The Effectiveness of Development Aid in Developmental States

(1)

University of Groningen, the Netherlands

International Economics & Business

August 2008

Master Thesis

The Effectiveness of Development Aid in

Developmental States

Author:

Katrien J. Wijsman – s1345613 First Supervisor:

Dhr. D. J. Bezemer – Faculty of Economics & Business Second Supervisor:

(2)

Table of contents

1. Introduction ... 4

2. Literature review ... 5

3. The developmental state... 13

4. Hypothesis ... 18 5. Methodology ... 19 6. Data ... 30 6.1 Dependent variable... 30 6.2 Independent variables... 31 6.3 Control variables ... 31 7. The model... 34 8. Data Analysis ... 36 9. Results ... 41 10. Conclusions ... 45

11. The way forward ... 46

12. Final remarks... 46

References ... 48

Appendix ... 54

Countries in the sample ... 54

Table 1. Variable description and their sources ... 54

Table 2. Descriptive statistics... 55

Table 3. Collinearity matrix ... 57

Table 4. Collinearity statistics developmental state characteristics ... 58

Table 5. Unit Root Test critical values... 58

Table 6. Unit Root Test results... 58

Graph 1. Distribution GDP growth per capita... 59

Table 7. Eviews output... 59

Table 8. Durbin-Watson test results Eviews output table 7 ... 60

Graph 2. Fifteen highest and lowest OS.1-DS Policy indexes... 61

Graph 5. Fifteen highest and lowest EW-DS Policy indexes... 62

(3)

Graph 7. OS.2-DS policy index versus GDP growth per capita, 1990-2001. ... 63

Graph 8. OS.3-DS policy index versus GDP growth per capita, 1990-2001. ... 64

Graph 9. EW-DS policy index versus GDP growth per capita, 1990-2001... 64

Graph 10. EW-DS policy index versus GDP growth per capita, 1990-2001 (separated per geographic area) ... 65

Table 9. Eviews output... 65

Table 11. Eviews output: EW-DS policy index ... 66

Table 13. Eviews output: EW-DS policy index excluding outliers ... 67

Table 15. Eviews output: OS.1-DS policy index ... 68

Table 17. Eviews output: OS.2- DS policy index ... 69

Table 19. Eviews output: OS.3-DS policy index ... 70

(4)

1. Introduction

Although developing countries are supported by billions of dollars of development aid each year, there are still many countries experiencing extreme poverty. Assessing the effectiveness of Official Development Aid (ODA) Boone (1994) and Burnside & Dollar (1997) conclude that during 1973-1993 aid had on average little impact on the economic growth. Over the years many economists have investigated this subject in order to define the ideal policy conditions that will increase the effectiveness of development aid. Which circumstances in the aid receiving country increase the effect of aid on the annual per capita Gross Domestic Product (GDP) growth rate?

Several studies (Easterly & Rebelo, 1993, Fischer, 1991 and Sachs & Warner, 1995) show that the economic growth in developing countries is dependent on the economic policies in that country. Burnside & Dollar (2000) include the effectiveness of development aid in this discussion. Their starting point is aid affects the economic growth, but that its impact is conditional on the policies that affect growth. Burnside & Dollar (2000: 847) define a set of ‘good fiscal, monetary and trade policies that are themselves important for growth’ and prove that these ‘good policies’ in fact increase the effectiveness of aid. Burnside & Dollar construct a policy index to measure the presence of these good policies in a country. This policy index contains the budget surplus, the inflation rate and the openness dummy developed by Sachs & Warner (1995). The positive relationship between the policy index and the effect of aid on economic growth encourages aid to become conditioned on the policy environment. Conditionality in this context refers to the commitment of the receiving country to the donors (often represented by the International Monetary Fund, IMF) to adopt certain policies (www.imf.org). When a country borrows from the IMF, its government makes the commitment to adopt certain economic and financial policies, a requirement that is known as conditionality. This assures that the developing country follows the economic model prescribed by the IMF. In return for compliance with these policy conditions the donor country commits itself to aid or loans.

(5)

Our investigation takes the research on the effect of policies on aid effectiveness in a different direction. Inspired by the economically fast growing East Asian countries such as Taiwan, Singapore, South Korea and Japan we have seen that there is not one form of policy that leads to economic prosperity. The high-growth East Asian capitalist countries and their so-called ‘developmental state policies’ have proven to be able to have a positive influence on the economic growth. These developmental state policies are characterised by an important role of the state in the economic development process of a country. Typical features of the developmental state policies include, among others, a high level of credibility and authority of the government, the presence of subsidies or tariffs in order to protect or stimulate certain industries, the focus on social stability, foreign exchange controls, the low inequality levels and the attention to human capital improvement. The question that arises is whether the effectiveness of aid increases in a developmental state policy environment? In other words, do the developmental state policies that had such a positive effect on the economic growth in many East Asian countries, also have a positive effect on the effectiveness of aid, measured as the increase in the annual GDP per capita growth rate. If we are able to confirm this supposition, we would be able to state the policies on which the development aid should be conditioned.

The structure of the paper is as follows: we will start with a brief overview of the literature on development economics and development aid. After this general description we will take a closer look at the developmental state theory and define and explain the policies characterising this model. The hypothesis will then be presented, after which the research methodology, the variables and the analysis of the data will be discussed. We will end this paper with the conclusion and some final remarks.

2. Literature review

(6)

Foreign aid is a ‘transfer of resources on concessional terms undertaken by official agencies has the promotion of economic development and welfare as its main objectives and has a grant element of 25 percent or more’ (Cassen, 1994: 2). The reasoning behind the development aid changed continuously together with the never-ending stream of research on the effectiveness of the aid conditioned on certain growth models. We will present a brief overview of the growth theories in the context of development aid and their effect on the economic situation in the developing countries.

The beginning of the international development assistance is often considered to be the Point Four Programme the American president Truman launched in 1949. The programme aimed to ‘aid the efforts of economically underdeveloped areas to develop their recourses and improve living conditions’ (Pronk, 2001: 611). The first theories of development economics were developed in the 1950s and were largely based on the success of the Marshall Plan, initiated by the minister of foreign affairs of the United States, George C. Marshall. This plan consisted of a financial injection in Europe after the devastating years of World War II. The theory behind the Marshall Plan consisted of the promotion of the economic growth through the focus on the accelerated accumulation of capital. Although the plan had shown to be successful in Europe, the effects in the developing world were not as expected. Critics blame the failure of this model due to the assumption that capital accumulation is sufficient to induce economic growth. Furthermore the assumption that the conditions of the developing countries are the same as those found in post-war Europe was highly criticised. The following years the aid policies kept on changing based on new theories such as the ‘big push’ (Rosenstein-Rodan), the different stages in economic growth (Rostow) and the classical ‘two sector’ development approach (Lewis) (Pronk, 2001). The development assistance became a process of learning by doing, highly debated and often criticised.

(7)

(‘leave-alone’) political and economical beliefs are based on the assumption that an economic system functions best when there is no interference by the government. The doctrines are based on the belief that the undisturbed economies will lead to the maximum well being of the individual and therefore of the community as a whole. One of these laissez-faire theories is the neoclassical theory.

The neoclassical model is based on the assumption that the economy will automatically find the optimal or at least satisfactory level of employment, national income and growth (Arestis & Sawyer, 2004). The only intervention needed is to keep the inflation at the desired level by means of a correct monetary policy. Neoclassicalism is totally different from the post-war consensus in which intervention in the areas of fiscal policy, income policy, industrial policy, planning and social harmony were seen as a governmental duty. According to neoclassicalism, the costs should be shifted from the state back to the individuals. In addition the (labour) market should be as flexible as possible, and the focus should be on privatisation, deregulation and liberalisation. In other words, free movement of capital, goods and services while creating the widest possible conditions for the markets to flourish. Gamble (2001) states that removing as many restrictions on the competition as possible (for example state subsidies, minimum wages and price controls) and empowering the market agents (by reducing the burdens of taxation) is the way to achieve a pareto optimal economic situation.

(8)

Furthermore Rodriquez & Rodrik (2000) state that trade openness may have crowded out other institutional reforms with potentially greater payoffs.

According to neoclassicalism political interference with the market is not compatible with the model. This part of the theory is based on the relative factor endowment model of trade, which is an important factor behind many policy advices of the IMF and the World Bank. The relative factor endowment model assumes constant returns to scale, the presence of competitive markets with many sellers and perfect factor mobility across sectors. ‘Development theory and policy during the last decade have been thoroughly dominated by the neoclassical paradigm …’ (Öniş, 1991: 109). The World Bank and the IMF became the principal conduit for the transmission of the neoclassical ideas to the developing countries (Beeson & Islam, 2005). Since the 1980s the aid programs of the IMF are mainly based on the neoclassical economic theories and favoured policies aimed at reducing the role of the state, the reduction or elimination of subsidies, the market liberalisation and the privatisation of the public enterprises (Buira, 2003). These views were brought to the developing world through structural aid-conditionality.

(9)

aid conditions removes the possible positive effect of the aid on the economic growth. ‘Successfully completed programs fell from over one half in the late seventies and early eighties to below thirty percent in the nineties’ (Buira, 2003: 25). In addition Dreher (2005) states that since in the empirical analysis compliance is controlled for, the remaining negative impact of the IMF programs can be explained by the ‘bad’ advice given by the IMF. Axel Dreher follows the conclusion of Przeworski & Vreeland (2000) that the IMF programs reduce growth while countries remain behind. Khor (2001) concludes that the conditionality policies did not work. ‘Instead of recovery, growth and getting out of debt, many recipient countries have experienced stagnation and many are still trapped into debt’ (Khor, 2001: 2). The lack of the effectiveness of aid conditioned on certain policy implications induced a sharp rise of criticism on the IMF conditionality in the 1990s. The programs were already criticised for being too short-run oriented, too focused on demand management and not paying adequate attention to the impact on growth, social spending and income distribution. In addition also the loss of focus became a major point of critique (Khor, 2001).

Contrary to the neoclassical growth model and the ideal of trade openness are the theories based on government intervention to encourage the economic growth. Gamble (2001) gives an example of successful government intervention using the economic recovery after the Depression and World War II. ‘The success of capitalism in recovering from the Depression of the 1930s and reorganising itself so successfully after 1945 was due to the much greater role played by the state’ (Gamble, 2001: 128). Vartianen (1999) shows that there is a lot of evidence that strong interventionist strategies can be very successful for the economic development. Using Taiwan, Korea, Finland and Australia as examples she shows that the state played an active role in the transformation of these countries. Johnson (1982) was the first to name this form of state intervention in which private ownership is united with state guidance a ‘developmental state’. The supporters of the developmental state model advocate that the economic development of a country requires a state, which can create and regulate the economic and political relationships that can support sustained industrialisation (Chang, 2002). Typical features of a developmental state are based on the East Asian economies characterised by government intervention.

(10)

economies that grew enormously under a regime of government intervention (Bagchi, 2000). In Britain the state encouraged growth, using policies intended to promote and protect their industries. These policies played a critical role in the economic development in the beginning of the 18th century. Free trade was only adopted around 1860, after Britain had already emerged as one of the most powerful nations in the world economically, militarily and politically. The role of the state has been underestimated because it was ‘invisible’ for many (Bagchi, 2000). But during 1560-1851 the national market was protected with tariffs and the import of various kinds of goods from foreign countries was banned. In Germany the States’ paternalism and protectionism and the sectored composition of the industrial growth made it an example of a well-functioning developmental state. The United States are also an example of successful government intervention; between the American Civil War (1861-1865) and the Second World War (1940-1945) the U.S. had the most heavily protected economy in the world. The national system called the American School (sometimes referred to as the protective system) had as its main element the protection of the industries through selective high tariffs and subsidies. During its American School period the United States became the largest economy in the world, with the highest standards of living, surpassing even Great Britain.

Johnson (1982) describes the economic growth miracle in Japan and explains why post-war Japan is a proof of effective government control. The aggressive policies of government control over sectored resource allocation brought about an increase in real income. In the 1950s and 1960s Japan experienced a period of spectacular recovery and growth. This extreme progress can be explained by the role of the leading state actor in the economy, the Ministry of International Trade and Industry (MITI), who had operated an intensive system of control. The governmental exchange control, the regulation of international trade and the focus on particular industries resulted in an impressive economic growth in Japan.

(11)

the increase in the relative profitability of exports a much more plausible explanation for the economic growth experience is the sharp increase in investment demand. Due to the strategic interventions, including investment subsidies, administrative guidance and the use of public enterprise, the intervention of the government could play a productive role in the economic growth process.

What is the reasoning behind the developmental state model and why is the free market or neoclassical model not capable of achieving the successes of the developmental states? In other words, on which points do both models differ so extensively that gives reason to prefer developmental state models? Öniş (1991) explains this difference using several examples. The first striking difference with the neoclassical model is the high degree of government intervention used to distort relative prices so that the desired levels of investment could materialise in strategic sectors. These investments in the chosen key areas of the economy are higher than would have occurred in a free market model without the intervention of the government.

The reasoning behind providing subsidies to firms differed decisively from the industrial policies that are characteristic to a number of neoclassical West European countries. Contrary to these West European countries, which were involved in the subsidisation of declining firms or industries experiencing financial difficulties, the developmental states refrained from bailing out firms that were badly managed in otherwise profitable markets (Öniş, 1991). The effect of the subsidies given by the developmental states increased by setting stringent performance criteria attached to these subsidies. In many economies based on the neoclassical model the government also provides subsidies but fail to integrate the subsidies with discipline over business performance. In this way the subsidies often proved to be counterproductive and emerged as a major avenue of rent seeking.

(12)

Of course not all interventionist countries turned out to be successful developmental states. By the 1970s, the frequency of a negative outcome of government interventionism caused the economic models to be based on the rejection of intervention. ‘In nearly every developing country the government involves itself in the economy and is powerful’ (Sen, 1983: 752) unfortunately often with poor results. Huff (1995) also doubts the multi-applicability of the developmental state model. ‘The experience of interventionism evolved in Singapore is unlikely to be applicable in any wholesale way to other developing countries’ (Huff, 1995: 1434). The bureaucratic-authoritarian states in Latin America, such as Argentina, Brazil, Chile, Mexico and Uruguay illustrate this problem. In contrast to the East Asian states these countries experienced severe legitimation problems. Mexico is also a striking example of how governmental interventionism can fail. The absence of bureaucratic autonomy has led to a situation in which the goals of the state were directly reduced to private interests. Furthermore Weiss (2000) argues that the capacity of the government has advanced considerably in Thailand and Indonesia. But although the states could be regarded ‘interventionists’, these countries cannot be seen as successful developmental states. The constellation of political priorities, state structure and government-business relations had differed significantly form the developmental state examples.

This general overview raises the question as to which conditions and/or policies have been, and in the future can be, of significant influence on the economic growth in a country. Nowadays many developing countries are often expected to adopt the so-called ‘global standard’ of policies and institutions. Global standard refers to the neoclassical model in which (government) control is regarded as inefficient. The principal examples of conditionalities of the IMF and the World Bank (based on the neoclassical model) are policies aimed at privatisation and liberalisation. In Africa many countries suffered from de-industrialisation as a result of low tariffs that were established as part of the IMF and the World Bank conditionality. Furthermore small farmers were struck when African countries were prevented from raising their low applied agricultural tariffs, allowing the cheap subsidised food imports from the developed countries to flood the local market (www.eurodad.org).

(13)

state as a key bottleneck in Africa’s ability to meet the Millennium Development Goals (Fritz & Rocha Menocal, 2006). This makes us doubt the economic models underlying the IMF aid conditionality. Why does the IMF persist in the neoclassical view, although the developmental state policies have shown to be able to matter in a positive way?

The central research question we investigate in this paper is whether the effect of aid on the economic growth in a developing country is conditional on the presence of the developmental state policies. In order to be able to state the corresponding hypothesis we need to clarify the definition of a developmental state. In other words, which characteristics make a country a developmental state?

3. The developmental state

(14)

Another characteristic of a developmental state, which stands in close relation to the credibility, is the level of authoritarianism. ‘If the government is truly committed to growth as its priority and has a co-operative disposition, a degree of authoritarianism may be advantageous to development’ (Huff, Dewitt & Oughton, 2001: 721). In a developmental state a dictatorship of development exists, which means that in contrast to a democratic regime, one person or an elite group is in control. A political system is present in which the bureaucracy is given sufficient latitude to take initiatives and operate effectively (Johnson, 1995). The state therefore guides the market with instruments formulated by an elite economic bureaucracy. In Korea Park Chung Hee was in control, in Singapore Lee Kuan Yew and in Taiwan it was Chian Ching-kuo. In Japan several ruling elites were in control. The power and autonomy of these elites were centred in the key ministries. Democracy defined as ‘a form of regularised conflict between political forces’ may make it difficult for the state to keep the country in control and achieve the reputation needed to build a developmental state (White, 1998: 42).

(15)

reasoning behind a government’s choice to support certain industries; is a government able to choose the right industries to promote?

We can distinguish two ways of promoting particular industries. One is the use of state controls on tariffs, import-substitution and non-tariff barriers to protect the own industries and improve the export status. In South Korea the government intervened to protect the local industry of cotton textiles from the (Japanese) competition. Later on the subsidies were used to encourage heavy industries. In addition price controls were used extensively. ‘At the end of 1986 as many as 110 commodities were controlled’, including flour, sugar, coffee, paper, automobiles and televisions (Amsden, 1989: 17). The quantitative import restrictions, government export incentives and tariffs provided the Korean businesses with abundant government support. Import substitution consists of the substitution of externally produced goods and services with locally produced ones. In other words, national industrial growth is encouraged in order to reduce the imports of manufactured goods. According to Wade (1990) the evidence suggests that, overall, the policies of import substitution in the 1950’s in Taiwan had a very important role in preparing the way for later success. This makes it plausible to assume that the high import tariffs and export subsidies, rather than a free market economy, could play a significant role in the economic development process of a country. The notion of import substitution was popularised in the 1950s and 1960s as a strategy to promote the economic independence and the development in developing countries (Bruton, 1998). Although import substitution and outward orientation offered a suitable solution to the development problems, a disadvantage of the approach needs to be mentioned; the possible inefficiency of the industries it creates.

(16)

to the targeted firms and industries. The government intervened and arranged international credit for favoured firms at rates far below the domestic rates. This government policy is against the prescriptions of the IMF and the World Bank. The IMF and the World Bank do not prescribe multiple interest and exchange rates but insist on a single interest and exchange rate. Maintaining a difference in the cost of borrowing at home and the cost of borrowing abroad gives governments the opportunity to discriminate in favour of particular industries and firms.

Apart from the level of credibility and authority and the promotion of certain industrial activity, savings is also a key characteristic of a developmental state. ‘The financial systems serve to transform the savings and other capital flows into various uses’ (Levi-Faur, 1998: 71). These savings develop the reserves to supply the financial funds, which make a high domestic saving rate a characteristic of a developmental state. South Korea offers a good example of this feature. In the post-war period South Korea failed to mobilise significant private savings. This changed in 1965 when Park’s administration doubled the ceiling of the interest rates, allowing a significant growth in savings. Consequently, the savings tripled in one year and doubled the following year (Cole & Layman, 1971). Singapore also shows that high levels of domestic saving positively influence the economic development. In Singapore the government’s achievement of control of high savings and correspondingly high investment is regarded as a key to rapid growth (Huff, 1995). There was little that was voluntary about the saving process; public sector saving was the driving force. Public saving consists of the surplus in the consolidated accounts of the public sector, which consists of the government and seven main statutory boards. The government used the monopoly power of the statutory boards to extract an economic surplus from consumers. ‘The implicit taxation of the consumers effectively mobilised savings’ (Huff, 1995: 1427).

(17)

exchange rates. During the miraculous growth phases East Asian countries successfully pegged their currencies to the dollar. Gala (2005) compares the exchange rate regime in East Asia with the situation in Latin America. While Latin American countries went through the well-known stabilisation episodes, East and Southeast Asian countries focused on their export-led growth strategy. This included a permanent stimulus for the export sector, avoiding periods of strong appreciation. Rodrik (1994) recognises the importance of competitive exchange rates in the development of East Asian countries. China is a very good example of the Asian style of exchange rate management. Following the steps of Japan and South Korea, China maintains an undervalued real exchange rate, one of the pillars of the East Asian Miracle (World Bank, 1993). It is important to acknowledge here that this kind of intervention in the exchange rate market creates a favourable distortion to growth (Gala, 2005). Williamson (1999) confirms that East Asian countries economically managed their currencies far better than Latin America or the African countries. They have not allowed their currencies to become so overvalued, by keeping the exchange rates fixed. In Korea and Taiwan we find clear examples of the influence of a fixed exchange rate on economic growth. Both countries started in the 1950s with a multiple exchange rate system. After a unification of the exchange rates, accompanied by devaluation, the rate stayed fixed and the exports increased very quickly afterwards. ‘Thanks to the appropriate macroeconomic and exchange rate policies, the export supply was adequate to meet the increase in import demand’ (Rodrik, 1995: 79). As the real exchange rate has been stable over the whole period from 1956 until 1985 this brought a great help to investors, who did not need to hedge against fluctuations in the exchange rate (Wade, 1990).

Another characteristic of a developmental state policy model is the influence of the model on inequality. Evidence of low income-inequality supports the developmental state model. Contrary to the neoclassical model, which might increase the income gap between the rich and poor, the developmental state increases the social stability and equality. The emphasis on high-quality basic education and augmenting labour demand reduce income inequalities. The East Asian economies have experienced rapid economic growth with relatively low levels of income inequality.

(18)

political and cultural unity of the states, and to cement the ideological hegemony of the dominant classes. Green compares the emphasis of a developmental state given on education with the examples of other countries such as the U.S., were education played a major part in assimilating immigrant cultures. According to Morris (1996) the expansion of the national educational systems was largely sequential, with first priority placed on primary education, later on secondary education followed by tertiary education. Investment in education therefore is our final developmental state characteristic. As noted by Abe (2006) an important conclusion is that, apart from ‘ordinary’ education, developmental states emphasise on values and moral education. This moral education in East Asian developmental states indirectly contributed the creation of national solidarity, which was necessary for the rapid economic development directed by the state.

Finally we should mention the degree of public-private cooperation that is evident in developmental states. Despite the autonomous character, the developmental states are deeply embedded in the society (Yoshimatsu, 2003). The close business-government relations enabled information sharing and guided the capital channelling into the favoured industries. The huge business conglomerates, for example the Keiretsu of Japan and the Chaebols in Korea, owed their phenomenal growth to the special incentives provided by the state (Öniş, 1991).

4. Hypothesis

The economic growth evident in many developmental states makes us wonder whether the developmental state policies could also positively influence the effectiveness of aid. The central research question we try to answer in this investigation is therefore whether the effect of aid on the economic growth is conditional on the presence of developmental state policies in the receiving country.

Hypothesis: The presence of developmental state policies in a country causes aid to have a more positive effect on the annual real growth rate of the GDP per capita.

(19)

5. Methodology

To be able to measure the effectiveness of development aid we need to choose the dependent variable. In order to assess the effect of aid we could investigate its effect on the economic growth or on the reduction of poverty. The effect on the economic growth could be measured using the annual GDP per capita growth rate, as is common in economic research. When using the GDP per capita growth rate as a proxy for economic development we need to be aware of a possible problem. If only the wealthy minority benefits form a substantial increase in the GDP per capita growth, we would not want this to be considered as successful economic development. Dollar & Kraay (2004) investigated this measurement problem and concluded that the growth rate of the income of the poor is closely related to the overall annual growth rate of GDP per capita. We can assume that an increase in the GDP per capita growth rate leads (on average) to a proportionate increase in the incomes of the poor. As poverty reduction is difficult to quantify and not commonly used in economic studies, we decided to use the annual growth rate of GDP per capita as the measurement of economic growth.

Measuring the influence of aid on the economic growth has been discussed in several studies; Dreher (2005) summarises all different methods in three models. First of all comparing the ‘before’ and ‘after’ situation is a way to define the effect of aid on the economic growth. The difficulty that arises in this model is the definition of the before situation. As participation in the IMF aid programs is often not exogenous but (usually) a consequence of a crisis, the before situation could be a dishonest representation of the economic situation in the country. The second method is a comparison of the growth rates in aid-program countries with the growth rates in a control group. The composition of an accurate control group is an obstacle in this method. At last Dreher mentions the regression analysis, the method used by most recent studies. ‘When endogeneity of the IMF-related variables is carefully taken into account, this method seems to be the most promising one’ (Dreher, 2005: 772). We agree with the conclusion of Dreher (2005) and will use the regression analysis in order to test our hypothesis.

(20)

growth causes a developmental state. Although in theory this might occur, it seems highly unexpected that the economic growth causes a country to become a developmental state. Endogeneity might become a problem with regard to aid and the economic growth, which could bias our regression outcome. An upward bias if donors give aid to countries that already are doing well, a downward bias if donors give aid to countries performing weakly in terms of growth. In order to overcome the endogeneity problem we will use the Generalised Method of Moments (GMM) analysis.

Our incentive is to base the overall set-up of our investigation on the paper of Burnside & Dollar (2000). The main advantage of following their structure is that it would enable us to compare the results. In other words, will the presence of the developmental state policies in an aid-receiving country have more effect on the effectiveness of aid than the good policies determined by Burnside & Dollar? Burnside & Dollar (2000) compose a policy index in order to answer this question. As we are interested in the effectiveness of aid in a developmental state we will replace their policy term by a variable consisting of the developmental state policies. We will refer to this index as the developmental state policy index (DS policy index). This index represents the presence of the most important developmental state policies in a country. We already explained the characteristic policies of a developmental state and will now provide an overview of the variables of which the DS policy index is compiled:

1. Credible status of the state

In order to define a developmental state we need to find a variable that enables us to measure the credibility of the state. An option is to use the level of corruption as a proxy for credibility. Schleifer & Vishny (1993) and Hellman, Jones, Kaufman & Schankerman (2000) conclude that countries with a high level of corruption have weak governments. But corruption does not necessarily have a negative influence on the credibility of the government. ‘Corruption within a society is independent of the reigning ideology’ (Kaufmann, 1998: 140). The level of public trust in a society is a major factor that distinguishes between countries suffering from corruption and countries able to perform well economically, even in a corrupt environment.

(21)

decisions. This has a destabilising effect on the economy and negatively affects the credibility of the government. The political stability in a country will thus be used to value the credibility of the government. According to Brunetti (1997) the political stability is proxied most accurately by the governmental change and the national election indicators. We will use the Database of Political Institutions (DPI) composed by Beck & al. (2001). Beck & al. have developed two variables on government stability based on the indicators of the turnover and tenure of governments. These variables capture the turnover of governments’ key decision-makers in a year. The variables are calculated by dividing the number of exits of veto players during year t by the total number of veto players in year t. ‘Veto players are defined as the president and the largest party in the legislature for a presidential system and as the prime minister and the parties in the government coalition for a parliamentary system’ (Beck & al., 2001: 168). When defining the stability a distinction is made between a change in the senate or not. Assuming the existence of a senate, the variable ‘STABS’ assumes the senate changes and represents the percentage of veto-players that dropped from government between year t and t + 1. The variable ‘STABNS’ represents this percentage assuming the senate did not change.

2. Degree of authority

(22)

The first indicator of the Polity IV dataset is executive recruitment. According to Eckstein & Gurr (1975) executive recruitment involves the way in which executives come to occupy their positions. The Polity IV dataset contains three indicators of the structural characteristics by which chief executives are recruited: the extent of institutionalisation of executive transfers, the competitiveness of executive selection and the openness of executive recruitment. All three component-variables are combined in one concept variable representing executive recruitment. Following Eckstein & Gurr (1975) another indispensable ingredient of the authority level in a country is the existence of the decision rules that provide the basic criteria under which the decisions are considered to been taken. This is measured with the executive constraints concept. Furthermore the political competition indicator combines the information of two component variables. The first component variable is the regulation of participation, which gives a rate to the development of institutional structures for political expression. The second component variable is the competitiveness of participation, which measures the extent to which the non-elites are able to access are taken into account.

In order to measure the authority level of a country we decided to use the executive constraints (EXCONST) indicator, as it resembles the authority level of developmental states most accurately. A seven-category scale is used, starting from unlimited authority (resembling a situation where there are no regular limitations on the executives actions) to executive parity or subordination (the accountability groups have effective authority equal to or greater than the executive in most areas of activity) (Polity IV Users Manual).

3. Promotion of industrial activity

After the credibility and authority measures, the promotion of industrial activity is a focus point of a developmental state model. We decided to separate this feature into two aspects:

A. The use of import substitutions and tariffs to protect own industries and improve the export status.

(23)

in mind that we are not sure of the impact of the aspect ‘other social benefits’ on the protection indicator. Ideally, we would exclude the ‘other benefits’ and measure only the subsidisation of private businesses by the government. These subsidies represent the transfer payments paid directly to private firms and public corporations, the subsidies of our interest with regard to the developmental states. As accurate data on this subject is unfortunately not available, we decided to use the measurement of the World Bank, including the other benefits.

B. The allocation of credit to strategic industries and the direct support for state-owned enterprises.

The allocation of credit to strategic industries and the support for state owned enterprises will be valued using a measurement of financial repression. ‘Financial repression refers to a set of policies, laws, formal regulations and informal controls imposed by the government on the financial sector to influence the financial intermediaries of an economy’ (Denizer, Desai & Gueorguiev, 1988: 3). The level of financial repression will be used as a measure of the influence of the government in a developmental state on the allocation of credit. To measure the level of financial repression we will use quantitative credit controls. ‘Credit ceilings are one of the most commonly used forms of selective credit controls’ (Yossifov, 2002: 3). By manipulating the modalities of their design, credit ceilings can be used by central banks or governments for a variety of purposes, for example the channelling of credit to specific sectors of the economy (Hilbers, 1993). Credit ceilings have been extensively used in the past in industrial countries. Following McKinnon (1982) financial repression will be proxied by the reserves ratio. The definition of the reserve ratio is the ratio of the commercial banks reserves to the money supply (M1 and quasi money). We will use the IMF Financial Statistics database in order to compose the financial repression variable. Reserves are measured as the domestic currency holdings and deposits with the monetary authorities. Money supply combines money (the sum of currency outside deposit money banks and demand deposits other than those of the central government) and quasi money (comprising time, savings and foreign currency deposits of resident sectors other than central government).

4. High domestic saving rate

(24)

GDP. A higher saving rate will positively influence the DS policy index without the necessity to define the rates as high or low. The gross domestic saving rate will be gathered from the World Development Indicators (WDI) database.

5. Exchange rate volatility

The developmental state is characterised by stable exchange rates. We can make a distinction between two types of exchange rate classification, namely de jure or de facto. The first method measures the intention of the monetary authorities; it incorporates information on the policy intention of the government and therefore retains an element of forward looking. Whereas for the de facto method the measurement is based on the actual behaviour of the exchange rate, regardless the political intentions. A comparison between the coarse natural classification, which is the de facto based classification proposed by Reinhart & Rogoff (2004), with the de jure classification of the IMF, showed huge discrepancies between both measurement methods. The investigation presents evidence for significant overstating the de jure classification, showing substantial differences between the actual behaviour of the exchange rate and the commitment made by policy makers. For these reasons we have chosen to use the de facto ‘Natural Classification’ database composed by Reinhart & Rogoff (2004). Reinhart & Rogoff innovate in the field of exchange rate rating in the sense that they incorporate both data on parallel and on dual exchange rate markets. Reinhart & Rogoff state that any exchange rate classification ignoring the existence of ‘black’ or parallel markets is flawed. Their initiative to incorporate data on parallel and dual exchange rates is based on the finding that the dual or parallel rate is a far better barometer of monetary policy than the official exchange rate and often most economically meaningful. In order to capture the wide range of arrangements Reinhart & Rogoff allow fourteen categories of exchange rate regimes, ranging from no separate legal tender or strict peg to a dysfunctional ‘freely falling’ or ‘hyperfloat’.

6. Inequality

(25)

plots the cumulative percentages of total income received against the cumulative number of recipients, starting with the poorest individual or household. The Gini index measures the area between the Lorenz curve and a hypothetical line of absolute equality, expressed as a percentage of the maximum area under the line. Thus a Gini index of 0 represents perfect equality, while an index of 100 implies perfect inequality(www.esds.mcc.ac.uk). The Gini index will be taken from the World Development Indicators.

7. Investment in education

The last characteristic we include in the compilation of the DS policy index is the investment in education. Although according to Holliday (2000) the East Asian developmental states were social laggards, Wong (2004) states that the East Asian developmental states invested most of their social capital in the provision of accessible education. This was justified as a means for human capital development and thus defined as an economic investment rather than a social policy. In order to give a rating to the investment in education we could investigate the enrolment rates of primary and secondary schools. A shortcoming of the school enrolment rates is that they only show the resultant effect of the investment in education. They cannot be used for determining the quality of the education and provide insufficient information on the interest and focus of the government on education. Therefore we decided to measure government’s incentive to improve education, proxied by the investment in education. The investment in education is measured as the educational expenditure of the government as percentage of total government expenses. The data is gathered from the World Bank Educational Statistics.

The described developmental state variables will be represented in one index, the DS policy index. When compiling this policy index we came across several challenges, which we will discuss briefly. An obstacle was the definition of the weights of each of the variables in the DS policy index. Is every characteristic equal in importance for the definition of a developmental state? Or do we need to assign a higher weight to certain developmental state characteristics in the compilation of the index? For example, we might expect the degree of authority and the level of credibility of the state to be of more importance for a developmental state than the level of economic equality. We would want this difference in importance to be reflected in the DS policy index by attributing weights to the variables of which the index is compiled.

(26)

growth will also increase the effectiveness of aid. Regressing their policy variables with economic growth provides the correlation coefficients of the effect of each policy on the GDP per capita growth rate. These coefficients will be used as the weights in the compilation of their policy index. The reasoning behind this compilation method is that Burnside & Dollar want the index to represent whether a country adopts the policies that are ‘good’ for the economic growth. The expectation is that a government that adopts these policies will be able to increase the effectiveness of development aid. In other words, development aid would contribute more to economic growth, conditioned on the presence of these policies. This method could also be used to compile the DS policy index, but the method certainly has some disadvantages. Although we do expect developmental state characteristics or policies to have a positive effect on the economic growth, this is not our main assumption. We hope to proof that the common effect of developmental state characteristics is positively correlated with the economic growth. One developmental state characteristic as such does not need to have a positive effect on the economic growth; the fact that these policies are simultaneously present in a country is what should induce economic growth. For example the credibility level of the government; although this is a typical developmental state characteristic, we do not expect this feature to be positively correlated with economic growth. Credibility is needed to successfully implement developmental state policies such as the promotion of certain industries, but credibility on itself is not growth enhancing. Furthermore we are, contrary to Burnside & Dollar, not interested in judging the policy environment to be ‘good’ or ‘bad’. In addition we should mention our doubts about the accuracy of the method. In our view this is not a satisfactory approach for arriving at the weights of the variables, as it will strongly influence the outcome. Giving a higher value to the characteristics of a developmental state according to their correlation with the economic growth strongly increases the chance of finding a positive correlation between the DS policy index and the economic growth.

On the other hand, we do have the intention to produce a research comparable to Burnside & Dollars’. In addition we expect a developmental state to be focused on economic growth and could assume a developmental state government pursues the policies enhancing this economic growth. Weighting a developmental state characteristic according to its correlation with economic growth therefore seems acceptable and we decided to follow Burnside & Dollar.

(27)

evidently adopted the developmental state policies and thus without doubt can be specified as a developmental state. An example of such a group is Japan, Singapore, Taiwan and South Korea. According to the number of times we find a particular developmental state characteristic present in this carefully selected group of developmental states we could assign a weight to this characteristic. In this way, each of the characteristics receives a certain degree of importance in the compilation of the DS policy index. The disadvantage of this method is the lack of sufficient developmental states to base our conclusions on. We therefore decided not to use this method.

Another option to assign weights to the developmental state characteristics is to use ordinal scores. In that way we would be able to derive a weight for each developmental state variable according to the expectation of the value, high or low. For each variable it is clear whether we expect a high or a low value to be characteristic for a developmental state. We will calculate the quartiles of each variable serie and give an ordinal rating to each quartile, from one to four. All ratings will be summed up which will give us the final DS policy index for each country for each time period (four year average). The higher the index for a country, the more developmental state policies are evident in this country. An example to make the calculation method clear: according to our theory we expect to see higher education expenses in a developmental state. All education expenses are divided into quartiles. For each country, for each time period we will rate the educational expenses. If their expense on education falls in the 4th quartile, we will appoint a score of four, if their educational expense falls in the 3rd quartile, an index score of three will be allocated, and so on. For the developmental state characteristics that we expect to be low in developmental state, the rating system works the other way around. A low index score will then be appointed to the highest quartile.

After having compiled the DS policy index we will run a regression taking the GDP per capita growth as the dependent variable against the set of control variables, aid per capita, the DS policy index and an interaction term of aid per capita and the DS policy index. The outcome of this regression will show whether a relation between the DS policy index and the effectiveness of development aid exists.

(28)

out to be the case for one or more of the developmental state variables, we could solve this problem by excluding one of these variables. In order to determine the presence of correlation we will compose the correlation matrix, providing the pair wise correlations coefficients. The interpretation of a correlation coefficient depends of course on the context and its purposes. Taking in mind that we are dealing with a model in which we have to account for the contribution of complicating factors, we will assume a correlation of R2 > 80% to proof the presence of correlation. Based on the correlation matrix we will decide if it is necessary to exclude variables from the regression. Afterwards we will also check for multivariate correlation, using the Variance Inflation Factor (VIF) method. The VIF represents to what extent the regression estimates are influenced by the existence of collinearity among the predictors. In theory a VIF estimate ranges from 1 to infinite, with larger numbers indicating collinearity. In practise a VIF value greater than 10 indicates that the predictor variable has a strong linear association (R2>0.95) with other variables included in the regression model (Kleinbaum, Kupper, Muller & Nizam, 1998). We will consider a VIF of 5 (corresponding an R2 of 80%) to be the critical value. If the VIF exceeds this critical value, the added explanatory value of including this variable in the index is so little that we are able to leave the variable out. The logical question that follows on this is of course how to define which variable to leave out. The overall rule is to base the exclusion on the theory. The variable with the least theoretical support should be dropped.

With regard to the ordinal score method a correlation would not cause problems as the variables will not enter the regression simultaneously. Therefore, although correlation between certain developmental state variables might be evident, this will not influence our outcome.

Summarising the DS policy index will be composed as follows:

Following Burnside & Dollar:

BD-DSindexit = [α1 * credibility] + [α2 * authority] + [α3 * subsidy] + [α4 * reserves ratio + [α5 *

savings rate] + [α6 * volatility of the exchange rate] + [α7 * Gini Coefficient] + [α8 * educational investment]

α0, 1, … 8 = representing the weight of the variable on economic growth.

Ordinal Scale compilation:

OS.1-DSindexit = Σ (ordinal score of each developmental state characteristic)

(29)

Variable Description Proxy for Measurement STABS – percentage of veto players that

dropped from government between year t and year t + 1, senate changes also ‘Credibility’

STABNS – percentage of veto players that dropped from government between year t and year t + 1, senate does not change

The credibility of the state and therefore the ability of the state to implement a strong developmental policy.

Percentage

‘Authority’ EXCONST - Executive constraints Authority measure of the government

Index

‘Subsidies’ SUBS - Subsidies and other transfers as percentage of GDP Promotion of particular industries Percentage ‘Reserves ratio’

RESERVES_M2 - Ratio of commercial banks reserves to the money supply (M1 and quasi money)

Degree of financial repression

Ratio

‘Savings’ GDSAVINGS - Gross domestic savings rate as percentage of GDP Governments’ ability to mobilise savings Percentage ‘Exchange rate’

RRRMCODE and RRMGCODE - De facto classification, measurement based on the actual behaviour of the exchange rate

Control of credit allocation

Index

‘Inequality index’

GINI - GINI index Inequality Index

‘Educational investment’

EDUCSTATS - Education expenditure of government as percentage of total government expenditures

Investment in education

Percentage

Alternative compositions of the DS policy index

Due to the fact that we used two variables to measure the level of credibility, namely STABS and STABNS, we unintentionally gave extra weight to these features in the compilation of the DS policy index. The same holds for the exchange rate where we used both RRMCODE and RRMGCODE. Therefore it would be interesting to see if (using the ordinal score method) taking the average weight of these variables gives us a different outcome. We will refer to the outcome of this compilation method as OS.2-DSindexit

(30)

policy index another time using the ordinal scale method, without taking the presence of the credibility and authority of the government into account. Of course one might claim that absence of credibility and authority disables a developmental state to be successful. Although in theory we might agree, the measurement uncertainties of both variables made us decide to investigate the effect of excluding them. We will refer to the outcome of this compilation method as

OS.3-DSindexit.

Finally it would be interesting to see the outcome of a totally different manner to compile the DS policy index. Although the ordinal scale method is chosen with reason, we should still be open for another composition possibility and the effect on the outcome. Therefore we decided to compile the DS policy index using an equal weighting scheme:

EW-DSindexit = Σ (1/8*DS characteristic) = (1/8*credibility) + (1/8*authority) + (1/8*subsidies) +

(1/8*financial repression) + (1/8*savings) + (1/8*educational investment) + (1/8*exchange rate) + (1/8* inequality)

6. Data

Our combined data set contains 23 variables for 112 countries over a time span of 40 years from 1962 until 2001. Burnside & Dollar (2000) used a dataset consisting of 275 observations in 56 countries, Easterly & al. (2003) increased this set to 62 countries and 356 observations. As we intend to include eight variables in our index as opposed to the three in the policy index of Burnside & Dollar and Easterly & al. we will need more data in order to keep our outputs meaningful. As regards the availability of the distinct variables for each country in the set we will decide whether we need to exclude certain countries due to too many data gaps. As observations might be missing for some years or countries, the dataset would become ‘unbalanced’. In order to overcome (part of) this problem we decided to group the variables into 4-year averages. We followed the grouping of Easterly & al. (2003) and divided the data into 10 sets of 4 years, starting in 1962. The data are mainly gathered from the World Development Indicators, the IMF and Easterly & Levine (1997). An overview of the data and its sources can be found in the appendix, table 1.

6.1 Dependent variable

The dependent variable measuring economic growth is the annual growth rate of GDP per capita. The data is gathered from the World Development Indicators database.

(31)

6.2 Independent variables

Development aid is measured by Official Development Assistance (ODA), measured in current US$. ODA consists of net disbursements of loans and grants made on concessional terms by official agencies of the members of the Development Assistance Committee (DAC) and certain Arab countries to promote economic development and welfare in recipient economies listed as developing by the DAC. Loans with a grant element of more than 25 percent are included in ODA (www.worldbank.org). The data is gathered from the WDI database. We already explained the variables composing the DS policy index.

6.3 Control variables

Levine & Renelt (1992) investigated the influence of control variables on the outcome of regressions and concluded that the results for cross-country regressions are highly sensitive to small changes in the conditioning/control set. We selected a set of control variables considering their importance as growth determinants defined by the literature. Included in the control set are variables that differ over time and across countries, as well as variables that vary across countries but are assumed constant over time. The influence of the control variables (summarised in a vector Z in the regression) on the economic growth is estimated by βi in the equation.

Levine & Renelt (1992) concluded that initial income, investment and the initial secondary school enrolment rates are found to be robustly correlated with the GDP per capita growth rate. A number of growth studies (Easterly & Levine, 1997, Burnside & Dollar, 2000, Dollar & Kraay, 2004 and Chang, Kaltani & Loayza, 2005) confirm their findings and add the following control variables: a measure of the financial depth (proxied by the average ratio of private credit to the GDP) and the level of public infrastructure (proxied by the average ratio of main telephone lines per capita) (www.worldbank.org). As already discussed in the literature review the impact of the level of trade openness is highly questioned in the growth literature. Broadly accepted in the literature is the favourable effect of exports on the rate of economic growth, especially in the less developed countries (Michaely, 1977, Tyler, 1981, Feder, 1982, Balassa, 1985, Ram, 1985, Chow, 1987 and Krueger, 1998).

(32)

persons from a given country do not belong to the same ethnic group (Mauro, 1995). Furthermore a measurement of the development of the financial system is included in the control set. Although a measurement of financial depth, calculated as private credit over the GDP, is also commonly used as control variable, we decided to remain with Burnside & Dollar and use the calculation of broad money over the GDP. We do not consider it necessary to include both of these measurements in the control vector because of the expected high correlation between both variables. Again following Burnside & Dollar the institutional quality of a country will be controlled for using the International Country Risk Guide (ICRG) of Knack & Keefer (1995). Following Knack & Keefer (1995) the measurement of the institutional quality captures security of the property rights and the efficiency of the governments’ bureaucracy. As this variable is not widely available before 1980 we decided to agree with Burnside & Dollar (2000) and use each country’s 1980 figure throughout the time, on the assumption that the institutional factors change very slowly over time.

Furthermore based on some additional growth studies we decided to control for the level of public infrastructure, the level of human capital proxied by the average rate of primary and secondary school enrolment, exports, and investments. A measurement of investment is obtained by dividing gross fixed capital formation by gross domestic product, both at purchasers’ prices. The rate of investment proxies the stimulus to economic development, reflecting the infusion of requisite capital to finance the development processes. Exports will be measured by the ratio of exports to the GDP. The data is gathered from the World Bank. The initial status of the economy will be taken into account by including the initial GDP per capita. Finally the geographic differences will be controlled for through the use of dummies for Sub Saharan Africa, East Asia and Latin

America and the Caribbean. The reason for including dummy variables for the different

geographic areas comes from the investigation of Barro (1991). Barro found significant, negative coefficients on both Africa and Latin American and the Caribbean cross-country regressions. The inability to explain the poor performance of Africa and Latin America and the Caribbean with variables designed to control for political, economic and other measurable characteristics, made the inclusion of geographical dummy variables very reasonable. Levine & Renelt (1992) also include dummy variables, following Barro (1991) and other previous researchers (Romer, 1989, 1990a and Grier & Tullock, 1989) who have found significant effects for the continent variables.

Concluding, the set of control variables is as follows:

(33)

γ 6(M2/GDP)i(t-1) + γ 7EDUCSTATSit + γ8TELit + γ9 EXPORTS it + γ10 INVESTMit +

γ 11 SSAi + γ12EASIAi + γ13LATINCAi

Z it = Vector composed of all control variables. GDP initial = The initial real per capita GDP.

ETHNIC it = Ethno linguistic fractionalisation representing societal conflict.

ASSASSit = Number of assassinations and attempted assassinations representing civil unrest. (ETHNIC*ASSASS) = Interaction term between ethnic fractionalisation and assassinations.

ICRGit = Institutional quality will be controlled for using the ICRG from Knack and Keefer (1995).

(M2/GDP)i(t-1) = Financial system proxied by the level of broad money. This variable is lagged by one time period due to the concern over the endogeneity problem.

EDUCSTATSit = Level of human capital proxied by the average primary and secondary school

enrolment rate.

TELit = Level of public infrastructure proxied by the number of main telephone lines in use per 100 inhabitants.

EXPORTSit =Ratio of exports to GDP.

INVESTMit = Investment is measured using the net investment share in GDP; net share of

investment is measured in relation to total production. SSAi = Dummy variable for sub-Saharan Africa.

EASIAi = Dummy variable for East Asia.

LATINCAi = Dummy variable for Latin America and the Caribbean.

As data about the number of assassinations (Banks, 2002) and the institutional quality of a country (Knack & Keefer, 1995) appeared not to be available free of charge we decided to use the data set of Easterly & al. of 2003, who used Banks (2002) and Knack & Keefers (1995) datasets. As the data is only available in four-year time periods, we will follow this. Furthermore after assessing the descriptive statistics of the control variables it became clear that the number of observations for primary and secondary school enrolment were very low. In order to keep the number of observations as high as possible we therefore decided to proxy the level of human capital by the educational expenditure of the government, measured as percentage of total government expenses. A summary of all variables and their sources is given in table 1 in the appendix.

(34)

expenditure is present in the DS policy index and simultaneously part of the control variable vector. Burnside & Dollar (2000) do not encounter this problem, as their policy index variables are not simultaneously control variables. Although this might look disputable (how can we compute an index based on a variable that is also a control factor?) this will not cause any problems. As long as this is not the only variable in the model (which evidently it is not) no errors should occur. We must be cautious though as it may cause serious multicollinearity. We need to overcome the possibility that the DS policy index becomes a direct function of the other independent variables.

7. The model

The following equation will be used to test whether the DS policy index positively influences the effect of aid as a fraction of the GDP, received by country i in period t, on the annual real GDP per capita growth rate.

GDPpc,it = β0 + β1AID it + β2DSindexit + β3[AID it*DSindexit] + βi Z it + u git + ε it (1)

GDPpc,it = Annual real GDP per capita growth rate. i = Country index

t = Time index

AIDit = Aid receipts relative to the GDP.

DSindex it = Developmental state policy index

AID it*DSindexit = Interaction term of aid and the DS policy index.

Z it = Vector of exogenous variables that might affect growth and the allocation of aid, capturing the so-called control variables, with βi as the vector of the coefficients of the control variables. u git = Unobserved fixed time and country specific effects.

ε it = Error terms