in Europe. It is thus noteworthy to mention that the results presented below may be subject to sample bias. However, across the European countries there are differences in population size, culture, GDP per capita, corruption, business freedom, trade freedom etc. Therefore the results obtained from the sample of Europe should reflect the results obtained from the population, i.e. the world, if data was available for all the world countries. Furthermore, to eliminate the possibility that potential differing results between the two measures of corruption is due to
re-coding of the CPI variable, the WGI CORR variable was also re-coded. Rere-coding the WGI CORR variable did not change the results presented below in the empirical analysis.
The time-aspect of the panel data is limited to seven years to maximize country coverage and data availability of other variables. Below follows an outline of the variables considered and used in the regression models.
CPI –Corruption Perceptions Index (Transparency International): The index is a “poll of polls” incorporating different surveys of expert opinions and business representatives. The original scale ranges from 1 to 10 where 10 is the least corrupt. The data was re-coded such that 1 represents least corruption and 10 highest level of corruption for the CPI variable. WGI CORR –World Governance Indicators Control of Corruption (World Bank): The Index measures six different indicators of world governance. This variable measure the control of corruption on a scale ranging from -2.5 to 2.5, where a higher score represents better control of corruption, i.e. lower corruption levels.
GDP/CAP –Gross Domestic Product per capita (EUROSTAT), measured in Euros per inhabitant.
IMP/GDP –Imports share in GDP (%) (EUROSTAT), measuring trade openness.
TRADEFR –A measure Trade Freedom (Heritage Foundation) which represents the restrictiveness of the economy to imports. Incorporated in the index are measures of trade weighted average in most cases as well as a penalty for how extensive the non-tariff barriers (NTB) applied by the country are. A higher value corresponds to less trade restrictiveness. TRADEWE & SIMPLEAV – Trade weighted and simple average tariff rates (UNCTAD). Data was only available for the years 2004-2007 for most countries.
BUSFR –A quantitative index of Business Freedom (Heritage Foundation) which measures the ability to create, operate, and close an enterprise quickly and easily. It measures regulatory rules that are most burdensome to business freedom. A higher value of the index represents less burdensome regulations.
GOVTSIZEFR –Government Size Freedom (Heritage Foundation) measures the freedom from government expenditure. It includes all government expenditures including consumption and transfers and their share in GDP. Higher government expenditures are represented by a lower point on the scale of the index.
FISCFR – Fiscal Freedom (Heritage Foundation) measures tax burden in terms of the top tax rate on income of individuals and corporations separately as well as overall amount of tax revenue as a share of GDP. Higher values of this measure represent less tax burden. RULELAW –Rule of Law (World Bank) measures perceptions about to which extent agents have confidence in and abide by the rules of society. Included in the measure are perceptions of quality of contract enforcement, property rights and likelihood of crime and violence. NATRES –Measures Natural Resource Abundance. The data was constructed by summing exports of fuels, mining products, iron and steel and calculating their share in total exports. Higher values represent higher natural resource abundance.
EU15 – A dummy variable taking on the value of 1 if the country belongs to one of the 15 first member countries of the European Union and 0 otherwise.
CENTRAL - A dummy variable taking on the value of 1 if the country has been under a centrally planned economic system within the last 20 years, and 0 otherwise.
PROTESTANT -A dummy variable taking on the value of 1 if the majority of the population in a country is protestant, and 0 otherwise.
Panel-data set was chosen in favor of cross-section data since a panel-data set allows a more dynamic analysis. It increases the sample size and the power of significance tests. More importantly, it allows one to control for country- and time-specific unobservable variables that may lead to bias. However, the time-dimension in the data set is, as in many panel-data sets, relatively short –eight years. This means that it may not be possible to account for time-specific effects since the data used in the empirical analysis is of such nature that it changes relatively little over time. Corruption and business freedom in particular are variables that depend on institutional factors and do not change much over the small time-period used in the analysis of this thesis. This implies that a longer period of time is needed for changes in business freedom, trade freedom, GDP per capita etc. to have an effect of corruption.
While the two corruption indices correlate highly with each other, the measures for trade policy (trade freedom and simple and weighted average tariff rates) and trade openness (imports share in GDP) have a low correlation. The trade policy measures of corruption are highly correlated among each other while imports share in GDP has a low correlation with all the variables measuring trade policy.
Table 4.1. Correlation between variables measuring trade openness and policy
IMP/GDP TRADEFR TRADEWE SIMPLEAV
IMP/GDP 1.000000 0.009879 -0.078328 -0.207076
TRADEFR 0.009879 1.000000 -0.705641 -0.527852
TRADEWE -0.078328 -0.705641 1.000000 0.822857
SIMPLEAV -0.207076 -0.527852 0.822857 1.000000
The correlation coefficients support the observation of Pritchett (1996) that there is a difference between trade openness resulting from trade intensity and policy influenced trade openness – Imports share in GDP has a low correlation with the measures of trade policy. Therefore the measures for policy trade openness (TRADEFREEDOM, TRADEWE and SIMPLEAV) will be used separately from those of trade intensity (IMP/GDP) to compare if there is any difference in effects on corruption.
The empirical analysis is carried out by first testing the effects of trade restrictiveness and openness on corruption when controlling for GDP per capita. The models are then expanded to including other forms of regulations. Finally the final regression models are presented including variables capturing effects that do not represent trade related factors but other effects that may have an effect on corruption. In all of the models used (except for those including dummy variables), an F-test was conducted that showed that a Fixed-effects model should be used. A problem when using fixed effects models is that the dummy variables for country-specific effects capture too much of the cross-section variation instead of this being reflected in the explanatory variables of interest. However, this risk needs to be balanced against specification errors which may arise if not using fixed effects models.
Based on the theory and previous evidence, it is probable that the models about to be tested suffer from simultaneity bias, or endogeneity. To correct for this possibility one could use instrumental variables, which affect only corruption but which cannot have been influenced by corruption. Finding such instrumental variables is easier for GDP/cap but harder for trade restrictions. Therefore, according to the theory, the thesis makes the assumption that trade restrictions are exogenous.
5 Empirical findings
The models used in testing the hypotheses described in the problem above are introduced in this chapter. Two measures of corruption will be used together with different measures of trade restrictions, government regulations as well as other possible explanatory variables. The analysis is performed by first testing models that include trade restrictiveness and trade openness and then, by including possibly influential variables step by step in order to test the robustness of the results from the first section. Sub-hypotheses will be used to carry the analysis forward. Regression analysis of Panel Least Squares is used.
5.1 Import restrictions and corruption
From the theory, it follows that higher trade restrictions should lead to higher levels of corruption, therefore countries with more restrictive trade policies should have higher corruption.
In this section hypothesis (1.1) and (1.2) will be tested in order to be able to draw conclusions on whether hypothesis (1) can be rejected or not.
Hypothesis (1.1):
Trade restrictiveness, measured by tariff levels, is associated with higher corruption levels.Hypothesis (1.2):
Trade openness, measured by imports share in GDP, is associated with lower corruption levels.Trade freedom is a good measure of trade restrictiveness since it takes into account both tariff barriers and non-tariff barriers. The value of is expected to be negative, since according to the theory, increases in trade freedom should affect the level of corruption negatively.
Model 1
/ (1)
The variable GDP per capita is included due to the large body of empirical work which suggests that more developed countries have lower corruption levels. The usual agreement is that countries with higher GDP per capita also have more developed institutions and thus lower levels of corruption. The variable GDP/CAP is thus expected to capture the effects of other variables usually used in analysis of causes of corruption such as literacy rate, life expectancy, schooling levels, effective institutions and access to technology and information. An alternative measure to the CPI is the Kaufmann index from World Governance Indicators, which is used as a robustness check of the models used in the analysis. If trade restrictiveness is associated with higher corruption levels it should be robust to the use of different corruption measures, therefore the next model is introduced.
Model 2
/ (2)
The CPI-variable has been re-coded such that higher values of the variable represent more corruption, where as for the WGI CORR variable, higher values represent lower corruption. Several empirical studies have used imports share in GDP as a measure of trade openness or restrictiveness. It is thus motivated to include this variable as a measure of trade restrictions and compare how the results differ when using trade freedom as a measure of trade restrictiveness. Model 3
Model 4
/ / (4)
The results from the models above are presented in Table 5.1. below. T-statistics are reported in parenthesis and ***=significant at 1 % level, **=significant at 5 % level ,*=significant at 10 % level.
Table 5.1. Trade restrictions, GDP per capita and Corruption levels
Independent Variable(s) Model1 CPI Model 2 WGI CORR Model 3 CPI Model 4 WGI CORR Intercept 9.877245*** (13.01710) -1.448766*** (-4.145148) 7.513927*** (31.56043) -0.619943*** (-5.735394) TRADEFR -0.040537*** (-3.989235) 0.014597*** (3.076767) IMP/GDP -0.010437*** (-2.842812) 0.003943** (2.335688) GDP/cap -0.000138*** (-26.47416) 6.47E-05*** (26.25203) -0.000154*** (-32.31221) 7.06E-05*** (31.95482) R 0.812993 0.825024 0.808780 0.823040 F-statistic 558.6398*** 544.5924*** 547.7307*** 541.8413 N 260 234 262 236
All slope coefficients have the correct sign as well as being significant at 1 per cent level. There is no large difference when using imports share in GDP in this model specification as a measure of trade restrictiveness, except from a slightly smaller R for both CPI and WGI CORR, than when using the trade freedom index. Furthermore, when using WGI CORR as dependent variable, the significance level of imports share in GDP decreases from 1 per cent to 5 per cent.
Autocorrelation may be a problem in this model. Using the Durbin Watson test is inapplicable in the presence of panel estimation. Graphical examination of the residuals via correlogram suggests that there is indeed autocorrelation present. However, running the same regressions and correcting for autocorrelation, using a Cochrane-Orcutt procedure (AR(1)), does not change the results extensively, other than decreasing the significance level to 10 % albeit still producing slope coefficients of the correct sign.
Before making conclusions about statistical tests, one needs to verify that the series are stationary. Regressing non-stationary series, that is, series which exhibit trends, can produce spurious regression results which can take the form of a high R-squared while the Durbin Watson statistic is fairly low. The regression results may be considered spurious if they show that there is a relationship between two variables when according to theory, no relationship is expected the two variables. Regressing the first differences of the non-stationary series (if they are integrated of order 1, that is, their first differences are stationary) on each other decreases R- squared towards zero and increases the Durbin Watson statistic closer to 2. For the models used in this analysis so far, autocorrelation probably occurred because one or more of the series used are non- stationary.
Both corruption variables are highly correlated with previous values of corruption. Including lagged variables of the two corruption measures as independent variables increases to above 0.98 (see appendix 6). However, including lagged variables of corruption as control variables serves no purpose from an analytical point of view. Knowing that corruption depends on previous levels of corruption does not help explaining what those previous levels of corruption were induced by in the first place.
Raw data on trade restrictions is difficult to obtain. For the countries used in the analysis of this thesis, data on simple average tariff rates and trade weighted average tariff rates are usually only available for two of the years from 2005-2007. One can run an unbalanced panel regression and
