Title: Economic Openness, Growth and Income Distribution: A Pro-Poor Growth Analysis of 61 Developing Countries.

0

Title: Economic Openness, Growth and Income Distribution: A Pro-Poor Growth Analysis of 61 Developing Countries.

University of Groningen Faculty of Economics and Business

International Economics and Trade

Supervisor: dr. R.K.J Maseland, Co-Supervisor: prof. dr.mr. C.J. Jepma

Date:14/06/2016

Peter Aardema, s2390515

p.aardema@student.rug.nl

Abstract

This paper empirically estimates the effect of economic openness on pro-poor growth. It ventures to estimate the relative and absolute benefit the lower income groups have been able to reap from globalization and the consequent intensifying international economic integration.

The empirical analysis is based on data from 61 developing countries in the period 1980-2011.

This paper uses a 2SLS as well as a generalized method of moments estimator GMM. Trade openness as well as inward and outward FDI are considered. The findings suggest that the poor benefit from trade openness in absolute terms, but that in relative terms trade does seem to advantage or disadvantage the poor. The poor do benefit from inward FDI in absolute terms, whilst the absolute effect of outward FDI on the poor remains ambiguous. However, both Inward FDI and outward FDI tend to adversely affect the poorest quantile of the population in relative terms.

Keywords - Economic Openness · Foreign direct investment · Poverty · Pro-Poor Growth

1 1. Introduction

At the millennium summit in 2000 world leaders declared a desire to achieve various collective objectives for worldwide development. Delegates from all UN member countries pledged to improve the economic and living conditions of the world population by 2015, with respect to these Millennium Development Goals (MDGs). Perhaps, the most pressing MDG was MDG1. MDG1 essentially consisted of 3 parts. 1) Halve the amount of absolute poverty, people with income under the (1,25$ a day) poverty line, 2) Increase the productive employment, especially amongst the poor and underprivileged, 3) Halve the amount of people suffering from hunger. Even though the pace of poverty reduction has been slowing down in recent years (Imai et al., 2010), many of the MDGs concerning poverty reduction and improving living standards have been reached (UN, 2015)

. In spite of several economic downturns, income conditions for the collective poor have massively improved over the past decade. In 1999, 29,1% of the world population lived under the new World Bank’s absolute poverty line of 1,90$ a day

, by 2012 approximately 12,7% lived under the poverty line, a reduction in absolute poverty by of over 55%.

This significant collective reduction in the number of absolute poor, has shifted the focus of leading international organizations, to stimulating countries to achieve a broader, more inclusive process of economic development. By not merely considering how absolute poverty can be reduced, but also how development affects the different income groups within a country.

One example of this is Sustainable Development Goal 10 of the United Nations, which amongst other objectives aims to reduce inequality within countries, and promote inclusive growth (UN, 2014). Likewise, other organizations such as the OECD and World Bank, have aimed to stimulate shared prosperity over the past decade (OECD, 2009; World Bank, 2013). The motivations for achieving a more inclusive growth process are diverse and include a plethora of socio-economic reasons. These include the effects that inequality within countries can have on, amongst others: the distribution and quality of health care (Veenstra, 2005), the effect on crime, internal conflict and social cohesion (Ostry et al. 2014), and to induce decreasing opportunities for social mobility (Mitnik et al., 2013). Besides the socio economic effects that inequality may have, the relationship between inequality and economic growth has garnered much attention (Castellò-Clement 2010; Halter et al. 2014). As well as the growth-inequality-

1 On a world-wide scale poverty reduction has been largely successful. This is mainly due to the successful economic performance of South-East Asia, China and India However, there is little uniformity across countries and regions. Poverty in Sub-Saharan Africa remains problematic, and as such MDGs were globally achieved, but failed to be achieved locally (UN, 2013).

2 The new poverty line of 1,90$ in 2011 prices is the same as the real value of the 1,25$ a day poverty line in 2005 prices, World Bank (2016).

2

poverty nexus (Bouruignon, 2002; 2004; Son and Kakwani, 2008). As such, several authors in conjunction with international organizations such as the World Bank have sought to introduce concepts that give further consideration to how growth and inequality affect poverty, and how the poor can reap benefits from economic growth. Concepts such as: pro-poor and inclusive growth (UN, 2001; UN, 2014).

Besides significant reductions in absolute poverty, there is a second major trend that has been observable over the past decades. Globalization and economic openness have increased considerably over the past decades (World Bank, 2013). There are basically two opposing points of view, concerning the benefits of this liberalization period has had on the developing countries’ poor. In extremis we can describe them as such. At one end are those who profess that globalization policies have assisted in increasing efficiency, and helped in generating technology and knowledge spillovers. These economic improvements caused by liberalization have accordingly benefited economic growth, and as such reduced poverty (Dollar and Kraay 2002; 2004, Dollar et al. 2016; Ravallion & Chen 1997). These authors indicate that globalisation and liberal trade policy have been largely beneficial for all income groups and should be stimulated. On the opposite side of the argument are those who acknowledge the importance of economic growth, and the role economic openness can play in stimulating economic growth. However, they express their concerns about the effect of liberalization on income inequality and the consequent effect on the developing countries’ ability to reduce poverty and achieve an inclusive growth process (Stiglitz, 2002; Wade, 2004; Huang et al., 2010).

This paper ventures to examine how economic openness has affected the poor. Whether

increasing trade and FDI flows helped or hindered their economic development. This paper will

contribute to the literature in 4 ways. 1) It will use the recently improved World Bank inequality

and poverty databases, to consider the effect of openness on lower incomes. 2) Even though, a

fair number of studies consider effects of variables on either growth or inequality, this paper

considers the Poverty-Inequality-Growth nexus, and proposes a simultaneous examination of

growth and inequality to determine how lower incomes have benefited in absolute terms. An

approach only sparingly used in the literature (Belke & Wernet, 2015). 3) It will consider the

effect economic openness has had on relative inclusiveness of growth. As of yet, there is little

empirical cross-country evidence and much ambiguity how economic openness has affected the

poor relative to the rest of society (World Bank, 2013). 4) Special consideration is given for the

role of outward FDI, which is infrequently explored in the literature in conjunction with its

effect on the poor (Huang et al. 2010).

3 2. Literature Review

The literature review is organized as follows. Section 2.1 will examine the Growth- Inequality-Poverty nexus, as well trends of these 3 factors. Section 2.2 will extend the theoretical model and consider the interaction between growth and inequality and the importance of initial conditions. Section 2.3 will consider the concept of pro poor growth, and whether growth has been pro-poor in the past decades. Finally, Section 2.4-2.6 will examine the role of economic openness, with respect to economic growth inequality. After which the relative performance of economic openness on relative performance is considered in 2.7.

2.1 Economic Growth, Income Inequality and Poverty.

In order to examine how the poor have fared over the past decades it is useful to first consider trends in absolute poverty. Absolute poverty is generally measured by considering income information gathered by household surveys, against a fixed poverty line over time (Klasen, 2008). Appendix 1 show the headcount ratio at 1,90$ a day and 3,10$ a day. Clearly visible is a general trend towards poverty reduction in the past decades, despite significant regional differences.

The reduction in poverty is in large part due to economic growth in the developing world. After the Great Divergence, which persisted until the 1970s, a partial economic convergence has been observable. Appendix 2 shows the average economic growth rates of developing countries in various regions, and large developing economies. Some developing countries and regions have seen considerable economic growth and have, at least in part, been able to converge to developed countries. This has caused the relative economic gravity of the world to shift towards these developing countries (Quah, 2011). This strong spell of economic growth in developing countries has had considerable effects on absolute poverty reduction.

Empirical evidence shows that this trend of high economic growth in developing countries, has been the primary mechanism through which absolute poverty has been reduced over the past decades (Ravallion and Huppi, 1991, White and Anderson 2001; Kraay 2003; Dollar and Kraay 2016).

However, economic growth is not the only mechanism that affects absolute poverty.

Bourguignon (2004) introduced the Growth-Inequality-Poverty (GIP) Triangle. This simple

nexus states that reductions in absolute poverty can be fully determined by income growth and

Inequality. Before the introduction, of the GIP triangle and the consequent extensive empirical

work on the growth-inequality poverty relationship, distribution was largely dismissed by

authors as a relatively negligible contributor to poverty relief (e.g. Demery and Squire 1996;

4

Dollar and Kraay, 2002). However more recently authors have identified, that even though economic growth is the main poverty alleviator, distribution should not be dismissed so definitively (Lopez 2004; Ravallion and Chen, 2007). Many countries have been able to drastically reduce poverty through inequality reduction (Fosu, 2011). Lopez (2004) remarks that the distribution of income is the filter, through which income growth reduces poverty. In other words, in situations of high income inequality, economic growth will be relatively less poverty reducing. The development of inequality has mostly contrasted the developments observed in economic growth. Developing countries have seen a 11% within country inequality increase in the period 1990-2010 (UN, 2014)

. However, as with growth, changes have not been uniform across regions, with increasing inequality in East Asian developing countries, and decreasing inequality in Latin America (Lustig and Lopez-Calva, 2013).

2.2 Extending the GIP Triangle, the Inequality-Growth link and Initial Conditions.

Expanding this simple nexus, Lopez (2004) puts policy on the left hand side. Policy causes changes in either income inequality, income growth or both. After which, these two mechanisms in turn affect the absolute amount of poverty. Average income growth has a decreasing effect on the amount of absolute poverty, whilst an increase in income inequality will reduce poverty reduction.

2.2.1 The Relationship between Economic Growth and Income Inequality

The GIP triangle in figure 1 indicated that both income inequality and income growth reduce poverty, but it also hints at a tight relationship between income inequality and income growth. The effect of economic growth on income inequality boils down to essentially two opposing views. The first is based on the work of Kuznets (1955), which argues that there are

3 Within country inequality, adjusted for population size. Developing country based on UN definition.

Figure 1: Growth-Inequality-Poverty Triangle.

+ Poverty Reduction Distribution

Income growth

Note: Based on the Poverty-Inequality-Growth Triangle Bourguignon (2004), adapted by (Lopez, 2004) Policy

5

effects of structural change on income inequality. The general theory of the first view is that initial economic development is accompanied by an increase in income inequality, the so-called inverted U-curve hypothesis (Kuznets 1955). For developing countries this means, that whilst in the transition from an agricultural society to an industrial society, inequality increases. For a long time, this idea was taken for granted and is still a widely held as a stylized fact of growth.

However, in the past decades, multiple authors have found little evidence for the existence of the U-curve hypothesis (Ravallion, 1995; Deininger and Squire, 1998). Furthermore, the East Asian miracle showed, that there are instances in which economic development and income inequality reduction or stabilization can go hand in hand (Stiglitz, 1996). Indicating that income inequality is not an inevitable consequence of economic growth.

As for the effect of inequality on growth, there are again two opposing views. They can be described as the “classical view” and the “modern view” (Thorbecke, 2013). The classical view underpins the importance of accumulation of capital in accordance with the Solow model.

Large income differences indicate relatively large amounts of capital reside with the rich, who tend to save more (Kaldor, 1956). This capital accumulation increases investment and thus stimulates economic growth. On the other hand, the modern theories emphasize the importance of various social and political aspects, that are influenced by inequality. The modern theory states, that through these various channels income inequality can be detrimental to economic growth. These channels include: social and political unrest, increasing uncertainty, increasing transaction costs and reduced security of property rights in an unequal society (Thorbecke and Charumilind, 2002). Shin (2012) in an extensive empirical survey acknowledges that empirical evidence is decidedly split. Some authors have found decisive negative relationships between economic growth and inequality, (White and Anderson, 2001; Sukiassyan 2007), some found a positive relationship (Forbes 2000; Aghion and Howitt 1998), and some have not found any conclusive evidence of effects of income inequality on economic growth at all (Shin et al. 2009).

Recent empirical evidence has tried to explain the conflicting and inconsistent results in

the empirical literature. by differentiating between developed and developing countries. This

distinction has ensured more consistent results. The effect of inequality on economic growth in

developed economies is generally positive, whilst the effect of inequality on economic growth

in developing countries is generally negative (Castellò-Clement 2010; Ostry et al., 2014; Halter

et al. 2014).

6 2.2.2 Initial Conditions in Economic Growth.

Figure 2 shows a further extension by including initial condition of economic development, based on the work of Lopez (2004) and Nissanke & Thorbecke (2010). As mentioned previously it is to be expected that initial conditions of economic development will have an effect on income growth levels (Chen, 2008). Several authors have hypothesized that levels of economic development can affect economic growth, through conditional convergence, or the catch-up effect. Theoretically, catch-up might happen through decreasing returns to capital (Kuznets 1955), or through the advantages of economic backwardness (Gerschenkron, 1962). Indicating that an increase in initial income, on average, tends to negatively influence growth. The empirical evidence of convergence was initially mixed and it seemed that during much of the 20 century the developed countries in Western-Europe and European offshoots were able to maintain sustainable growth, whilst developing countries were struggling to generate economic growth (Frank, 2001). As noted previously there is significant evidence that developing countries, since the 1980’s, as a collective have been able to converge towards developed countries (Quah, 2011; Rodrik 2011). Though, it should be noted that the large contrast between developing countries, this statement does not necessarily hold true for countries individually, (David, 1997).

Figure 2. Initial conditions and the Interrelationship of Income Inequality and Income growth

Poverty Reduction Income

Inequality

Income growth

Initial Income Level

Kuznets +/-

Classical + Modern -

Note: Based on the Poverty Inequality Growth Triangle Bourguignon (2004), adapted by Thorbecke (2010) +

- Policy

7

2.3 The Extent of Poverty Reduction. What are Pro-Poor and Inclusive Growth?

In the PIG-Triangle inequality is used as a filter through which poverty reduction is achieved. However, it does not consider the relative benefits the poor reap from economic growth. The aforementioned contrary trends in developing countries of on the one hand high economic growth, and the other hand increasing income inequality, have been met with increasing concern by international organizations (World Bank 2013). In order to fill the gap in the literature concerning the benefits the poor can reap from growth, terms such as inclusive growth and pro-poor growth have been introduced. At its conception, pro-poor growth was broadly defined by international organizations as: growth that leads to a “significant” reduction in absolute poverty, (UN 2000; OECD, 2001). However, this definition of pro-poor growth was too broad for research. Since, as Kraay (2004) remarks, what is deemed to be a significant reduction? In the policy oriented discussion of pro-poor growth, the varying definitions can be essentially concentrated into two groups: absolute and relative pro-poor growth (OECD 2006;

Klasen 2008). The absolute definition states that growth is pro poor to the extent that it reduces poverty (Ravallion & Chen, 2002). After all, when income grows for the poor, it reduces poverty regardless of how income changes for other groups (Ravallion, 2004). On the other side are those that advocate the use of the relative measure of pro-poor growth. Relative pro-poor growth occurs when economic growth is accompanied by distributional shifts in favour of the poor (Kakwani & Son, 2008). In other words, growth is found to be relatively pro-poor if it disproportionately favours the lower incomes, in comparison to average economic growth.

Proponents of using the relative definition argue that simply measuring poverty reduction in a vacuum, is not enough to interpret whether growth has benefited the poor, it should be contrasted to the countries’ average performance (Kakwani et al., 2004). One should acknowledge that the term pro-poor in itself implies economic growth with special consideration of the lower income groups (Chen 2003).

As noted by Cord et al. (2003) neither are completely satisfactory in considering the benefits the poor reap from economic growth. Only examining the reduction of absolute poverty has considerable limitations. According to the absolute definition, pro-poor growth can be achieved, whilst there are minimal improvements for the poor. A scenario where average income growth would increase with 5%, whilst the increase of income of the poor was 1%, would still be deemed pro poor in an absolute sense. Empirical evidence demonstrates nearly every growth process to be pro-poor (White and Anderson, 2001; Ravallion & Chen, 2003;

Chen, 2013).

8

On the other hand, solely considering the relative definition can lead to blind spots as well. A simple relative measures will not show the possible benefits of economic growth for the poor. If low incomes grew relatively slower or at the same pace, compared to the national average, they are not considered to be pro-poor. Kraay (2004) notes that according to this definition, China’s growth process would not be considered pro-poor. Even though, the percentage of people living under the poverty line was reduced by more than 80%

in the period 1990-2010. This mismatch can put the focus too much on policy that invites the idea of inequality reduction, through government intervention and social spending, regardless of its effect on economic growth (Dollar et al., 2016). Deterring growth for more equality might actually hurt the poor, since poverty relief through average income growth might be a more welcome and beneficial policy for all income groups, than polices attempting to reduce inequality. The empirical estimations of relative pro poor growth, have more diverging results in the literature. A large portion of the literature has argued that growth has been relatively distribution neutral, (Ravallion and Chen, 1997; Dollar and Kraay, 2002; Dollar et al., 2016).

Dollar and Kraay (2002) find that the bottom 20% of average per capita income in developing countries has seen similar growth rates compared to the average and top income per capita growth. They conclude that the elasticity of the income of the bottom 20% with respect to mean income is statistically not significantly different from 1. Their methodology was questioned by several authors who indicated that the Dollar Kraay dataset was not robust (Ashley, 2008;

Deaton, 2005). Dollar et al. (2016), re-estimate their calculations with a more recent time period and a more robust data set, Dollar et al. 2016 find that even though growth is still good for the poor, evidence for an elasticity of 1 is weakened, the relative benefit of the poor seems lower than the average. This is corroborated by Chen (2013), who also indicate an increase in the percentage of relatively poor.

Beyond the scope of the extremes of relative and absolute pro-poor growth, several authors have attempted to unite these two measures of pro- poor growth into one measure that considers changes in both income growth and inequality, in one empirical estimation (e.g.

Ravallion and Datt 2000; Ali and Son 2007; Son and Kakwani 2008; Habito 2009). In the literature this is broadly described as inclusive growth, (Ranieri and Almeida Ramos, 2013).

However, as noted by White (2012), both conceptually and empirically definitions and measurements are very diverse and it seems that the term inclusive is used across the spectrum.

4 Povcalnet headcount ratio data. Percentage of people living under less 1,90$ (2011) a day.

9

From definitions close to relative pro-poor growth (UN, 2014), to definitions similar to absolute pro-poor growth, see (Habito, 2009).

As such, it might be more useful to consider the motivation behind the used method.

Klasen (2008) concludes, that there is no right definition of pro-poor growth or inclusive growth, but that different estimates pro-poor growth serves different purposes. Relative definition is suited when one wants to analyse what Klasen calls the state of pro-poor growth.

The state of pro-poor growth indicates how much the opportunities afforded by a given growth rate disproportionally helped or hindered the poor. This definition is more suited towards the purpose of this paper than absolute pro-poor growth, which mainly concerns poverty relief.

This is because trends of increasing inequality and economic openness raise questions about how the poor have benefited from this globalization period. For the purposes of this paper in accordance with sustainable development goal 10, this paper is more concerned with relative performance. As such it will use the relative definition of pro-poor growth.

2.3.1 Relative pro-poor growth and Inequality, two sides of the same coin?

Relative pro-poor growth in the definition of Kakwani and Son (2008) considers the income changes of a fixed proportion of the population relative to other groups, or to the mean income growth. Practically, this means that relative pro-poor growth is estimated using quintile income data, see (Dollar and Kraay 2002; 2004, White and Anderson 2001, Tsai & Huang, 2010). Since income share data is an alternative type of inequality data, relative pro-poor growth has a reasonable amount of common ground with inequality, often measured by the GINI index.

However, there is an important distinction between inequality and relative pro-poor growth.

Changes in the GINI take into account the entire distribution, whilst changes in relative poor growth consider income changes of a fixed bottom quantile, relative to other quantiles or the mean. The difference between bottom quantile growth and inequality is important, since the shape of the distribution affects income changes for the poor. Similarly, the UN’s Sustainable development goal 10 proposes to consider economic growth in the bottom 40% relative to average growth, as it better captures the effect of income changes on the poor. Furthermore, Voitchovsky, (2005), notes that policies are often associated with different parts of the income distribution, which cannot be properly captured by the GINI index. Accordingly, it is useful to consider more than just the GINI index in estimating relative performance of income groups (Cingano, 2014).

Relating this back to the earlier discussion of determinants of poverty, the remainder of

this paper considers a two-fold approach. First, in line with the work of Lopez (2004), Belke

10

and Wernet (2015) and the GIP Triangle, it will consider the effects of globalization on the two factors determining poverty: income growth and distribution. This helps us in understanding how changes in openness affect poverty, through either economic growth, inequality or both.

The second step focuses on the relative performance of the poor. It will consider relative pro- poor growth and in line with the work of: (Dollar and Kraay, 2002; 2004; 2016; Tsai & Huang, 2007; Huang et al., 2010; White and Anderson 2001), and consider how economic openness has affected lower income groups compared to the average.

2.4 Globalization and Economic Openness

Even though, there are numerous definitions of globalization, these definitions often revolve around a single theme, which is international integration. A commonly used definition of globalization by Albrow and King (1990) states that globalization encompasses the international integration and interchanging of culture, products and world views. Bordo et al.

(2007) identify 3 overarching economic factors that are influenced by globalization. These are:

the international openness of commodity, capital and labour markets. It could be argued that these factors could be complemented, by other aspects of integration, that arose during globalization. These include technology and knowledge transfers, due to the introduction of information technology (Thorbecke, 2010). However, for the purposes of the analysis in this paper, 2 of the factors identified by Bordo et al. (2007) are considered: The capital and commodity markets. More specifically Trade Openness and FDI will be used. These two factors have become standard in assessing economic openness, the extent to which economies are integrated in the world markets (Harrison 2007).

The rise of increased global economic integration has various sources. Efforts of

international organizations such as the WTO to reduce trade barriers in the form of tariffs and

non-tariff barriers have contributed to increased trade and FDI (World Bank, 2013). Other

factors such as improved communication and transportation technology, as well as relative shift

from central planning economies to market economies, have contributed to increased economic

integration (Gao, 2000). The period of globalization has significantly increased the volume of

trade, as well as amount and intensity of FDI (Bordo et al. 2007). Trade has increased

significantly, on all continents, since 1990, (World Bank, 2013). Inward FDI stock as a

percentage of GDP in developing countries has more than doubled since 1990, from 12,5% to

25,5% in 2011 (UNCTAD, 2016). Indicating that both Trade and FDI has not merely increased

in absolute terms, but have become a larger part of the respective national economies.

11

In order to decompose the effect of openness into a growth effect and an income inequality effect, the model of by Nisanke & Thorbecke (2010) integrates globalization and the resulting economic openness in the GIP triangle, figure 4. In the model by Nisanke and Thorbecke, economic openness has a positive effect on economic growth and a negative effect on income inequality. This paper will consider both the effect of FDI as well as trade openness on income growth and income inequality respectively. This gives us an indication whether economic openness affects poverty reduction through inequality and growth as described in figure 4. As noted previously the second step entails relative performance.

2.5 Trade Liberalization, Economic Growth and Income Inequality 2.5.1 Trade Openness and Economic Growth.

Trade can theoretically affect economic growth in two ways: productivity growth and capital accumulation (Andersen and Babula, 2008). Empirical evidence suggests that the effect of trade on capital accumulation is a relatively minor factor (Hall and Jones 1999), whilst the effects of trade on productivity gains are important towards economic growth (Frankel and Romer, 1999). Productivity growth can occur through efficiency gains and improvements in technology. Traditional trade models, such as the Ricardian and the Heckscher-Ohlin model predict efficiency gains from trade. Trade allows countries to have a more efficient resource allocation, which improves the level of productivity, and as such raises the level of national

Poverty Reduction Economic

Openness

Income Inequality

Globalization

Economic Openness:

Capital Trade

Technology Knowledge

Income growth -

+

Initial Income Level

Kuznets +/-

Classical + Modern -

Note: Based on the Poverty Inequality Growth Triangle (Bourguignon, 2004), The Openness-Growth- Distribution Nexus (Thorbecke, 2010)

Figure 4: Globalization Openness Growth distribution and Poverty

+ +

-

12

income. In considering the effects of trade it useful to distinguish between increases in the level of productivity also known as static gains, and factors that increase productivity on a long-term basis also known as dynamic gains. Economic development from traditional trade models comes in the form of static gains. This means that despite improving the level of national income, they do not increase the rate of economic growth (Lopez, 2005). Other static efficiency gains can occur through more efficient capital allocation. In the Melitz (2003) model of firm heterogeneity, efficient firms will export, whilst less efficient firms will not. As such, productive firms increase their relative production, compared to less productive firms as trade increases. This generates a relatively more effective allocation of capital, causing a level increase in productivity (Nordas et al. 2006).

There seems to be an unsatisfyingly low amount of empirically tested theoretical frameworks and models, that consider the exact channels through which trade openness affects long-run economic growth, i.e. provides dynamic gains (Walde and Wood 2004; Ulasan 2012).

Empirical evidence has shown that dynamic gains are primarily gained through technology spillovers (Nordas et al. 2006). The next section outlines the New-Growth school, one of the scarce theories concerning trade and economic growth.

The New Growth

school argues that technology is the key factor through which countries’ growth rates can benefit from trade. Technology spillovers may happen in a number of ways. Trade is likely to create spillovers, by transmitting knowledge and technology from developed to developing economies. This yields direct gains in productivity and as such in economic growth (Grossman and Helpman, 1991). However, there are indirect effects of trade as well. Imports, and the consequent technology spillovers, from developed economies to developing economies, guides developing countries on a path of innovation and imitation. This path is paved with significant externalities, through a process of ‘learning-to-learn’ (Connolly and Valderrama 2005). Through successful imitation and innovation in the past, developing countries become more successful in garnering knowledge from developed countries. This can increase the level of productivity growth. Another indirect effect of trade and the consequent spillovers is that if a country imports human capital intensive goods, trade reduces the derived demand for human capital. This causes the costs associated with innovation to decrease (Grossman and Helpman, 1991).

Besides technology transfers, there are other dynamic gains to be made from increasing one’s involvement in trade. Firstly, trade can influence domestic macroeconomic stability.

5 For a more extensive discussion concerning the new growth school please refer to (Lopez 2005; Andersen Babula 2008).

13

Government policy might be influenced by trade policy. International agreements can put implicit or explicit pressure on governments to engage in policies that strengthen macroeconomic stability. These macroeconomic policies are likely to favourably affect growth by: reducing public deficit, debt levels, uncertainty, encourage price stability, and it can allow domestic firms to be more competitive in the international markets (Fischer, 1993). A second way in which trade openness might influence economic growth, is in regard to the acquisition of capital goods. Trade openness might increase the ability of domestic agents to import previously unavailable capital goods, removing restraints on investment, increasing capital accumulation. These imported capital goods, also tend to embody more recent technologies, thus improving productivity (Wacziarg, 2001). These theories are particularly applicable to developing countries, where the knowledge and technology gap to the developed economies is considerable and the import of human capital intensive goods is relatively high.

The cross-country empirical literature surrounding trade openness and growth is large in size, but also full of pitfalls. During the 90’s a large body of work found positive effects of trade openness on economic growth (Edwards, 1993;1998; Dollar,1992; Sachs and Warner, 1995). However, this literature suffered from several shortcomings, concerning estimation techniques. Rodriquez and Rodrik, (2001) found that the results of many previous estimations did not hold up under further scrutiny, especially concerning issues of unaddressed endogeneity.

The wider availability of more advanced econometric techniques, including instrumental variable approaches, has allowed for more reliable estimations. A second concern of the estimation of trade openness has been that trade openness can be measured through a rather wide variety of variables. Appendix 3 shows a selective survey of literature. Singh (2010) notes that due to data availability in cross country analyses, trade openness is generally measured either by the volume of trade

or by tariffs. Trade openness through either, volume of trade (Frankel and Rose 2002, Dollar and Kraay 2003; 2004; Lee et al, 2004), or through the reduction of tariffs (Clemens and Williamson, 2004; Lee et al. 2004), have yielded positive effects on economic growth.

2.5.2 Trade Openness and Inequality

Trade has not merely a growth effect, it influences inequality as well. The Stolper- Samuelson extension of the Heckscher-Ohlin model predicts that trade influences factor rewards, such as real wages and returns to capital. The Heckscher-Ohlin model argues labour abundant (developing) economies will increase their production in labour abundant products,

6 Volume of trade is generally measured by: (𝐸𝑥𝑝𝑜𝑟𝑡 + 𝐼𝑚𝑝𝑜𝑟𝑡)/𝐺𝐷𝑃

14

for export. On the other hand, it will increase imports of capital abundant products from capital abundant (developed) economies. If the two factors are skilled labour intensive (capital intensive goods) and unskilled labour intensive goods (labour intensive), assuming skilled labour to earn higher wages, developed countries should see an increase in inequality. This is because relative demand for capital intensive goods compared to labour increases. Whilst the opposite is true in developing countries, leading to a decrease inequality (Harrison et al. 2011).

However, due to the lack of empirical evidence concerning the basic Heckscher Ohlin model, as well as the Stolper-Samuelsen theorem, the proposed effect of factor abundance, skill intensity and inequality are in doubt as well (Davis and Mirsha, 2007; Helpman et al., 2016).

Indeed, empirical evidence tends to indicate that trade openness exacerbates inequality.

The aggravation of income inequality has been supported by several panel data analyses as well as individual country studies, see appendix 3 for a selective overview of the literature. The theories that explain this exacerbation are diverse. It can occur through a number of mechanisms, such as changes in bargaining power, or labour market frictions, these factors are caused or intensified by trade (Harrison et al. 2010). The next sections will give a brief overview of these various channels.

Perhaps the most considerable body of literature that considers the effect of trade on inequality, are the changes in wages that occur through heterogeneous firms in the Melitz model. Besides economic growth the Melitz model also has implications for income inequality.

The first approach considers how effective firms will expand their production, as they have to serve the foreign market conjointly with the domestic market. In order to do so they will increase the number of employees at the productive firms. These productive firms pay higher wages, leading to a relative shift of weight to higher wage labour. This guarantees that with an increase in trade, the averages wages will increase compared to the wages of those who work at less productive firms, leading to increased inequality (Helpman et al. 2010). The second approach considers the effect of trade further exacerbating inequality through frictions and bargaining in the labour market. Helpman et al. (2010) theorize that more productive firms will screen more diligently compared to less productive firms. This occurs because more productive firms pay higher wages and expect a higher contribution of production per labourer they hire.

As such, more stringent scrutiny and higher thresholds are worthwhile. Only a relatively

selective group of potential employees will be considered for the productive firms. This

increases the respective potential employees’ bargaining power, which in turn increases wages

at productive firms, relative to the average. Trade exacerbates this effect, as in the Melitz model

less productive firms, which do not export incur a reduction in sales, whilst more productive

15

firms do export and increase their sales. The previously mentioned diligent screening process becomes more worthwhile, further increasing wages at productive firms, whilst the screening process in less productive firms, who see their sales reduced, becomes less worthwhile, which decreases wages at less productive firms (Helpman et al. 2010). The final way through which trade can increase inequality has largely to do with previous approach. As trade increases, sales of unproductive firms will relatively decrease compared to productive firms. These productive firms can produce with relatively less labour, leading to an increase in unemployment. This will increase the fraction of workers receiving no income, which consequently leads to higher inequality (Egger & Kreickemeier 2009). Empirical evidence strongly supports the notion that trade openness has increased wages at globalized firms, whilst trade openness has decreased wages in firms that only serve the domestic markets (Amitit & Davis 2008; Helpman et al.

2016). This supports the theories discussed in this section, that trade openness can cause a relative shift of wages through heterogeneous firms.

The second way in which trade can increase inequality is through offshoring. Naturally

offshoring has significantly increased the amount of trade and is largely facilitated by trade

policies trade agreements. Feenstra & Hanson (1996) argue that offshoring can cause increased

inequality in both developing and developed countries. Following a Heckscher-Ohlin model,

the propose a model with two countries, one of which is skill abundant and one of which is low-

skill abundant. Headquarters of the firm are positioned in skill abundant country. If off-shoring

becomes easier due to decreasing trade costs, the range of tasks performed in the skill-poor

abundant country will increase, by the logic of cost minimization. The tasks that will be

outsourced will be relatively low in skill intensity This ensures that tasks in skill-abundant

countries become more skill intensive. This ensures a skill-premium in the skill abundant

country, which causes relative wages for high skilled labour to increase, causing an increase in

inequality in the skill abundant country. However, increased range of tasks that are outsourced

by skill abundant countries are in fact relatively more skill intensive for the skill-poor country

than the range of tasks previously performed. As such, a similar trend will occur in skill-poor

countries. The skill premium will increase and inequality will increase. Empirical approaches

of this model are abundant and considerably different approaches are considered throughout the

literature, in an extensive discussion of the empirical evidence (Harisson et al., 2011) conclude

that there is extensive empirical evidence that supports the Fenestra & Hanson model.

16

2.6 Foreign Direct Investment, Economic Growth and Income Inequality.

2.6.1 FDI and Economic Growth

Inward-FDI can influence economic growth in essentially two ways. First, it may simply increase the available capital. This contribution to capital accumulation, will increase the aggregate investment in the host country, providing long-term economic growth (Borensztein et al. 1998). Secondly, in accordance to endogenous growth theory, knowledge and technology spillovers may increase productivity and thus economic growth (Clark et al. 2011).

Technological spillovers may happen through a number of channels. Firstly, labour mobility can stimulate horizontal spillovers by transferring knowledge from a foreign Multinational Company (MNC) to a domestic firm (Fosfuri et al. 2001). It should be noted that this effect may be limited, due to counter measures by foreign firms, trying to prevent this outflow of knowledge. For example, by offering higher wages (Sinani & Meyer, 2004). Another possibility for horizontal spillovers is the imitation effect, whereby foreign firms are able to demonstrate to the host countries’ firms that management techniques or expensive technologies are worth utilizing, also known as the “demonstration effect” (Giovannetti and Ricchiuti, 2005).

Not only does FDI affect horizontal spillovers, backward and forward linkages can provide ample possibilities for spillovers as well. MNCs engaging in FDI, might require their suppliers in the host country to become more productive or adhere to certain standards. This support may come through: training; technological support or capital. Secondly, foreign entry might encourage competition, forcing domestic firms to use resources more effectively and to use more productive technologies and inputs (Rodriquez-Clare, 1996).

In Accordance with endogenous growth theory, the extent to which countries are able to benefit in terms of the accumulation of capital and technology growth, the so-called absorptive capacity of a country, is dependent on a countries’ home characteristics, (de Mello, 1999). Absorptive capacity can depend on a number of factor. There is significant evidence that the FDI can benefit economic growth, if the human capital is in place (Borensztein et al., 1998;

Ram & Zhang, 2002). Others have found that the development of financial markets, Alfaro et al. (2004), or trade openness (Durham, 2004), have an important contribution in allowing FDI to cause the desired effect on economic growth, see appendix 4 for a selective overview of the literature)

An important, but somewhat less explored aspect of openness is the effect of outward FDI. developing countries have started to play a much more significant role in providing outward FDI as well. Theoretically, FDI outflows could harm and help the domestic economy.

Stevens & Lipsey (1992) identify two ways relevant for developing economies, through which

17

FDI outflows could affect domestic productivity. First, an outflow of capital means a loss of savings in the home country. Due to imperfect financial markets this outflow of capital from the home country cannot be substituted. Less capital accumulation means lower investments, causing a reduction in economic growth. However, on the other hand, firms that invested abroad, might be seeking efficiency gains through vertical FDI, this can improve the competitive position of the domestic firms through cost reductions. Outward FDI may also cause gains in the level of technology, if new technologies or knowledge can be acquired abroad, which could benefit the home economy. The available empirical evidence that researches outward FDI and economic growth, generally finds a positive relationship between outward FDI on domestic economic growth, see (Huang et al. 2010; Herzer, 2010).

2.6.2 FDI and Inequality

Theoretically there are two channels through which FDI inflows can decrease income inequality. The first is the possibility of FDI to utilize large amounts of labour. Theoretically, if enough low income low-skilled labour is utilized, FDI can reduce income inequality, (Deardorff & Stern 1994). The second channel through which FDI inflow can reduce inequality is the FDI increases the amount of capital available. Larger amounts of capital would decrease the return to capital and increase the return to labour, (Clark et al. 2011). Though there is some evidence that suggests the inflow of FDI decreases inequality, see (Milanovic, 2002), most empirical evidence indicates that increasing FDI inflows actually generate higher income inequality (Choi 2004; 2006; Tsai & Huang 2007; Huang et al. 2010).

There are several theoretical channels through which FDI can increase inequality.

Firstly, there is significant evidence of a skill-bias of FDI in developing countries. Relatively higher skilled labourers benefit more from the inflow of FDI, increasing the wage gap between skilled and unskilled labour (Gopinath and Chen, 2003; Basu Guariglia 2007 Figini 2006).

Wages are the primary aspect of income, especially for lower incomes, unequal wage growth

can lead to increasing income inequality (Figini & Görg, 2011). The second factor that needs

to be considered is the sectoral imbalance that FDI can create (Tsai 1995). Only certain sectors

are stimulated by the foreign investments. These sectors will see improvements in wages, whilst

other sectors do not. The large income differences between rural and urban areas in China, is

partially stimulated by foreign investments. These investments have brought economic gains in

urban areas, but not in rural areas This has caused a significant wage gap between those in rural

and urban areas, and has contributed to rising inequality in China (Wei et al. 2008).

18

Theoretically there are also reasons why the outflow of FDI might affect inequality. As opposed to inflow of FDI, an outflow of capital should increase the returns to capital, and diminish the relative returns to labour. This would increase inequality. Empirically, evidence indeed suggests, that indeed FDI outflow decreases economic growth for lower incomes, (Tsai and Huang, 2007; Huang et al. 2010). They find outward FDI to negatively affect the poor, both in absolute and relative terms. On top of that Calderon (2014) outward FDI can increase inequality through asymmetric tax burdens, as outward flows of FDI might occur through tax havens, which increases inequality.

2.7 The Effect of Economic Openness on Quantile Growth and Relative Pro-Poor Growth As noted previously, the definition of relative pro-poor growth used in this paper, considers inequality in the bottom part of the distribution compared to the average. This means that theoretical mechanisms behind the effect of trade and FDI on relative pro-poor growth are essentially similar to the mechanisms behind the effect of trade and FDI on inequality. There is no indication from the literature that these mechanisms have affected bottom quantiles differently. As, such only further consideration will be given towards empirical evidence. The empirical literature concerning economic openness and bottom quantile growth is rather narrow. Tsai and Huang (2010) uses a level analysis of average income in the bottom quantile in a 2SLS model. Concerning trade openness, measured by trade volume, they find an absolute positive effect average bottom quintile income. However, they find distinct negative effects of both inward and outward FDI on absolute levels of income. This is contrasted by others who, using similar instrumental variable approaches and found significant positive effect of FDI and trade on absolute bottom quantile growth (Dollar and Kraay; 2003; Otsubo & Hirano, 2016).

Earlier models investigating the relative effect of trade and FDI on economic growth,

have considerable methodological problems. White and Anderson (2001) use a simple OLS

estimation and find no effect of trade, as well as an insignificant effect of FDI on the bottom

quantile relative to the average. As mentioned previously these type of estimations were

plagued by endogeneity, which is not accounted for by OLS or fixed effects estimators. Dollar

and Kraay (2003; 2004) are some of the first to consider relative performance by comparing

bottom quintile growth to other quantiles, using a 2SLS model. As noted previously the Dollar

and Kraay estimations were met with considerable criticism, see (Deaton, 2005). More recent

empirical estimations find trends of FDI to be similar to the effects of FDI on inequality as

estimated by the GINI. Otsubo & Hirano (2016) use a 2SLS model and find that bottom

quintiles do not benefit from FDI, whilst the highest two quantiles has tended to be positively

19

affected by FDI. This evidence is supported by (Im and Mcaleren, 2015), who find that FDI relatively benefits higher incomes compared to lower incomes.

Otsoubo & Hirona also estimate trade openness through trade volume. However, they find no differences in quantiles wen regressing trade openness, both benefit equally from trade.

It should be noted that others have found a relative negative effect of trade openness on economic growth (Jeanneney and Kpodar, 2011; Singh and Huang 2011) using similar instrumental variable approaches.

Table 1. Hypotheses concerning the effect of Economic Openness on Economic Growth, Income Inequality and Relative pro-poor growth

 H1: Trade openness has a positive effect on economic growth

 H2: Trade Openness will increase Income Inequality.

 H3a: Inward FDI have appositive effect on economic growth

 H3b: Outward FDI have a positive effect on economic growth

 H4a: Inward FDI will increase Income Inequality.

 H4b: Outward FDI will increase Income Inequality.

 H5a: Trade Openness will have a relative negative impact on the poor, increasing the difference in relative economic growth between the lower incomes and average income growth.

 H5b: Inward FDI will have a relative negative impact on the poor, increasing the difference in relative economic growth between the lower incomes and average income growth.

 H5c: Outward FDI will have a relative negative impact on the poor, increasing the difference in relative economic growth between the lower incomes and average income growth.

3. Methodology and Data

The panel in this paper assesses a sample of 61 developing countries, with annual observations for the period 1980-2011. For a full list of countries please refer to Appendix 5.

The length of the panel is primarily based on data availability; World Bank data only extends

back to 1980. Mainly through the relative scarcity of inequality data, only a relatively select

number of time periods and countries are considered. Taking a cursory glance at the data, it

reveals skewness in a number of ways. First, Geographical skewness, the majority of

observations comes from the World Bank classified regions of: Europe and Central Asia, Latin

America and East Asia and Pacific. Only a small minority of observations are available from

Africa and the Middle East. Secondly, observations tend to be skewed to more recent years.

20

The final type of imbalance implicates that the data is slightly skewed towards countries that are currently more developed.

3.1 Data & Main variables 3.1.1. Dependent variables

Following the model by Bourguignon (2004) and the literature review thus far, three main points of interest were reviewed: economic growth, inequality and relative pro-poor growth. Economic growth will simply be measured by GDP per capita growth, retrieved from the Penn World Tables (Timmer et al., 2014). Inequality will be measured by the GINI coefficient retrieved from the WDI (World Development Indicators) of the World Bank. The third point of interest, relative pro-poor growth, will be measured through two variables, following the method of White and Anderson (2001). Firstly, the variable SH20 (or SH40), by:

Y

− Y

Y

− Y

(1)

The variable calculated in 1 denotes the she shares of incremental income received, by the poorest 20% or 40% respectively. Where, Y

indicates income of the respective quantile, and 𝑌

indicates, the average income. The quintile data is retrieved from the WDI database.

Inequality measures over time and comparisons across countries have long been cumbersome, due to different definitions between countries and changes in definition over time. However, WDI has made significant improvements in comparability (Latner and Smeeding, 2015). The second dependent variable of relative pro-poor growth used in this paper estimates the change in the share of income DQ20 (DQ40):

Y

Y

− Y

Y

(2)

As noted by White and Anderson (2001), there is significant correlation between the dependent variables, SH and DQ

. The SH variable has an unfortunate characteristic, that it generates large variables if growth is very small. As such several outliers of SH-20 (40) have to be removed from the sample. For a comprehensive overview and detailed definition of the variables please refer to Appendix 6.

7 In the sample presented in this paper r=0.72 for the SH20 and DQ20, and r=0.68 for the DQ40 and SH40 dependent variables.

21 3.1.2 Economic Openness variables

This paper measures 3 variables that concern economic openness. As alluded to in the literature discussion, measures of trade openness are diverse. David (2007) distinguishes between essentially 6 different measures of trade openness

. The most widely used variable measuring trade openness in cross-country analyses is a simple trade ratio. This measure has some attractive properties, but one must also be aware of its limitations. David (2007) notes that its wide availability, especially in data scarce developing countries, has made it popular proxy for trade openness. However, country specific effects make it incomparable between countries, if not carefully used. Furthermore, since it is an outcome measure reliant on many factors, there is little evidence whether it is strongly related to actual trade policies. As such, it is important to be aware of the limitations of the conclusions that trade data can provide towards openness policies. Despite these shortcomings, data restraints

, as well as common usage in the literature (see appendix 3) ensures that out of the 6 options presented by David only trade ratios are a viable option for the type of analysis performed in this paper.

The effect of FDI will be measured through FDI flows. This paper measures de facto Capital account Openness. Inward FDI is measured through FDI net inflows as a percentage of GDP and outward FDI is measured by outward FDI, through FDI net outflows as a percentage of GDP. The choice of de facto measure of account openness is motivated by availability and prevalence in the literature (see appendix 4). Both inward FDI, as well as outward FDI, are widely available for a vast number of countries with yearly updates, from World Bank databases.

3.2 Estimating The effect of Economic Openness on Economic Growth and Income Inequality.

Similar to the literature review, we first distinguish between the determinants of poverty:

economic growth and Inequality, after which relative pro-poor growth is explored (section 3.4).

In order to measure the effect of economic openness on both inequality and growth on we use the methodology of Lopez (2004) and GIP triangle. In econometric form this looks like:

𝑃

= 𝑃(𝑌, 𝐺) (1)

8 1) Trade ratios; 2) Adjusted trade flows; 3) Price-based; 4) Tariffs; 5) Non-tariff barriers; 6) Composite Indices.

9 Average Tariff rates were considered in assessing openness to trade. However, deficient data availability reduced the number of observations significantly.

22

Where 𝑃

the poverty reduction, is measured income per capita Y and G, the Gini coefficient.

We can dissect, poverty into the effect of income growth and the effect of redistribution, like this:

𝜕𝑃

𝜕𝑋 𝑋 𝑃 = 𝜕𝑌

𝜕𝑋 𝑋

𝑌 × 𝛾 + 𝜕𝐺

𝜕𝑋 𝑋

𝑃𝐺 + 𝜑 (2)

The effect of an economic policy measure X, the depends on 4 effects. 1) the effect of economic activity on growth, 2) On the poverty elasticity of growth, 𝛾 3) on the change in income distribution at the same time 4) on the poverty elasticity of the distribution effect 𝜑.

The elasticities of poverty and distribution 𝛾 𝑎𝑛𝑑 𝜑 can be seen as independent from the policy measure, but they naturally depend on the present conditions. The present condition of initial inequality and conditions of initial GDP per capita levels.

Empirically the effect of globalization on growth and income inequality is estimated by the following equations:

GROWTH

= 𝛽

GROWTH

+ 𝛽

𝐺𝐼𝑁𝐼

+ 𝛽

𝐹𝐷𝐼_𝐼𝑁𝐹𝐿𝑂𝑊

+ 𝛽

𝐹𝐷𝐼_𝑂𝑈𝑇𝐹𝐿𝑂𝑊

+ 𝛽

𝑇𝑅𝐴𝐷𝐸

+ 𝛽

𝑋

+ 𝛽

𝐺𝐷𝑃 + 𝛼

+ 𝜂

+ 𝜀

(3)

𝐺𝐼𝑁𝐼

= 𝛽

+ 𝛽

GINI

+ 𝛽

𝐹𝐷𝐼_𝐼𝑁𝐹𝐿𝑂𝑊

+ 𝛽

𝐹𝐷𝐼_𝑂𝑈𝑇𝐹𝐿𝑂𝑊

+ 𝛽

𝑇𝑅𝐴𝐷𝐸

+ 𝛽

𝑋

+ 𝛽

𝐺𝐷𝑃 + 𝛼

+ 𝜂

+ 𝜀

(4)

With X indicating the control variables, which will be discussed in section 3.5, and i denoting the country and t denoting the time. The equation with GINI as a dependent variable is similarly estimated to economic growth. The models incorporate Inequality and the 3 determinants of Economic Openness. 𝛼

are the time-invariant country specific effects and 𝜀

the country specific error term. The initial level of GDP, is also integrated, as it was found in the literature review that convergence ensures that countries with higher incomes will have generally slower growth. For income, economic growth is used as a dependent variable, whilst concerning inequality the level of the GINI is used. This is in accordance with the estimations of Lopez and Serven (2006) and Belke and Wernet (2015) and the literature discussion presented in this paper. Inequality is the filter through which growth affects poverty. As such, the main concern of this paper is not with change in inequality, but the level of inequality.

A second important aspect is that the lagged variant of the dependent variable is used as

an independent variable. Taking into account the effect of previous years’ economic growth

23

and level of inequality, causes the estimations in 3 and 4 to become dynamic. This type of model is theoretically justified for Inequality and growth, since income

and distribution are deemed to be persistent over time (Shen and Yao, 2008). This is mainly due to the structural circumstances, within the economy such as institutions and factor endowments, (Boix, 2010), which cause current GINI and Growth to be reasonable good predictors of future Inequality and economic growth.

3.3 Fist differences and system GMM

In a dynamic panel model with a large number of countries there are generally two main concerns (Lopez, 2004). The first aspect is the endogeneity issue, which inevitably shows its head in dynamic panel models (Han and Phillips, 2010). Endogeneity in dynamic models indicates that the lagged variant of the dependent variable, used as an explanatory variable, is correlated with the error term. The second concern is that using limited variables to estimate economic growth or inequality provides inadequate explanatory power. This is chiefly due to the large number of country specific variables that influence growth and inequality, and are not properly taken into account. Factors such as being land-locked, language, whether it is a former colony etc. Several authors that results have observed that estimation results of cross country panel equations such as those presented in (3;4), can be inaccurate if unmeasured country differences are not properly taken into account (Amann et al., 2006; Ravallion, 2001).

First there is the need to resolve the highly likely possible endogeneity

issue. On top of this concern in a dynamic panel, the literature reviews also revealed possibilities of reverse causality between FDI and trade, and GDP, adding to the concerns of endogeneity issues. In this case neither fixed effects, nor ordinary least (OLS) estimators are a good fit. In such cases instrumental variables (IV) are a considerably better option (Anderson and Hsiao, 1981).

However, the dataset presented in this paper has a number of characteristics that lend itself better to a special type of IV approach, the Arellano Bond (1991) GMM estimator. The Arrellano Bond GMM estimators hinges on the idea of moment conditions. The idea behind using the moments conditions is that instead of using the theoretical expected value, it is replaced by the empirical variant. The deviation of the perfect sample variant of the moments conditions, are captured by the so called cost-function, the sum of the cost function of the

10 It must be noted that income growth is more volatile than income per capita. As such there is a larger chance that economic growth is not deemed to be persistent over time. However, there is still significant evidence of correlation between growth in previous years and growth in time t (Shen and Yao, 2008).

11 Results of the Durbin Watson test indeed show a high likelihood of autocorrelation for the GROWTH and GINI dependent variables. See table 2 and 3 respectively.